festival/Makefile0000644000175000017500000001055607614354707012463 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996-2002 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## The Festival Speech Synthesis System ## ## ## ## Authors: Alan W Black, Paul Taylor, Richard Caley and others ## ## Date: January 2003 ## ## ## ########################################################################### TOP=. DIRNAME=. BUILD_DIRS = src lib examples bin doc ALL_DIRS=config $(BUILD_DIRS) testsuite CONFIG=configure configure.in config.sub config.guess \ missing install-sh mkinstalldirs FILES = Makefile README ACKNOWLEDGMENTS NEWS COPYING INSTALL $(CONFIG) VERSION=$(PROJECT_VERSION) LOCAL_CLEAN= Templates.DB ALL = .config_error $(BUILD_DIRS) # Try and say if config hasn't been created config_dummy := $(shell test -f config/config || ( echo '*** '; echo '*** Making default config file ***'; echo '*** '; ./configure; ) >&2) # force a check on the system file system_dummy := $(shell $(MAKE) -C $(TOP)/config -f make_system.mak TOP=.. system.mak) include $(TOP)/config/common_make_rules backup: time-stamp @ $(RM) -f $(TOP)/FileList @ $(MAKE) file-list @ sed 's/^\.\///' .file-list-all @ (cd ..; tar cvf - `cat festival/.file-list-all` festival/.time-stamp | gzip > festival/festival-$(VERSION)-$(PROJECT_STATE).tar.gz ) @ $(RM) -f $(TOP)/.file-list-all @ ls -l festival-$(VERSION)-$(PROJECT_STATE).tar.gz time-stamp : @ echo festival $(VERSION) >.time-stamp # Too many randomly different unices out there .... # @ echo `id -u -n`@`hostname`.`domainname` >>.time-stamp @ date >>.time-stamp test: @ $(MAKE) --no-print-directory -C testsuite test config/config: config/config.in config.status ./config.status configure: configure.in autoconf include $(EST)/config/rules/top_level.mak include $(EST)/config/rules/install.mak festival/README0000644000175000017500000000442712432155242011666 0ustar awbawb The Festival Speech Synthesis System version 2.4 December 2014 This directory contains the Festival Speech Synthesis System, developed at CSTR, University of Edinburgh. The project was originally started by Alan W Black and Paul Taylor but many others have been involved (see ACKNOWLEDGEMENTS file for full list). Festival offers a general framework for building speech synthesis systems as well as including examples of various modules. As a whole it offers full text to speech through a number APIs: from shell level, though a Scheme command interpreter, as a C++ library, and an Emacs interface. Festival is multi-lingual (currently English (US and UK) and Spanish are distributed but a host of other voices have been developed by others) though English is the most advanced. The system is written in C++ and uses the Edinburgh Speech Tools for low level architecture and has a Scheme (SIOD) based command interpreter for control. Documentation is given in the FSF texinfo format which can generate, a printed manual, info files and HTML. COPYING Festival is free. Earlier versions were restricted to non-commercial use but we have now relaxed those conditions. The licence is an X11 style licence thus it can be incorporated in commercial products and free source products without restriction. See COPYING for the actual details. INSTALL Festival should run on any standard Unix platform. It has already run on Solaris, SunOS, Linux and FreeBSD. It requires a C++ compiler (GCC 2.7.2, 2.8.1, 2.95.[123], 3.2.3 3.3.2 RedHat "gcc-2.96", gcc 3.3, gcc 4.4.x and gcc-4.5.x are our standard compilers) to install. A port to Windows XP/NT/95/98 and 2000 using either Cygnus GNUWIN32, this is still new but many people are successfully using it. A detailed description of installation and requirements for the whole system is given in the file INSTALL read that for details. NEWS Keep abreast of Festival News by regularly checking the Festival homepage http://www.cstr.ed.ac.uk/projects/festival/ or the US site http://festvox.org/festival/ New in Festival 2.2 updates to hts (hts_engine 1.07) and clustergen New in Festival 2.1 Support for various new GCC compilers Improved support for hts, clustergen, clunits and multisyn voices lots of wee bugs fixed festival/ACKNOWLEDGMENTS0000644000175000017500000000703212432155373013156 0ustar awbawbFestival is currently actively developed by: Alan W Black (Carnegie Mellon University) Rob Clark (Edinburgh University) Junichi Yamagishi (Edinburgh University) Keiichiro Oura (Nagoya Institute of Technology) The following people and organisations have contributed to the development of Festival in various ways. It is their work that makes it all possible. Alan W Black Overall design, most of the front end and software control Paul Taylor Overall design, most of the back end Richard Caley for doing lots of difficult and boring bits Rob Clark Intonation, multisyn voice building, general developement and maintenance. Keiichiro Oura Updated HTS engine and API Junichi Yamagishi HTS voices Korin Richmond Multisyn engine, swig wrappers and general developement. Heiga Zen HTS engine Brian Foley Mac OSX support Kevin Lenzo for speaking a bunch of different nonsense words, design and improvements to the clunits module, and co-author of the whole festvox project Alistair Conkie various low level code points and some design work Spanish synthesis, recording Roger Steve Isard design of diphone schema, LPC diphone code, and directorship EPSRC who funded awb and pault Carnegie Mellon University who fund awb David Huggins Daines (Cepstral, LLC) configure, and lots of Linux associated bugs Sun Microsystems Laboratories For believing in us and their generosity. AT&T Research Labs For providing funding and using our work Paradigm Assoc. and George Carrett For Scheme In One Defun CNET, France Telecom for use of Donovan diphones and some code in modules/donovan (used with permission) The beta testers Thanks for wanting to use the system, you make it worth doing. (And thanks for helping me debug my code.) You all responded to my requests fast and accurately thanks, even when I dumped last minute changes on you Andy Donovan for speaking a bunch of nonsense words Roger Burroughes for speaking another bunch of nonsense words Kurt Dusterhoff for speaking another bunch of nonsense words Amy Isard for her SSML project and related synthesizer Mike Macon for signal processing advice Richard Tobin for answering all those difficult questions, and the socket code, and rxp the XML parser Simmule Turner and Rich Salz command line editor: editline Borja Etxebarria For Spanish synthesis and answer signal processing questions Briony Williams Welsh synthesis Jacques H. de Villiers from CSLU at OGI, for the TCL interface. ATR and Nick Campbell for first allowing Paul and Alan to work together Oxford Text Archive For the computer users version of Oxford Advanced Learners' Dictionary redistributed with permission Reading University for access to MARSEC from which the phrase break model was trained. Mari Ostendorf For giving access to the FM Radio Corpus from which some models were trained. LDC & Penn Tree Bank from which the POS tagger was trained, redistribution of the models is with permission from the LDC. Grady Ward for the MOBY pronunciation lexicon FSF for G++, make, .... and others too. festival/NEWS0000644000175000017500000002053207304555116011506 0ustar awbawb Note that not all features discussed in this file are included in the standard distribution. HISTORY June 21st 2001 1.4.2 Release Various new gcc's support Visual C++ 6.0 support uses configure (though could do so even more) substantial updates to the clunits unit selection module lots of wee bugs fixed a few very hard bugs fixed (client/server race condition) (dropped bytes in reading files when machine overloaded) (FreeBSD memory/gc problem) default waverform type is now RIFF. Nov 21st 1999 1.4.1 Release SSFF (for emulabel) track support AIX support Java fixes various minor bug fixes WFST with proper quoting Wagon sample counts gcc-2.95.1 support June 20th 1999 1.4.0 Release becomes free software June 6th 1999 1.3.95 Beta size/speed/memory leak overhaul (no memory leaks) XML support for relation loading (for SOLE support) JSAPI initial support GalaxyCommunicator architecture interface ked_mttilt_diphone voice built Parser trained on MARSEC (prosodic) brackets rather than syntax Unisyn_selection fully integrated Unisyn_phonology fully integrated viterbi cart/ngram/wfst base LTS prediction (did improve but BIG) viterbi cart/ngram based accent prediction (didn't improve) tilt working (again) audioin (na_record) for many architectures viterbi from Scheme (with cart, ngram, wfst models) January 26th 1999 1.3.1 Release egcs-1.1.1 support tobi_rules update (GM) replace readline with editline (+ extensions) Lots of little bug fixes cluster code tidied up kal voice ked power normalization updated lexicons with addenda for US and UK New LTS models for US and UK English "Building Voices in Festival" document August 24th 1998 1.3.0 Release UniSun/groupfile optimizations Java client support Fixed ESPS so both track and wave output works Retraining of most modules with new architecture (durations improved) rxp, (Richard's XML parser) integrated and Sable XMLified Fringe display program for labels and utterances Metrical tree synthesis A new utterance architecture (Relations and Items) utterance save and load work properly now Trainable LTS system Lexicon cache system Substantial optimization of front end (twice the speed) UniSyn, new signal processing and generic waveform synthesis module OLS code added WFST support for kk rules, regular grammars etc, simple English morphology November 30th 1997 1.2.4 BETA Tilt analysis and Tilt intonation modules added. make_utts substantially improved (> 100 times faster) text2wave script added Pitch synchronous lpc analysis and support rab consonant clusters labelled New duration tree (wagon stepwise) much smaller if not better SCFG grammar and parser (scfg_parse_text added as festival script) change config stuff (again) October 1st 1997 1.2.1 RELEASE preliminary support for Visual C++ Use path-append rather than string-append (in buckets of places) Minor bugs fixes throughout the code (end silences are now *always* inserted in tts) Linux socket bug fixed (get_url didn't work) native irix audio support September 5th 1997 1.2.0 RELEASE Proclaim modules and voices automatic detection of voices Phonset, lexicon, ltsrules listing and printing 16 bit linear native support for Solaris i386 (sb16) Update Festival Tutorial to 1.2.0 Aygust 15th 1997 1.1.99 Beta release Win NT (and 95) initial support Cygnus win32 and Visual C++ 100 more pages of documentation LPC analysis for voices now ESPS independent Spanish el voices tidy up (Borja) ToBI by rule implementation Confirmed support for gcc-2.7.2, gcc-2.6.3, Linux, FreeBSD, SunOS Alpha and SunCC port on Solaris reference card added return s-expressions in server/client mode OGI markup mode added. Native support for sun16, linux16 and freebsd (compile time option) Changed names of .C files to .cc files for bILL wagon integrated into speech tools (plus docs) auto-text-mode-alist for automatic selection of text mode from file name Associated token tests added Many more tokens dealt with (numbers, money, roman, phone, etc.) (analysed databases to see what coverage is like) A probablistic chart parser (no significant grammars though) RJC's new database/units/join/modify modules taking shape Some more examples added to the tutorial (with answers) Integrated CSLU changes for OGItoolkit including TCL support stml support for phrase types and words inline ssml -> stml Postlexical rules done in Scheme rather than C++ Rest of functions to allow any manipulation of utterance from scheme New duration models trained for both English and American New lexicon (CMU based) Consonant cluster support (for kd) American diphone set Cluster unit selection algorithm more robust Ngram backoff smoothing Token pos, for numbers (97.5%) but does poor on phone numbers New lexicon with final Rs and r deletion as postlex rule Update pos prediction (ts39) and phrase break ngrams (faster to load) New ngram format (binary files, and smoothing) Vowel reduction module Sun CC port New string class (rjc) remove dependence on libg++ Update of course notes and new section on building models from dbs Yarowsky homograph disambiguation Jan 24th 1997 1.1.1 release (first public release) a number of configuration and INSTALL documentation bugs fixed SSML tidied up and a festival script provided for it. Diphones, again, checked and copyright explicitly added Jan 6th 1997 1.1.0 release Roger diphones now default speaker A new unit clustering algorithm with acoustic costs and optimal coupling BSD socket client/server support A format function in Scheme (fprintf-like) A short course on Speech Synthesis in Festival (with course notes and exercises) A programmable form of text modes including externally customizable token to word rules. Fully programmable intonation module (for ToBI-like theories) Backtrace facility in Lisp Externally specified Utterance end (for all tts modes) with lookahead Roger diphones, first draft Nov 8th 1.0.0 release Substantial bug fixes, stabilization and documentation updates Added residual excited LPC synthesizer and removed PSOLA code. Made sucs and taylor optional modules, new modules can be added without modifying the base code MOBY lexicon (not as good as cuvoald but free) New diphone grouping software A new diphone database module (free from adc) Sept 30th 0.1 release MBROLA support (good example of external module) latest news: read out the latest news (from Time Warner, Pathfinder) audio spooler --language option on command line Spanish synthesis Letter to sound rules as external system (replacing all the NRL code) Welsh synthesis, making the whole system more language independent sucs spoke in reasonably way (gsw_450 and f2b dbs) document strings for functions (built in and user) and variables access from command line and dumped automatically into texinfo cleaned up SSML implementation break prediction integrated using viterbi and pos sucs module started (selection of units for concatenative synthesis) a part-of-speech tagging system (ngram/viterbi based) viterbi code added fixes in SIOD for running batch and stdin, also sub_prompts added saytime example Memory leaks fixed, no leaks for tts July 30th 0.0 release (just for the sake of it) a significiant start at documentation (texinfo -> info & html) festival scripts using #! on first line donovan diphone support can compile (with too many warnings) under g++ 2.7.2 copyrights on all files memory leak checks (only 8 bytes for "unknown" words) SSML (and tts file modes) cuvoald cmu and beep lexicons lexicon compilation web page, emacs interface June 2nd Klatt duration module syllabification in phones from letter to sound rules Linear Regression model for F0 prediction (from ToBI labels) CART (wagon) built trees for duration (zscores), phrase boundaries, accent and endtone prediction. ffeatures allowing specification of features of an utterance May integrated Taylor diphone module US Naval Research letter to sound rules CSTR lexicon 12th April first words "hello" start with speech_tools library, scheme-in-one-defun and readline and external CSTR diphone synthesizer 7th April 1996 work started festival/COPYING0000644000175000017500000001266312432155255012046 0ustar awbawbThe system as a whole and most of the files in it are distributed under the following copyright and conditions The Festival Speech Synthesis System Centre for Speech Technology Research University of Edinburgh, UK Copyright (c) 1996-2014 All Rights Reserved. Permission is hereby granted, free of charge, to use and distribute this software and its documentation without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of this work, and to permit persons to whom this work is furnished to do so, subject to the following conditions: 1. The code must retain the above copyright notice, this list of conditions and the following disclaimer. 2. Any modifications must be clearly marked as such. 3. Original authors' names are not deleted. 4. The authors' names are not used to endorse or promote products derived from this software without specific prior written permission. THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. Some further comments: Every effort has been made to ensure that Festival does not contain any violation of intellectual property rights through disclosure of trade secrets, copyright or patent violation. Considerable time and effort has been spent to ensure that this is the case. However, especially with patent problems, it is not always within our control to know what has or has not been restricted. If you do suspect that some part of Festival cannot be legally distributed please inform us so that an alternative may be sought. Festival is only useful if it is truly free to distribute. As of 1.4.0 the core distribution (and speech tools) is free. Unlike previous versions which had a commercial restriction. You are free to incorporate Festival in commercial (and of course non-commercial systems), without any further communication or licence from us. However if you are seriously using Festival within a commercial application we would like to know, both so we know we are contributing and so we can keep you informed of future developments. Also if you require maintenance, support or wish us to provide consultancy feel free to contact us. The voices however aren't all free. At present the US voices, kal and ked are free. Our British voices are free themselves but they use OALD which is restricted for non-commercial use. Our Spanish voice is also so restricted. Note other modules that festival supports e.g MBROLA and OGI extensions, may have different licencing please take care when using the system to understand what you are actually using. -------------------------------------------------- A number of individual files in the system fall under a different copyright from the above. All however are termed "free software" but most people. ./src/arch/festival/tcl.c * Copyright (C)1997 Jacques H. de Villiers * Copyright (C)1997 Center for Spoken Language Understanding, * Oregon Graduate Institute of Science & Technology See conditions in file. This is the standard TCL licence and hence shouldn't cause problems from most people. ./examples/festival_client.pl # Copyright (C) 1997 # Kevin A. Lenzo (lenzo@cs.cmu.edu) 7/97 See condition in file ./src/modules/clunits/* ./lib/*clunits* Joint copyright University of Edinburgh and Carnegie Mellon University Conditions remain as free software like the rest of distribution ./src/modules/clustergen/* Copyright Carnegie Mellon University Nagoya Insitute of Technology (derived code from HTS) HTS_vocoder_me.c HTS_vocoder_me.h contains additions from Aby Louw Copyright (c) 2012 The Department of Arts and Culture, The Government of the Republic of South Africa. under a free software licence compatible with the other free software copyrights in this work ./src/modules/hts_engine/* ./lib/hts.scm The HMM-based speech synthesis system (HTS) hts_engine API version 1.07 (http://hts-engine.sourceforge.net/) Copyright (C) 2001-2012 Nagoya Institute of Technology 2001-2008 Tokyo Institute of Technology All rights reserved. distributed under the Modified BSD licence. ./lib/festival.el ;;; Copyright (C) Alan W Black 1996 copyright under FSF General Public Licence Please also read the COPYING section of speech_tools/README for the conditions on those files. festival/INSTALL0000644000175000017500000005260612442333465012047 0ustar awbawbInstallation ************ This section describes how to install Festival from source in a new location and customize that installation. * Menu: * Requirements:: Software/Hardware requirements for Festival * Configuration:: Setting up compilation * Site initialization:: Settings for your particular site * Checking an installation:: But does it work ... Requirements ============ In order to compile Festival you first need the following source packages `festival-2.4-release.tar.gz' Festival Speech Synthesis System source `speech_tools-2.4-release.tar.gz' The Edinburgh Speech Tools Library `festlex_NAME.tar.gz' The lexicon distribution, where possible, includes the lexicon input file as well as the compiled form, for your convenience. The lexicons have varying distribution policies, but are all free except OALD, which is only free for non-commercial use (we are working on a free replacement). In some cases only a pointer to an ftp'able file plus a program to convert that file to the Festival format is included. `festvox_NAME.tar.gz' You'll need a speech database. A number are available (with varying distribution policies). Each voice may have other dependencies such as requiring particular lexicons `festdoc_2.0.tar.gz' Full postscript, info and html documentation for Festival and the Speech Tools. The source of the documentation is available in the standard distributions but for your conveniences it has been pre-generated. In addition to Festival specific sources you will also need _A UNIX machine_ Currently we have compiled and tested the system under Solaris (2.5(.1), 2.6, 2.7, 2.8 and 2.9), SunOS (4.1.3), FreeBSD 3.x, 4.x Linux (Redhat 4.1, 5.0, 5.1, 5.2, 6.[012], 7.[01], 8.0, 9, FC1 and other Linux distributions), and it should work under OSF (Dec Alphas) SGI (Irix), HPs (HPUX). But any standard UNIX machine should be acceptable. We have now successfully ported this version to Windows XP, Windows NT and Windows 95 (using the Cygnus GNU win32 environment). This is still a young port but seems to work. _A C++ compiler_ Note that C++ is not very portable even between different versions of the compiler from the same vendor. Although we've tried very hard to make the system portable, we know it is very unlikely to compile without change except with compilers that have already been tested. The currently tested systems are * Sun Sparc Solaris 2.5, 2.5.1, 2.6, 2.7, 2.9: GCC 2.95.1, GCC 3.2 * FreeBSD for Intel 3.x and 4.x GCC 2.95.1, GCC 3.0 * Linux for Intel (RedHat 4.1/5.0/5.1/5.2/6.0/7.x/8.0): GCC 2.7.2, GCC 2.7.2/egcs-1.0.2, egcs 1.1.1, egcs-1.1.2, GCC 2.95.[123], GCC "2.96", GCC 3.0, GCC 3.0.1, GCC 3.2, GCC 3.2.1 GCC 3.2.3, GCC 3.3.2 * Windows NT 4.0: GCC 2.7.2 plus egcs (from Cygnus GNU win32 b19), Visual C++ PRO v5.0, Visual C++ v6.0 Note if GCC works on one version of Unix it usually works on others. We have compiled both the speech tools and Festival under Windows NT 4.0 and Windows 95 using the GNU tools available from Cygnus. http://www.cygwin.com/ _GNU make_ Due to there being too many different `make' programs out there we have tested the system using GNU make on all systems we use. Others may work but we know GNU make does. _Audio hardware_ You can use Festival without audio output hardware but it doesn't sound very good (though admittedly you can hear less problems with it). A number of audio systems are supported (directly inherited from the audio support in the Edinburgh Speech Tools Library): NCD's NAS (formerly called netaudio) a network transparent audio system (which can be found at `ftp://ftp.x.org/contrib/audio/nas/'); `/dev/audio' (at 8k ulaw and 8/16bit linear), found on Suns, Linux machines and FreeBSD; and a method allowing arbitrary UNIX commands. *Note Audio output::. Earlier versions of Festival mistakenly offered a command line editor interface to the GNU package readline, but due to conflicts with the GNU Public Licence and Festival's licence this interface was removed in version 1.3.1. Even Festival's new free licence would cause problems as readline support would restrict Festival linking with non-free code. A new command line interface based on editline was provided that offers similar functionality. Editline remains a compilation option as it is probably not yet as portable as we would like it to be. In addition to the above, in order to process the documentation you will need `TeX', `dvips' (or similar), GNU's `makeinfo' (part of the texinfo package) and `texi2html' which is available from `http://wwwcn.cern.ch/dci/texi2html/'. However the document files are also available pre-processed into, postscript, DVI, info and html as part of the distribution in `festdoc-1.4.X.tar.gz'. Ensure you have a fully installed and working version of your C++ compiler. Most of the problems people have had in installing Festival have been due to incomplete or bad compiler installation. It might be worth checking if the following program works if you don't know if anyone has used your C++ installation before. #include int main (int argc, char **argv) { cout << "Hello world\n"; } Unpack all the source files in a new directory. The directory will then contain two subdirectories speech_tools/ festival/ Configuration ============= First ensure you have a compiled version of the Edinburgh Speech Tools Library. See `speech_tools/INSTALL' for instructions. The system now supports the standard GNU `configure' method for set up. In most cases this will automatically configure festival for your particular system. In most cases you need only type gmake and the system will configure itself and compile, (note you need to have compiled the Edinburgh Speech Tools `speech_tools-1.2.4' first. In some case hand configure is require. All of the configuration choice are held in the file `config/config' For the most part Festival configuration inherits the configuration from your speech tools config file (`../speech_tools/config/config'). Additional optional modules may be added by adding them to the end of your config file e.g. ALSO_INCLUDE += clunits Adding and new module here will treat is as a new directory in the `src/modules/' and compile it into the system in the same way the `OTHER_DIRS' feature was used in previous versions. If the compilation directory being accessed by NFS or if you use an automounter (e.g. amd) it is recommend to explicitly set the variable `FESTIVAL_HOME' in `config/config'. The command `pwd' is not reliable when a directory may have multiple names. There is a simple test suite with Festival but it requires the three basic voices and their respective lexicons install before it will work. Thus you need to install festlex_CMU.tar.gz festlex_OALD.tar.gz festlex_POSLEX.tar.gz festvox_don.tar.gz festvox_kedlpc16k.tar.gz festvox_rablpc16k.tar.gz If these are installed you can test the installation with gmake test To simply make it run with a male US English voiuce it is sufficient to install just festlex_CMU.tar.gz festlex_POSLEX.tar.gz festvox_kallpc16k.tar.gz Note that the single most common reason for problems in compilation and linking found amongst the beta testers was a bad installation of GNU C++. If you get many strange errors in G++ library header files or link errors it is worth checking that your system has the compiler, header files and runtime libraries properly installed. This may be checked by compiling a simple program under C++ and also finding out if anyone at your site has ever used the installation. Most of these installation problems are caused by upgrading to a newer version of libg++ without removing the older version so a mixed version of the `.h' files exist. Although we have tried very hard to ensure that Festival compiles with no warnings this is not possible under some systems. Under SunOS the system include files do not declare a number of system provided functions. This a bug in Sun's include files. This will causes warnings like "implicit definition of fprintf". These are harmless. Under Linux a warning at link time about reducing the size of some symbols often is produced. This is harmless. There is often occasional warnings about some socket system function having an incorrect argument type, this is also harmless. The speech tools and festival compile under Windows95 or Windows NT with Visual C++ v5.0 using the Microsoft `nmake' make program. We've only done this with the Professional edition, but have no reason to believe that it relies on anything not in the standard edition. In accordance to VC++ conventions, object files are created with extension .obj, executables with extension .exe and libraries with extension .lib. This may mean that both unix and Win32 versions can be built in the same directory tree, but I wouldn't rely on it. To do this you require nmake Makefiles for the system. These can be generated from the gnumake Makefiles, using the command gnumake VCMakefile in the speech_tools and festival directories. I have only done this under unix, it's possible it would work under the cygnus gnuwin32 system. If `make.depend' files exist (i.e. if you have done `gnumake depend' in unix) equivalent `vc_make.depend' files will be created, if not the VCMakefiles will not contain dependency information for the `.cc' files. The result will be that you can compile the system once, but changes will not cause the correct things to be rebuilt. In order to compile from the DOS command line using Visual C++ you need to have a collection of environment variables set. In Windows NT there is an installation option for Visual C++ which sets these globally. Under Windows95 or if you don't ask for them to be set globally under NT you need to run vcvars32.bat See the VC++ documentation for more details. Once you have the source trees with VCMakefiles somewhere visible from Windows, you need to copy `peech_tools\config\vc_config-dist' to `speech_tools\config\vc_config' and edit it to suit your local situation. Then do the same with `festival\config\vc_config-dist'. The thing most likely to need changing is the definition of `FESTIVAL_HOME' in `festival\config\vc_config_make_rules' which needs to point to where you have put festival. Now you can compile. cd to the speech_tools directory and do nmake /nologo /fVCMakefile and the library, the programs in main and the test programs should be compiled. The tests can't be run automatically under Windows. A simple test to check that things are probably OK is: main\na_play testsuite\data\ch_wave.wav which reads and plays a waveform. Next go into the festival directory and do nmake /nologo /fVCMakefile to build festival. When it's finished, and assuming you have the voices and lexicons unpacked in the right place, festival should run just as under unix. We should remind you that the NT/95 ports are still young and there may yet be problems that we've not found yet. We only recommend the use the speech tools and Festival under Windows if you have significant experience in C++ under those platforms. Most of the modules `src/modules' are actually optional and the system could be compiled without them. The basic set could be reduced further if certain facilities are not desired. Particularly: `donovan' which is only required if the donovan voice is used; `rxp' if no XML parsing is required (e.g. Sable); and `parser' if no stochastic paring is required (this parser isn't used for any of our currently released voices). Actually even `UniSyn' and `UniSyn_diphone' could be removed if some external waveform synthesizer is being used (e.g. MBROLA) or some alternative one like `OGIresLPC'. Removing unused modules will make the festival binary smaller and (potentially) start up faster but don't expect too much. You can delete these by changing the `BASE_DIRS' variable in `src/modules/Makefile'. Site initialization =================== Once compiled Festival may be further customized for particular sites. At start up time Festival loads the file `init.scm' from its library directory. This file further loads other necessary files such as phoneset descriptions, duration parameters, intonation parameters, definitions of voices etc. It will also load the files `sitevars.scm' and `siteinit.scm' if they exist. `sitevars.scm' is loaded after the basic Scheme library functions are loaded but before any of the festival related functions are loaded. This file is intended to set various path names before various subsystems are loaded. Typically variables such as `lexdir' (the directory where the lexicons are held), and `voices_dir' (pointing to voice directories) should be reset here if necessary. The default installation will try to find its lexicons and voices automatically based on the value of `load-path' (this is derived from `FESTIVAL_HOME' at compilation time or by using the `--libdir' at run-time). If the voices and lexicons have been unpacked into subdirectories of the library directory (the default) then no site specific initialization of the above pathnames will be necessary. The second site specific file is `siteinit.scm'. Typical examples of local initialization are as follows. The default audio output method is NCD's NAS system if that is supported as that's what we use normally in CSTR. If it is not supported, any hardware specific mode is the default (e.g. sun16audio, freebas16audio, linux16audio or mplayeraudio). But that default is just a setting in `init.scm'. If for example in your environment you may wish the default audio output method to be 8k mulaw through `/dev/audio' you should add the following line to your `siteinit.scm' file (Parameter.set 'Audio_Method 'sunaudio) Note the use of `Parameter.set' rather than `Parameter.def' the second function will not reset the value if it is already set. Remember that you may use the audio methods `sun16audio'. `linux16audio' or `freebsd16audio' only if `NATIVE_AUDIO' was selected in `speech_tools/config/config' and your are on such machines. The Festival variable `*modules*' contains a list of all supported functions/modules in a particular installation including audio support. Check the value of that variable if things aren't what you expect. If you are installing on a machine whose audio is not directly supported by the speech tools library, an external command may be executed to play a waveform. The following example is for an imaginary machine that can play audio files through a program called `adplay' with arguments for sample rate and file type. When playing waveforms, Festival, by default, outputs as unheadered waveform in native byte order. In this example you would set up the default audio playing mechanism in `siteinit.scm' as follows (Parameter.set 'Audio_Method 'Audio_Command) (Parameter.set 'Audio_Command "adplay -raw -r $SR $FILE") For `Audio_Command' method of playing waveforms Festival supports two additional audio parameters. `Audio_Required_Rate' allows you to use Festivals internal sample rate conversion function to any desired rate. Note this may not be as good as playing the waveform at the sample rate it is originally created in, but as some hardware devices are restrictive in what sample rates they support, or have naive resample functions this could be optimal. The second addition audio parameter is `Audio_Required_Format' which can be used to specify the desired output forms of the file. The default is unheadered raw, but this may be any of the values supported by the speech tools (including nist, esps, snd, riff, aiff, audlab, raw and, if you really want it, ascii). For example suppose you run Festival on a remote machine and are not running any network audio system and want Festival to copy files back to your local machine and simply cat them to `/dev/audio'. The following would do that (assuming permissions for rsh are allowed). (Parameter.set 'Audio_Method 'Audio_Command) ;; Make output file ulaw 8k (format ulaw implies 8k) (Parameter.set 'Audio_Required_Format 'ulaw) (Parameter.set 'Audio_Command "userhost=`echo $DISPLAY | sed 's/:.*$//'`; rcp $FILE $userhost:$FILE; \ rsh $userhost \"cat $FILE >/dev/audio\" ; rsh $userhost \"rm $FILE\"") Note there are limits on how complex a command you want to put in the `Audio_Command' string directly. It can get very confusing with respect to quoting. It is therefore recommended that once you get past a certain complexity consider writing a simple shell script and calling it from the `Audio_Command' string. A second typical customization is setting the default speaker. Speakers depend on many things but due to various licence (and resource) restrictions you may only have some diphone/nphone databases available in your installation. The function name that is the value of `voice_default' is called immediately after `siteinit.scm' is loaded offering the opportunity for you to change it. In the standard distribution no change should be required. If you download all the distributed voices `voice_rab_diphone' is the default voice. You may change this for a site by adding the following to `siteinit.scm' or per person by changing your `.festivalrc'. For example if you wish to change the default voice to the American one `voice_ked_diphone' (set! voice_default 'voice_ked_diphone) Note the single quote, and note that unlike in early versions `voice_default' is not a function you can call directly. A second level of customization is on a per user basis. After loading `init.scm', which includes `sitevars.scm' and `siteinit.scm' for local installation, Festival loads the file `.festivalrc' from the user's home directory (if it exists). This file may contain arbitrary Festival commands. Checking an installation ======================== Once compiled and site initialization is set up you should test to see if Festival can speak or not. Start the system $ bin/festival Festival Speech Synthesis System 2.0:release July 2004 Copyright (C) University of Edinburgh, 1996-2004. All rights reserved. For details type `(festival_warranty)' festival> ^D If errors occur at this stage they are most likely to do with pathname problems. If any error messages are printed about non-existent files check that those pathnames point to where you intended them to be. Most of the (default) pathnames are dependent on the basic library path. Ensure that is correct. To find out what it has been set to, start the system without loading the init files. $ bin/festival -q Festival Speech Synthesis System 1.4.3:release Jan 2003 Copyright (C) University of Edinburgh, 1996-2003. All rights reserved. For details type `(festival_warranty)' festival> libdir "/projects/festival/lib/" festival> ^D This should show the pathname you set in your `config/config'. If the system starts with no errors try to synthesize something festival> (SayText "hello world") Some files are only accessed at synthesis time so this may show up other problem pathnames. If it talks, you're in business, if it doesn't, here are some possible problems. If you get the error message Can't access NAS server You have selected NAS as the audio output but have no server running on that machine or your `DISPLAY' or `AUDIOSERVER' environment variable is not set properly for your output device. Either set these properly or change the audio output device in `lib/siteinit.scm' as described above. Ensure your audio device actually works the way you think it does. On Suns, the audio output device can be switched into a number of different output modes, speaker, jack, headphones. If this is set to the wrong one you may not hear the output. Use one of Sun's tools to change this (try `/usr/demo/SOUND/bin/soundtool'). Try to find an audio file independent of Festival and get it to play on your audio. Once you have done that ensure that the audio output method set in Festival matches that. Once you have got it talking, test the audio spooling device. festival> (intro) This plays a short introduction of two sentences, spooling the audio output. Finally exit from Festival (by end of file or `(quit)') and test the script mode with. $ examples/saytime A test suite is included with Festival but it makes certain assumptions about which voices are installed. It assumes that `voice_rab_diphone' (`festvox_rabxxxx.tar.gz') is the default voice and that `voice_ked_diphone' and `voice_don_diphone' (`festvox_kedxxxx.tar.gz' and `festvox_don.tar.gz') are installed. Also local settings in your `festival/lib/siteinit.scm' may affect these tests. However, after installation it may be worth trying gnumake test from the `festival/' directory. This will do various tests including basic utterance tests and tokenization tests. It also checks that voices are installed and that they don't interfere with each other. These tests are primarily regression tests for the developers of Festival, to ensure new enhancements don't mess up existing supported features. They are not designed to test an installation is successful, though if they run correctly it is most probable the installation has worked. festival/configure0000755000175000017500000012010207262620276012713 0ustar awbawb#! /bin/sh # Guess values for system-dependent variables and create Makefiles. # Generated automatically using autoconf version 2.13 # Copyright (C) 1992, 93, 94, 95, 96 Free Software Foundation, Inc. # # This configure script is free software; the Free Software Foundation # gives unlimited permission to copy, distribute and modify it. # Defaults: ac_help= ac_default_prefix=/usr/local # Any additions from configure.in: # Initialize some variables set by options. # The variables have the same names as the options, with # dashes changed to underlines. build=NONE cache_file=./config.cache exec_prefix=NONE host=NONE no_create= nonopt=NONE no_recursion= prefix=NONE program_prefix=NONE program_suffix=NONE program_transform_name=s,x,x, silent= site= srcdir= target=NONE verbose= x_includes=NONE x_libraries=NONE bindir='${exec_prefix}/bin' sbindir='${exec_prefix}/sbin' libexecdir='${exec_prefix}/libexec' datadir='${prefix}/share' sysconfdir='${prefix}/etc' sharedstatedir='${prefix}/com' localstatedir='${prefix}/var' libdir='${exec_prefix}/lib' includedir='${prefix}/include' oldincludedir='/usr/include' infodir='${prefix}/info' mandir='${prefix}/man' # Initialize some other variables. subdirs= MFLAGS= MAKEFLAGS= SHELL=${CONFIG_SHELL-/bin/sh} # Maximum number of lines to put in a shell here document. ac_max_here_lines=12 ac_prev= for ac_option do # If the previous option needs an argument, assign it. if test -n "$ac_prev"; then eval "$ac_prev=\$ac_option" ac_prev= continue fi case "$ac_option" in -*=*) ac_optarg=`echo "$ac_option" | sed 's/[-_a-zA-Z0-9]*=//'` ;; *) ac_optarg= ;; esac # Accept the important Cygnus configure options, so we can diagnose typos. case "$ac_option" in -bindir | --bindir | --bindi | --bind | --bin | --bi) ac_prev=bindir ;; -bindir=* | --bindir=* | --bindi=* | --bind=* | --bin=* | --bi=*) bindir="$ac_optarg" ;; -build | --build | --buil | --bui | --bu) ac_prev=build ;; -build=* | --build=* | --buil=* | --bui=* | --bu=*) build="$ac_optarg" ;; -cache-file | --cache-file | --cache-fil | --cache-fi \ | --cache-f | --cache- | --cache | --cach | --cac | --ca | --c) ac_prev=cache_file ;; -cache-file=* | --cache-file=* | --cache-fil=* | --cache-fi=* \ | --cache-f=* | --cache-=* | --cache=* | --cach=* | --cac=* | --ca=* | --c=*) cache_file="$ac_optarg" ;; -datadir | --datadir | --datadi | --datad | --data | --dat | --da) ac_prev=datadir ;; -datadir=* | --datadir=* | --datadi=* | --datad=* | --data=* | --dat=* \ | --da=*) datadir="$ac_optarg" ;; -disable-* | --disable-*) ac_feature=`echo $ac_option|sed -e 's/-*disable-//'` # Reject names that are not valid shell variable names. if test -n "`echo $ac_feature| sed 's/[-a-zA-Z0-9_]//g'`"; then { echo "configure: error: $ac_feature: invalid feature name" 1>&2; exit 1; } fi ac_feature=`echo $ac_feature| sed 's/-/_/g'` eval "enable_${ac_feature}=no" ;; -enable-* | --enable-*) ac_feature=`echo $ac_option|sed -e 's/-*enable-//' -e 's/=.*//'` # Reject names that are not valid shell variable names. if test -n "`echo $ac_feature| sed 's/[-_a-zA-Z0-9]//g'`"; then { echo "configure: error: $ac_feature: invalid feature name" 1>&2; exit 1; } fi ac_feature=`echo $ac_feature| sed 's/-/_/g'` case "$ac_option" in *=*) ;; *) ac_optarg=yes ;; esac eval "enable_${ac_feature}='$ac_optarg'" ;; -exec-prefix | --exec_prefix | --exec-prefix | --exec-prefi \ | --exec-pref | --exec-pre | --exec-pr | --exec-p | --exec- \ | --exec | --exe | --ex) ac_prev=exec_prefix ;; -exec-prefix=* | --exec_prefix=* | --exec-prefix=* | --exec-prefi=* \ | --exec-pref=* | --exec-pre=* | --exec-pr=* | --exec-p=* | --exec-=* \ | --exec=* | --exe=* | --ex=*) exec_prefix="$ac_optarg" ;; -gas | --gas | --ga | --g) # Obsolete; use --with-gas. with_gas=yes ;; -help | --help | --hel | --he) # Omit some internal or obsolete options to make the list less imposing. # This message is too long to be a string in the A/UX 3.1 sh. cat << EOF Usage: configure [options] [host] Options: [defaults in brackets after descriptions] Configuration: --cache-file=FILE cache test results in FILE --help print this message --no-create do not create output files --quiet, --silent do not print \`checking...' messages --version print the version of autoconf that created configure Directory and file names: --prefix=PREFIX install architecture-independent files in PREFIX [$ac_default_prefix] --exec-prefix=EPREFIX install architecture-dependent files in EPREFIX [same as prefix] --bindir=DIR user executables in DIR [EPREFIX/bin] --sbindir=DIR system admin executables in DIR [EPREFIX/sbin] --libexecdir=DIR program executables in DIR [EPREFIX/libexec] --datadir=DIR read-only architecture-independent data in DIR [PREFIX/share] --sysconfdir=DIR read-only single-machine data in DIR [PREFIX/etc] --sharedstatedir=DIR modifiable architecture-independent data in DIR [PREFIX/com] --localstatedir=DIR modifiable single-machine data in DIR [PREFIX/var] --libdir=DIR object code libraries in DIR [EPREFIX/lib] --includedir=DIR C header files in DIR [PREFIX/include] --oldincludedir=DIR C header files for non-gcc in DIR [/usr/include] --infodir=DIR info documentation in DIR [PREFIX/info] --mandir=DIR man documentation in DIR [PREFIX/man] --srcdir=DIR find the sources in DIR [configure dir or ..] --program-prefix=PREFIX prepend PREFIX to installed program names --program-suffix=SUFFIX append SUFFIX to installed program names --program-transform-name=PROGRAM run sed PROGRAM on installed program names EOF cat << EOF Host type: --build=BUILD configure for building on BUILD [BUILD=HOST] --host=HOST configure for HOST [guessed] --target=TARGET configure for TARGET [TARGET=HOST] Features and packages: --disable-FEATURE do not include FEATURE (same as --enable-FEATURE=no) --enable-FEATURE[=ARG] include FEATURE [ARG=yes] --with-PACKAGE[=ARG] use PACKAGE [ARG=yes] --without-PACKAGE do not use PACKAGE (same as --with-PACKAGE=no) --x-includes=DIR X include files are in DIR --x-libraries=DIR X library files are in DIR EOF if test -n "$ac_help"; then echo "--enable and --with options recognized:$ac_help" fi exit 0 ;; -host | --host | --hos | --ho) ac_prev=host ;; -host=* | --host=* | --hos=* | --ho=*) host="$ac_optarg" ;; -includedir | --includedir | --includedi | --included | --include \ | --includ | --inclu | --incl | --inc) ac_prev=includedir ;; -includedir=* | --includedir=* | --includedi=* | --included=* | --include=* \ | --includ=* | --inclu=* | --incl=* | --inc=*) includedir="$ac_optarg" ;; -infodir | --infodir | --infodi | --infod | --info | --inf) ac_prev=infodir ;; -infodir=* | --infodir=* | --infodi=* | --infod=* | --info=* | --inf=*) infodir="$ac_optarg" ;; -libdir | --libdir | --libdi | --libd) ac_prev=libdir ;; -libdir=* | --libdir=* | --libdi=* | --libd=*) libdir="$ac_optarg" ;; -libexecdir | --libexecdir | --libexecdi | --libexecd | --libexec \ | --libexe | --libex | --libe) ac_prev=libexecdir ;; -libexecdir=* | --libexecdir=* | --libexecdi=* | --libexecd=* | --libexec=* \ | --libexe=* | --libex=* | --libe=*) libexecdir="$ac_optarg" ;; -localstatedir | --localstatedir | --localstatedi | --localstated \ | --localstate | --localstat | --localsta | --localst \ | --locals | --local | --loca | --loc | --lo) ac_prev=localstatedir ;; -localstatedir=* | --localstatedir=* | --localstatedi=* | --localstated=* \ | --localstate=* | --localstat=* | --localsta=* | --localst=* \ | --locals=* | --local=* | --loca=* | --loc=* | --lo=*) localstatedir="$ac_optarg" ;; -mandir | --mandir | --mandi | --mand | --man | --ma | --m) ac_prev=mandir ;; -mandir=* | --mandir=* | --mandi=* | --mand=* | --man=* | --ma=* | --m=*) mandir="$ac_optarg" ;; -nfp | --nfp | --nf) # Obsolete; use --without-fp. with_fp=no ;; -no-create | --no-create | --no-creat | --no-crea | --no-cre \ | --no-cr | --no-c) no_create=yes ;; -no-recursion | --no-recursion | --no-recursio | --no-recursi \ | --no-recurs | --no-recur | --no-recu | --no-rec | --no-re | --no-r) no_recursion=yes ;; -oldincludedir | --oldincludedir | --oldincludedi | --oldincluded \ | --oldinclude | --oldinclud | --oldinclu | --oldincl | --oldinc \ | --oldin | --oldi | --old | --ol | --o) ac_prev=oldincludedir ;; -oldincludedir=* | --oldincludedir=* | --oldincludedi=* | --oldincluded=* \ | --oldinclude=* | --oldinclud=* | --oldinclu=* | --oldincl=* | --oldinc=* \ | --oldin=* | --oldi=* | --old=* | --ol=* | --o=*) oldincludedir="$ac_optarg" ;; -prefix | --prefix | --prefi | --pref | --pre | --pr | --p) ac_prev=prefix ;; -prefix=* | --prefix=* | --prefi=* | --pref=* | --pre=* | --pr=* | --p=*) prefix="$ac_optarg" ;; -program-prefix | --program-prefix | --program-prefi | --program-pref \ | --program-pre | --program-pr | --program-p) ac_prev=program_prefix ;; -program-prefix=* | --program-prefix=* | --program-prefi=* \ | --program-pref=* | --program-pre=* | --program-pr=* | --program-p=*) program_prefix="$ac_optarg" ;; -program-suffix | --program-suffix | --program-suffi | --program-suff \ | --program-suf | --program-su | --program-s) ac_prev=program_suffix ;; -program-suffix=* | --program-suffix=* | --program-suffi=* \ | --program-suff=* | --program-suf=* | --program-su=* | --program-s=*) program_suffix="$ac_optarg" ;; -program-transform-name | --program-transform-name \ | --program-transform-nam | --program-transform-na \ | --program-transform-n | --program-transform- \ | --program-transform | --program-transfor \ | --program-transfo | --program-transf \ | --program-trans | --program-tran \ | --progr-tra | --program-tr | --program-t) ac_prev=program_transform_name ;; -program-transform-name=* | --program-transform-name=* \ | --program-transform-nam=* | --program-transform-na=* \ | --program-transform-n=* | --program-transform-=* \ | --program-transform=* | --program-transfor=* \ | --program-transfo=* | --program-transf=* \ | --program-trans=* | --program-tran=* \ | --progr-tra=* | --program-tr=* | --program-t=*) program_transform_name="$ac_optarg" ;; -q | -quiet | --quiet | --quie | --qui | --qu | --q \ | -silent | --silent | --silen | --sile | --sil) silent=yes ;; -sbindir | --sbindir | --sbindi | --sbind | --sbin | --sbi | --sb) ac_prev=sbindir ;; -sbindir=* | --sbindir=* | --sbindi=* | --sbind=* | --sbin=* \ | --sbi=* | --sb=*) sbindir="$ac_optarg" ;; -sharedstatedir | --sharedstatedir | --sharedstatedi \ | --sharedstated | --sharedstate | --sharedstat | --sharedsta \ | --sharedst | --shareds | --shared | --share | --shar \ | --sha | --sh) ac_prev=sharedstatedir ;; -sharedstatedir=* | --sharedstatedir=* | --sharedstatedi=* \ | --sharedstated=* | --sharedstate=* | --sharedstat=* | --sharedsta=* \ | --sharedst=* | --shareds=* | --shared=* | --share=* | --shar=* \ | --sha=* | --sh=*) sharedstatedir="$ac_optarg" ;; -site | --site | --sit) ac_prev=site ;; -site=* | --site=* | --sit=*) site="$ac_optarg" ;; -srcdir | --srcdir | --srcdi | --srcd | --src | --sr) ac_prev=srcdir ;; -srcdir=* | --srcdir=* | --srcdi=* | --srcd=* | --src=* | --sr=*) srcdir="$ac_optarg" ;; -sysconfdir | --sysconfdir | --sysconfdi | --sysconfd | --sysconf \ | --syscon | --sysco | --sysc | --sys | --sy) ac_prev=sysconfdir ;; -sysconfdir=* | --sysconfdir=* | --sysconfdi=* | --sysconfd=* | --sysconf=* \ | --syscon=* | --sysco=* | --sysc=* | --sys=* | --sy=*) sysconfdir="$ac_optarg" ;; -target | --target | --targe | --targ | --tar | --ta | --t) ac_prev=target ;; -target=* | --target=* | --targe=* | --targ=* | --tar=* | --ta=* | --t=*) target="$ac_optarg" ;; -v | -verbose | --verbose | --verbos | --verbo | --verb) verbose=yes ;; -version | --version | --versio | --versi | --vers) echo "configure generated by autoconf version 2.13" exit 0 ;; -with-* | --with-*) ac_package=`echo $ac_option|sed -e 's/-*with-//' -e 's/=.*//'` # Reject names that are not valid shell variable names. if test -n "`echo $ac_package| sed 's/[-_a-zA-Z0-9]//g'`"; then { echo "configure: error: $ac_package: invalid package name" 1>&2; exit 1; } fi ac_package=`echo $ac_package| sed 's/-/_/g'` case "$ac_option" in *=*) ;; *) ac_optarg=yes ;; esac eval "with_${ac_package}='$ac_optarg'" ;; -without-* | --without-*) ac_package=`echo $ac_option|sed -e 's/-*without-//'` # Reject names that are not valid shell variable names. if test -n "`echo $ac_package| sed 's/[-a-zA-Z0-9_]//g'`"; then { echo "configure: error: $ac_package: invalid package name" 1>&2; exit 1; } fi ac_package=`echo $ac_package| sed 's/-/_/g'` eval "with_${ac_package}=no" ;; --x) # Obsolete; use --with-x. with_x=yes ;; -x-includes | --x-includes | --x-include | --x-includ | --x-inclu \ | --x-incl | --x-inc | --x-in | --x-i) ac_prev=x_includes ;; -x-includes=* | --x-includes=* | --x-include=* | --x-includ=* | --x-inclu=* \ | --x-incl=* | --x-inc=* | --x-in=* | --x-i=*) x_includes="$ac_optarg" ;; -x-libraries | --x-libraries | --x-librarie | --x-librari \ | --x-librar | --x-libra | --x-libr | --x-lib | --x-li | --x-l) ac_prev=x_libraries ;; -x-libraries=* | --x-libraries=* | --x-librarie=* | --x-librari=* \ | --x-librar=* | --x-libra=* | --x-libr=* | --x-lib=* | --x-li=* | --x-l=*) x_libraries="$ac_optarg" ;; -*) { echo "configure: error: $ac_option: invalid option; use --help to show usage" 1>&2; exit 1; } ;; *) if test -n "`echo $ac_option| sed 's/[-a-z0-9.]//g'`"; then echo "configure: warning: $ac_option: invalid host type" 1>&2 fi if test "x$nonopt" != xNONE; then { echo "configure: error: can only configure for one host and one target at a time" 1>&2; exit 1; } fi nonopt="$ac_option" ;; esac done if test -n "$ac_prev"; then { echo "configure: error: missing argument to --`echo $ac_prev | sed 's/_/-/g'`" 1>&2; exit 1; } fi trap 'rm -fr conftest* confdefs* core core.* *.core $ac_clean_files; exit 1' 1 2 15 # File descriptor usage: # 0 standard input # 1 file creation # 2 errors and warnings # 3 some systems may open it to /dev/tty # 4 used on the Kubota Titan # 6 checking for... messages and results # 5 compiler messages saved in config.log if test "$silent" = yes; then exec 6>/dev/null else exec 6>&1 fi exec 5>./config.log echo "\ This file contains any messages produced by compilers while running configure, to aid debugging if configure makes a mistake. " 1>&5 # Strip out --no-create and --no-recursion so they do not pile up. # Also quote any args containing shell metacharacters. ac_configure_args= for ac_arg do case "$ac_arg" in -no-create | --no-create | --no-creat | --no-crea | --no-cre \ | --no-cr | --no-c) ;; -no-recursion | --no-recursion | --no-recursio | --no-recursi \ | --no-recurs | --no-recur | --no-recu | --no-rec | --no-re | --no-r) ;; *" "*|*" "*|*[\[\]\~\#\$\^\&\*\(\)\{\}\\\|\;\<\>\?]*) ac_configure_args="$ac_configure_args '$ac_arg'" ;; *) ac_configure_args="$ac_configure_args $ac_arg" ;; esac done # NLS nuisances. # Only set these to C if already set. These must not be set unconditionally # because not all systems understand e.g. LANG=C (notably SCO). # Fixing LC_MESSAGES prevents Solaris sh from translating var values in `set'! # Non-C LC_CTYPE values break the ctype check. if test "${LANG+set}" = set; then LANG=C; export LANG; fi if test "${LC_ALL+set}" = set; then LC_ALL=C; export LC_ALL; fi if test "${LC_MESSAGES+set}" = set; then LC_MESSAGES=C; export LC_MESSAGES; fi if test "${LC_CTYPE+set}" = set; then LC_CTYPE=C; export LC_CTYPE; fi # confdefs.h avoids OS command line length limits that DEFS can exceed. rm -rf conftest* confdefs.h # AIX cpp loses on an empty file, so make sure it contains at least a newline. echo > confdefs.h # A filename unique to this package, relative to the directory that # configure is in, which we can look for to find out if srcdir is correct. ac_unique_file=src/include/festival.h # Find the source files, if location was not specified. if test -z "$srcdir"; then ac_srcdir_defaulted=yes # Try the directory containing this script, then its parent. ac_prog=$0 ac_confdir=`echo $ac_prog|sed 's%/[^/][^/]*$%%'` test "x$ac_confdir" = "x$ac_prog" && ac_confdir=. srcdir=$ac_confdir if test ! -r $srcdir/$ac_unique_file; then srcdir=.. fi else ac_srcdir_defaulted=no fi if test ! -r $srcdir/$ac_unique_file; then if test "$ac_srcdir_defaulted" = yes; then { echo "configure: error: can not find sources in $ac_confdir or .." 1>&2; exit 1; } else { echo "configure: error: can not find sources in $srcdir" 1>&2; exit 1; } fi fi srcdir=`echo "${srcdir}" | sed 's%\([^/]\)/*$%\1%'` # Prefer explicitly selected file to automatically selected ones. if test -z "$CONFIG_SITE"; then if test "x$prefix" != xNONE; then CONFIG_SITE="$prefix/share/config.site $prefix/etc/config.site" else CONFIG_SITE="$ac_default_prefix/share/config.site $ac_default_prefix/etc/config.site" fi fi for ac_site_file in $CONFIG_SITE; do if test -r "$ac_site_file"; then echo "loading site script $ac_site_file" . "$ac_site_file" fi done if test -r "$cache_file"; then echo "loading cache $cache_file" . $cache_file else echo "creating cache $cache_file" > $cache_file fi ac_ext=c # CFLAGS is not in ac_cpp because -g, -O, etc. are not valid cpp options. ac_cpp='$CPP $CPPFLAGS' ac_compile='${CC-cc} -c $CFLAGS $CPPFLAGS conftest.$ac_ext 1>&5' ac_link='${CC-cc} -o conftest${ac_exeext} $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS 1>&5' cross_compiling=$ac_cv_prog_cc_cross ac_exeext= ac_objext=o if (echo "testing\c"; echo 1,2,3) | grep c >/dev/null; then # Stardent Vistra SVR4 grep lacks -e, says ghazi@caip.rutgers.edu. if (echo -n testing; echo 1,2,3) | sed s/-n/xn/ | grep xn >/dev/null; then ac_n= ac_c=' ' ac_t=' ' else ac_n=-n ac_c= ac_t= fi else ac_n= ac_c='\c' ac_t= fi ac_aux_dir= for ac_dir in $srcdir $srcdir/.. $srcdir/../..; do if test -f $ac_dir/install-sh; then ac_aux_dir=$ac_dir ac_install_sh="$ac_aux_dir/install-sh -c" break elif test -f $ac_dir/install.sh; then ac_aux_dir=$ac_dir ac_install_sh="$ac_aux_dir/install.sh -c" break fi done if test -z "$ac_aux_dir"; then { echo "configure: error: can not find install-sh or install.sh in $srcdir $srcdir/.. $srcdir/../.." 1>&2; exit 1; } fi ac_config_guess=$ac_aux_dir/config.guess ac_config_sub=$ac_aux_dir/config.sub ac_configure=$ac_aux_dir/configure # This should be Cygnus configure. # Do some error checking and defaulting for the host and target type. # The inputs are: # configure --host=HOST --target=TARGET --build=BUILD NONOPT # # The rules are: # 1. You are not allowed to specify --host, --target, and nonopt at the # same time. # 2. Host defaults to nonopt. # 3. If nonopt is not specified, then host defaults to the current host, # as determined by config.guess. # 4. Target and build default to nonopt. # 5. If nonopt is not specified, then target and build default to host. # The aliases save the names the user supplied, while $host etc. # will get canonicalized. case $host---$target---$nonopt in NONE---*---* | *---NONE---* | *---*---NONE) ;; *) { echo "configure: error: can only configure for one host and one target at a time" 1>&2; exit 1; } ;; esac # Make sure we can run config.sub. if ${CONFIG_SHELL-/bin/sh} $ac_config_sub sun4 >/dev/null 2>&1; then : else { echo "configure: error: can not run $ac_config_sub" 1>&2; exit 1; } fi echo $ac_n "checking host system type""... $ac_c" 1>&6 echo "configure:573: checking host system type" >&5 host_alias=$host case "$host_alias" in NONE) case $nonopt in NONE) if host_alias=`${CONFIG_SHELL-/bin/sh} $ac_config_guess`; then : else { echo "configure: error: can not guess host type; you must specify one" 1>&2; exit 1; } fi ;; *) host_alias=$nonopt ;; esac ;; esac host=`${CONFIG_SHELL-/bin/sh} $ac_config_sub $host_alias` host_cpu=`echo $host | sed 's/^\([^-]*\)-\([^-]*\)-\(.*\)$/\1/'` host_vendor=`echo $host | sed 's/^\([^-]*\)-\([^-]*\)-\(.*\)$/\2/'` host_os=`echo $host | sed 's/^\([^-]*\)-\([^-]*\)-\(.*\)$/\3/'` echo "$ac_t""$host" 1>&6 echo $ac_n "checking target system type""... $ac_c" 1>&6 echo "configure:594: checking target system type" >&5 target_alias=$target case "$target_alias" in NONE) case $nonopt in NONE) target_alias=$host_alias ;; *) target_alias=$nonopt ;; esac ;; esac target=`${CONFIG_SHELL-/bin/sh} $ac_config_sub $target_alias` target_cpu=`echo $target | sed 's/^\([^-]*\)-\([^-]*\)-\(.*\)$/\1/'` target_vendor=`echo $target | sed 's/^\([^-]*\)-\([^-]*\)-\(.*\)$/\2/'` target_os=`echo $target | sed 's/^\([^-]*\)-\([^-]*\)-\(.*\)$/\3/'` echo "$ac_t""$target" 1>&6 echo $ac_n "checking build system type""... $ac_c" 1>&6 echo "configure:612: checking build system type" >&5 build_alias=$build case "$build_alias" in NONE) case $nonopt in NONE) build_alias=$host_alias ;; *) build_alias=$nonopt ;; esac ;; esac build=`${CONFIG_SHELL-/bin/sh} $ac_config_sub $build_alias` build_cpu=`echo $build | sed 's/^\([^-]*\)-\([^-]*\)-\(.*\)$/\1/'` build_vendor=`echo $build | sed 's/^\([^-]*\)-\([^-]*\)-\(.*\)$/\2/'` build_os=`echo $build | sed 's/^\([^-]*\)-\([^-]*\)-\(.*\)$/\3/'` echo "$ac_t""$build" 1>&6 test "$host_alias" != "$target_alias" && test "$program_prefix$program_suffix$program_transform_name" = \ NONENONEs,x,x, && program_prefix=${target_alias}- # Extract the first word of "gcc", so it can be a program name with args. set dummy gcc; ac_word=$2 echo $ac_n "checking for $ac_word""... $ac_c" 1>&6 echo "configure:637: checking for $ac_word" >&5 if eval "test \"`echo '$''{'ac_cv_prog_CC'+set}'`\" = set"; then echo $ac_n "(cached) $ac_c" 1>&6 else if test -n "$CC"; then ac_cv_prog_CC="$CC" # Let the user override the test. else IFS="${IFS= }"; ac_save_ifs="$IFS"; IFS=":" ac_dummy="$PATH" for ac_dir in $ac_dummy; do test -z "$ac_dir" && ac_dir=. if test -f $ac_dir/$ac_word; then ac_cv_prog_CC="gcc" break fi done IFS="$ac_save_ifs" fi fi CC="$ac_cv_prog_CC" if test -n "$CC"; then echo "$ac_t""$CC" 1>&6 else echo "$ac_t""no" 1>&6 fi if test -z "$CC"; then # Extract the first word of "cc", so it can be a program name with args. set dummy cc; ac_word=$2 echo $ac_n "checking for $ac_word""... $ac_c" 1>&6 echo "configure:667: checking for $ac_word" >&5 if eval "test \"`echo '$''{'ac_cv_prog_CC'+set}'`\" = set"; then echo $ac_n "(cached) $ac_c" 1>&6 else if test -n "$CC"; then ac_cv_prog_CC="$CC" # Let the user override the test. else IFS="${IFS= }"; ac_save_ifs="$IFS"; IFS=":" ac_prog_rejected=no ac_dummy="$PATH" for ac_dir in $ac_dummy; do test -z "$ac_dir" && ac_dir=. if test -f $ac_dir/$ac_word; then if test "$ac_dir/$ac_word" = "/usr/ucb/cc"; then ac_prog_rejected=yes continue fi ac_cv_prog_CC="cc" break fi done IFS="$ac_save_ifs" if test $ac_prog_rejected = yes; then # We found a bogon in the path, so make sure we never use it. set dummy $ac_cv_prog_CC shift if test $# -gt 0; then # We chose a different compiler from the bogus one. # However, it has the same basename, so the bogon will be chosen # first if we set CC to just the basename; use the full file name. shift set dummy "$ac_dir/$ac_word" "$@" shift ac_cv_prog_CC="$@" fi fi fi fi CC="$ac_cv_prog_CC" if test -n "$CC"; then echo "$ac_t""$CC" 1>&6 else echo "$ac_t""no" 1>&6 fi if test -z "$CC"; then case "`uname -s`" in *win32* | *WIN32*) # Extract the first word of "cl", so it can be a program name with args. set dummy cl; ac_word=$2 echo $ac_n "checking for $ac_word""... $ac_c" 1>&6 echo "configure:718: checking for $ac_word" >&5 if eval "test \"`echo '$''{'ac_cv_prog_CC'+set}'`\" = set"; then echo $ac_n "(cached) $ac_c" 1>&6 else if test -n "$CC"; then ac_cv_prog_CC="$CC" # Let the user override the test. else IFS="${IFS= }"; ac_save_ifs="$IFS"; IFS=":" ac_dummy="$PATH" for ac_dir in $ac_dummy; do test -z "$ac_dir" && ac_dir=. if test -f $ac_dir/$ac_word; then ac_cv_prog_CC="cl" break fi done IFS="$ac_save_ifs" fi fi CC="$ac_cv_prog_CC" if test -n "$CC"; then echo "$ac_t""$CC" 1>&6 else echo "$ac_t""no" 1>&6 fi ;; esac fi test -z "$CC" && { echo "configure: error: no acceptable cc found in \$PATH" 1>&2; exit 1; } fi echo $ac_n "checking whether the C compiler ($CC $CFLAGS $LDFLAGS) works""... $ac_c" 1>&6 echo "configure:750: checking whether the C compiler ($CC $CFLAGS $LDFLAGS) works" >&5 ac_ext=c # CFLAGS is not in ac_cpp because -g, -O, etc. are not valid cpp options. ac_cpp='$CPP $CPPFLAGS' ac_compile='${CC-cc} -c $CFLAGS $CPPFLAGS conftest.$ac_ext 1>&5' ac_link='${CC-cc} -o conftest${ac_exeext} $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS 1>&5' cross_compiling=$ac_cv_prog_cc_cross cat > conftest.$ac_ext << EOF #line 761 "configure" #include "confdefs.h" main(){return(0);} EOF if { (eval echo configure:766: \"$ac_link\") 1>&5; (eval $ac_link) 2>&5; } && test -s conftest${ac_exeext}; then ac_cv_prog_cc_works=yes # If we can't run a trivial program, we are probably using a cross compiler. if (./conftest; exit) 2>/dev/null; then ac_cv_prog_cc_cross=no else ac_cv_prog_cc_cross=yes fi else echo "configure: failed program was:" >&5 cat conftest.$ac_ext >&5 ac_cv_prog_cc_works=no fi rm -fr conftest* ac_ext=c # CFLAGS is not in ac_cpp because -g, -O, etc. are not valid cpp options. ac_cpp='$CPP $CPPFLAGS' ac_compile='${CC-cc} -c $CFLAGS $CPPFLAGS conftest.$ac_ext 1>&5' ac_link='${CC-cc} -o conftest${ac_exeext} $CFLAGS $CPPFLAGS $LDFLAGS conftest.$ac_ext $LIBS 1>&5' cross_compiling=$ac_cv_prog_cc_cross echo "$ac_t""$ac_cv_prog_cc_works" 1>&6 if test $ac_cv_prog_cc_works = no; then { echo "configure: error: installation or configuration problem: C compiler cannot create executables." 1>&2; exit 1; } fi echo $ac_n "checking whether the C compiler ($CC $CFLAGS $LDFLAGS) is a cross-compiler""... $ac_c" 1>&6 echo "configure:792: checking whether the C compiler ($CC $CFLAGS $LDFLAGS) is a cross-compiler" >&5 echo "$ac_t""$ac_cv_prog_cc_cross" 1>&6 cross_compiling=$ac_cv_prog_cc_cross echo $ac_n "checking whether we are using GNU C""... $ac_c" 1>&6 echo "configure:797: checking whether we are using GNU C" >&5 if eval "test \"`echo '$''{'ac_cv_prog_gcc'+set}'`\" = set"; then echo $ac_n "(cached) $ac_c" 1>&6 else cat > conftest.c <&5; (eval $ac_try) 2>&5; }; } | egrep yes >/dev/null 2>&1; then ac_cv_prog_gcc=yes else ac_cv_prog_gcc=no fi fi echo "$ac_t""$ac_cv_prog_gcc" 1>&6 if test $ac_cv_prog_gcc = yes; then GCC=yes else GCC= fi ac_test_CFLAGS="${CFLAGS+set}" ac_save_CFLAGS="$CFLAGS" CFLAGS= echo $ac_n "checking whether ${CC-cc} accepts -g""... $ac_c" 1>&6 echo "configure:825: checking whether ${CC-cc} accepts -g" >&5 if eval "test \"`echo '$''{'ac_cv_prog_cc_g'+set}'`\" = set"; then echo $ac_n "(cached) $ac_c" 1>&6 else echo 'void f(){}' > conftest.c if test -z "`${CC-cc} -g -c conftest.c 2>&1`"; then ac_cv_prog_cc_g=yes else ac_cv_prog_cc_g=no fi rm -f conftest* fi echo "$ac_t""$ac_cv_prog_cc_g" 1>&6 if test "$ac_test_CFLAGS" = set; then CFLAGS="$ac_save_CFLAGS" elif test $ac_cv_prog_cc_g = yes; then if test "$GCC" = yes; then CFLAGS="-g -O2" else CFLAGS="-g" fi else if test "$GCC" = yes; then CFLAGS="-O2" else CFLAGS= fi fi if test "x$GCC" = "xyes"; then CFLAGS="$CFLAGS -Wall" fi # Extract the first word of "ranlib", so it can be a program name with args. set dummy ranlib; ac_word=$2 echo $ac_n "checking for $ac_word""... $ac_c" 1>&6 echo "configure:862: checking for $ac_word" >&5 if eval "test \"`echo '$''{'ac_cv_prog_RANLIB'+set}'`\" = set"; then echo $ac_n "(cached) $ac_c" 1>&6 else if test -n "$RANLIB"; then ac_cv_prog_RANLIB="$RANLIB" # Let the user override the test. else IFS="${IFS= }"; ac_save_ifs="$IFS"; IFS=":" ac_dummy="$PATH" for ac_dir in $ac_dummy; do test -z "$ac_dir" && ac_dir=. if test -f $ac_dir/$ac_word; then ac_cv_prog_RANLIB="ranlib" break fi done IFS="$ac_save_ifs" test -z "$ac_cv_prog_RANLIB" && ac_cv_prog_RANLIB=":" fi fi RANLIB="$ac_cv_prog_RANLIB" if test -n "$RANLIB"; then echo "$ac_t""$RANLIB" 1>&6 else echo "$ac_t""no" 1>&6 fi if test $host != $build; then ac_tool_prefix=${host_alias}- else ac_tool_prefix= fi # Extract the first word of "${ac_tool_prefix}ar", so it can be a program name with args. set dummy ${ac_tool_prefix}ar; ac_word=$2 echo $ac_n "checking for $ac_word""... $ac_c" 1>&6 echo "configure:898: checking for $ac_word" >&5 if eval "test \"`echo '$''{'ac_cv_prog_AR'+set}'`\" = set"; then echo $ac_n "(cached) $ac_c" 1>&6 else if test -n "$AR"; then ac_cv_prog_AR="$AR" # Let the user override the test. else IFS="${IFS= }"; ac_save_ifs="$IFS"; IFS=":" ac_dummy="$PATH" for ac_dir in $ac_dummy; do test -z "$ac_dir" && ac_dir=. if test -f $ac_dir/$ac_word; then ac_cv_prog_AR="${ac_tool_prefix}ar" break fi done IFS="$ac_save_ifs" test -z "$ac_cv_prog_AR" && ac_cv_prog_AR="ar" fi fi AR="$ac_cv_prog_AR" if test -n "$AR"; then echo "$ac_t""$AR" 1>&6 else echo "$ac_t""no" 1>&6 fi echo $ac_n "checking whether byte ordering is bigendian""... $ac_c" 1>&6 echo "configure:929: checking whether byte ordering is bigendian" >&5 if eval "test \"`echo '$''{'ac_cv_c_bigendian'+set}'`\" = set"; then echo $ac_n "(cached) $ac_c" 1>&6 else ac_cv_c_bigendian=unknown # See if sys/param.h defines the BYTE_ORDER macro. cat > conftest.$ac_ext < #include int main() { #if !BYTE_ORDER || !BIG_ENDIAN || !LITTLE_ENDIAN bogus endian macros #endif ; return 0; } EOF if { (eval echo configure:947: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then rm -rf conftest* # It does; now see whether it defined to BIG_ENDIAN or not. cat > conftest.$ac_ext < #include int main() { #if BYTE_ORDER != BIG_ENDIAN not big endian #endif ; return 0; } EOF if { (eval echo configure:962: \"$ac_compile\") 1>&5; (eval $ac_compile) 2>&5; }; then rm -rf conftest* ac_cv_c_bigendian=yes else echo "configure: failed program was:" >&5 cat conftest.$ac_ext >&5 rm -rf conftest* ac_cv_c_bigendian=no fi rm -f conftest* else echo "configure: failed program was:" >&5 cat conftest.$ac_ext >&5 fi rm -f conftest* if test $ac_cv_c_bigendian = unknown; then if test "$cross_compiling" = yes; then { echo "configure: error: can not run test program while cross compiling" 1>&2; exit 1; } else cat > conftest.$ac_ext <&5; (eval $ac_link) 2>&5; } && test -s conftest${ac_exeext} && (./conftest; exit) 2>/dev/null then ac_cv_c_bigendian=no else echo "configure: failed program was:" >&5 cat conftest.$ac_ext >&5 rm -fr conftest* ac_cv_c_bigendian=yes fi rm -fr conftest* fi fi fi echo "$ac_t""$ac_cv_c_bigendian" 1>&6 if test $ac_cv_c_bigendian = yes; then cat >> confdefs.h <<\EOF #define WORDS_BIGENDIAN 1 EOF fi trap '' 1 2 15 cat > confcache <<\EOF # This file is a shell script that caches the results of configure # tests run on this system so they can be shared between configure # scripts and configure runs. It is not useful on other systems. # If it contains results you don't want to keep, you may remove or edit it. # # By default, configure uses ./config.cache as the cache file, # creating it if it does not exist already. You can give configure # the --cache-file=FILE option to use a different cache file; that is # what configure does when it calls configure scripts in # subdirectories, so they share the cache. # Giving --cache-file=/dev/null disables caching, for debugging configure. # config.status only pays attention to the cache file if you give it the # --recheck option to rerun configure. # EOF # The following way of writing the cache mishandles newlines in values, # but we know of no workaround that is simple, portable, and efficient. # So, don't put newlines in cache variables' values. # Ultrix sh set writes to stderr and can't be redirected directly, # and sets the high bit in the cache file unless we assign to the vars. (set) 2>&1 | case `(ac_space=' '; set | grep ac_space) 2>&1` in *ac_space=\ *) # `set' does not quote correctly, so add quotes (double-quote substitution # turns \\\\ into \\, and sed turns \\ into \). sed -n \ -e "s/'/'\\\\''/g" \ -e "s/^\\([a-zA-Z0-9_]*_cv_[a-zA-Z0-9_]*\\)=\\(.*\\)/\\1=\${\\1='\\2'}/p" ;; *) # `set' quotes correctly as required by POSIX, so do not add quotes. sed -n -e 's/^\([a-zA-Z0-9_]*_cv_[a-zA-Z0-9_]*\)=\(.*\)/\1=${\1=\2}/p' ;; esac >> confcache if cmp -s $cache_file confcache; then : else if test -w $cache_file; then echo "updating cache $cache_file" cat confcache > $cache_file else echo "not updating unwritable cache $cache_file" fi fi rm -f confcache trap 'rm -fr conftest* confdefs* core core.* *.core $ac_clean_files; exit 1' 1 2 15 test "x$prefix" = xNONE && prefix=$ac_default_prefix # Let make expand exec_prefix. test "x$exec_prefix" = xNONE && exec_prefix='${prefix}' # Any assignment to VPATH causes Sun make to only execute # the first set of double-colon rules, so remove it if not needed. # If there is a colon in the path, we need to keep it. if test "x$srcdir" = x.; then ac_vpsub='/^[ ]*VPATH[ ]*=[^:]*$/d' fi trap 'rm -f $CONFIG_STATUS conftest*; exit 1' 1 2 15 # Transform confdefs.h into DEFS. # Protect against shell expansion while executing Makefile rules. # Protect against Makefile macro expansion. cat > conftest.defs <<\EOF s%#define \([A-Za-z_][A-Za-z0-9_]*\) *\(.*\)%-D\1=\2%g s%[ `~#$^&*(){}\\|;'"<>?]%\\&%g s%\[%\\&%g s%\]%\\&%g s%\$%$$%g EOF DEFS=`sed -f conftest.defs confdefs.h | tr '\012' ' '` rm -f conftest.defs # Without the "./", some shells look in PATH for config.status. : ${CONFIG_STATUS=./config.status} echo creating $CONFIG_STATUS rm -f $CONFIG_STATUS cat > $CONFIG_STATUS </dev/null | sed 1q`: # # $0 $ac_configure_args # # Compiler output produced by configure, useful for debugging # configure, is in ./config.log if it exists. ac_cs_usage="Usage: $CONFIG_STATUS [--recheck] [--version] [--help]" for ac_option do case "\$ac_option" in -recheck | --recheck | --rechec | --reche | --rech | --rec | --re | --r) echo "running \${CONFIG_SHELL-/bin/sh} $0 $ac_configure_args --no-create --no-recursion" exec \${CONFIG_SHELL-/bin/sh} $0 $ac_configure_args --no-create --no-recursion ;; -version | --version | --versio | --versi | --vers | --ver | --ve | --v) echo "$CONFIG_STATUS generated by autoconf version 2.13" exit 0 ;; -help | --help | --hel | --he | --h) echo "\$ac_cs_usage"; exit 0 ;; *) echo "\$ac_cs_usage"; exit 1 ;; esac done ac_given_srcdir=$srcdir trap 'rm -fr `echo "config/config" | sed "s/:[^ ]*//g"` conftest*; exit 1' 1 2 15 EOF cat >> $CONFIG_STATUS < conftest.subs <<\\CEOF $ac_vpsub $extrasub s%@SHELL@%$SHELL%g s%@CFLAGS@%$CFLAGS%g s%@CPPFLAGS@%$CPPFLAGS%g s%@CXXFLAGS@%$CXXFLAGS%g s%@FFLAGS@%$FFLAGS%g s%@DEFS@%$DEFS%g s%@LDFLAGS@%$LDFLAGS%g s%@LIBS@%$LIBS%g s%@exec_prefix@%$exec_prefix%g s%@prefix@%$prefix%g s%@program_transform_name@%$program_transform_name%g s%@bindir@%$bindir%g s%@sbindir@%$sbindir%g s%@libexecdir@%$libexecdir%g s%@datadir@%$datadir%g s%@sysconfdir@%$sysconfdir%g s%@sharedstatedir@%$sharedstatedir%g s%@localstatedir@%$localstatedir%g s%@libdir@%$libdir%g s%@includedir@%$includedir%g s%@oldincludedir@%$oldincludedir%g s%@infodir@%$infodir%g s%@mandir@%$mandir%g s%@host@%$host%g s%@host_alias@%$host_alias%g s%@host_cpu@%$host_cpu%g s%@host_vendor@%$host_vendor%g s%@host_os@%$host_os%g s%@target@%$target%g s%@target_alias@%$target_alias%g s%@target_cpu@%$target_cpu%g s%@target_vendor@%$target_vendor%g s%@target_os@%$target_os%g s%@build@%$build%g s%@build_alias@%$build_alias%g s%@build_cpu@%$build_cpu%g s%@build_vendor@%$build_vendor%g s%@build_os@%$build_os%g s%@CC@%$CC%g s%@RANLIB@%$RANLIB%g s%@AR@%$AR%g CEOF EOF cat >> $CONFIG_STATUS <<\EOF # Split the substitutions into bite-sized pieces for seds with # small command number limits, like on Digital OSF/1 and HP-UX. ac_max_sed_cmds=90 # Maximum number of lines to put in a sed script. ac_file=1 # Number of current file. ac_beg=1 # First line for current file. ac_end=$ac_max_sed_cmds # Line after last line for current file. ac_more_lines=: ac_sed_cmds="" while $ac_more_lines; do if test $ac_beg -gt 1; then sed "1,${ac_beg}d; ${ac_end}q" conftest.subs > conftest.s$ac_file else sed "${ac_end}q" conftest.subs > conftest.s$ac_file fi if test ! -s conftest.s$ac_file; then ac_more_lines=false rm -f conftest.s$ac_file else if test -z "$ac_sed_cmds"; then ac_sed_cmds="sed -f conftest.s$ac_file" else ac_sed_cmds="$ac_sed_cmds | sed -f conftest.s$ac_file" fi ac_file=`expr $ac_file + 1` ac_beg=$ac_end ac_end=`expr $ac_end + $ac_max_sed_cmds` fi done if test -z "$ac_sed_cmds"; then ac_sed_cmds=cat fi EOF cat >> $CONFIG_STATUS <> $CONFIG_STATUS <<\EOF for ac_file in .. $CONFIG_FILES; do if test "x$ac_file" != x..; then # Support "outfile[:infile[:infile...]]", defaulting infile="outfile.in". case "$ac_file" in *:*) ac_file_in=`echo "$ac_file"|sed 's%[^:]*:%%'` ac_file=`echo "$ac_file"|sed 's%:.*%%'` ;; *) ac_file_in="${ac_file}.in" ;; esac # Adjust a relative srcdir, top_srcdir, and INSTALL for subdirectories. # Remove last slash and all that follows it. Not all systems have dirname. ac_dir=`echo $ac_file|sed 's%/[^/][^/]*$%%'` if test "$ac_dir" != "$ac_file" && test "$ac_dir" != .; then # The file is in a subdirectory. test ! -d "$ac_dir" && mkdir "$ac_dir" ac_dir_suffix="/`echo $ac_dir|sed 's%^\./%%'`" # A "../" for each directory in $ac_dir_suffix. ac_dots=`echo $ac_dir_suffix|sed 's%/[^/]*%../%g'` else ac_dir_suffix= ac_dots= fi case "$ac_given_srcdir" in .) srcdir=. if test -z "$ac_dots"; then top_srcdir=. else top_srcdir=`echo $ac_dots|sed 's%/$%%'`; fi ;; /*) srcdir="$ac_given_srcdir$ac_dir_suffix"; top_srcdir="$ac_given_srcdir" ;; *) # Relative path. srcdir="$ac_dots$ac_given_srcdir$ac_dir_suffix" top_srcdir="$ac_dots$ac_given_srcdir" ;; esac echo creating "$ac_file" rm -f "$ac_file" configure_input="Generated automatically from `echo $ac_file_in|sed 's%.*/%%'` by configure." case "$ac_file" in *Makefile*) ac_comsub="1i\\ # $configure_input" ;; *) ac_comsub= ;; esac ac_file_inputs=`echo $ac_file_in|sed -e "s%^%$ac_given_srcdir/%" -e "s%:% $ac_given_srcdir/%g"` sed -e "$ac_comsub s%@configure_input@%$configure_input%g s%@srcdir@%$srcdir%g s%@top_srcdir@%$top_srcdir%g " $ac_file_inputs | (eval "$ac_sed_cmds") > $ac_file fi; done rm -f conftest.s* EOF cat >> $CONFIG_STATUS <> $CONFIG_STATUS <<\EOF exit 0 EOF chmod +x $CONFIG_STATUS rm -fr confdefs* $ac_clean_files test "$no_create" = yes || ${CONFIG_SHELL-/bin/sh} $CONFIG_STATUS || exit 1 festival/configure.in0000644000175000017500000000511107262616703013317 0ustar awbawbdnl######################################################################## dnl ## dnl Centre for Speech Technology Research ## dnl University of Edinburgh, UK ## dnl Copyright (c) 1996-2001 ## dnl All Rights Reserved. ## dnl ## dnl Permission is hereby granted, free of charge, to use and distribute ## dnl this software and its documentation without restriction, including ## dnl without limitation the rights to use, copy, modify, merge, publish, ## dnl distribute, sublicense, and/or sell copies of this work, and to ## dnl permit persons to whom this work is furnished to do so, subject to ## dnl the following conditions: ## dnl 1. The code must retain the above copyright notice, this list of ## dnl conditions and the following disclaimer. ## dnl 2. Any modifications must be clearly marked as such. ## dnl 3. Original authors' names are not deleted. ## dnl 4. The authors' names are not used to endorse or promote products ## dnl derived from this software without specific prior written ## dnl permission. ## dnl ## dnl THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## dnl DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## dnl ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## dnl SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## dnl FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## dnl WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## dnl AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## dnl ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## dnl THIS SOFTWARE. ## dnl ## dnl######################################################################## AC_INIT(src/include/festival.h) AC_CANONICAL_SYSTEM AC_PROG_CC if test "x$GCC" = "xyes"; then CFLAGS="$CFLAGS -Wall" fi AC_PROG_RANLIB AC_CHECK_TOOL(AR, ar) AC_C_BIGENDIAN AC_OUTPUT(config/config) festival/config.sub0000755000175000017500000010541212254116520012763 0ustar awbawb#! /bin/sh # Configuration validation subroutine script. # Copyright 1992-2013 Free Software Foundation, Inc. timestamp='2013-10-01' # This file is free software; you can redistribute it and/or modify it # under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 3 of the License, or # (at your option) any later version. # # This program is distributed in the hope that it will be useful, but # WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU # General Public License for more details. # # You should have received a copy of the GNU General Public License # along with this program; if not, see . # # As a special exception to the GNU General Public License, if you # distribute this file as part of a program that contains a # configuration script generated by Autoconf, you may include it under # the same distribution terms that you use for the rest of that # program. This Exception is an additional permission under section 7 # of the GNU General Public License, version 3 ("GPLv3"). # Please send patches with a ChangeLog entry to config-patches@gnu.org. # # Configuration subroutine to validate and canonicalize a configuration type. # Supply the specified configuration type as an argument. # If it is invalid, we print an error message on stderr and exit with code 1. # Otherwise, we print the canonical config type on stdout and succeed. # You can get the latest version of this script from: # http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.sub;hb=HEAD # This file is supposed to be the same for all GNU packages # and recognize all the CPU types, system types and aliases # that are meaningful with *any* GNU software. # Each package is responsible for reporting which valid configurations # it does not support. The user should be able to distinguish # a failure to support a valid configuration from a meaningless # configuration. # The goal of this file is to map all the various variations of a given # machine specification into a single specification in the form: # CPU_TYPE-MANUFACTURER-OPERATING_SYSTEM # or in some cases, the newer four-part form: # CPU_TYPE-MANUFACTURER-KERNEL-OPERATING_SYSTEM # It is wrong to echo any other type of specification. me=`echo "$0" | sed -e 's,.*/,,'` usage="\ Usage: $0 [OPTION] CPU-MFR-OPSYS $0 [OPTION] ALIAS Canonicalize a configuration name. Operation modes: -h, --help print this help, then exit -t, --time-stamp print date of last modification, then exit -v, --version print version number, then exit Report bugs and patches to ." version="\ GNU config.sub ($timestamp) Copyright 1992-2013 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE." help=" Try \`$me --help' for more information." # Parse command line while test $# -gt 0 ; do case $1 in --time-stamp | --time* | -t ) echo "$timestamp" ; exit ;; --version | -v ) echo "$version" ; exit ;; --help | --h* | -h ) echo "$usage"; exit ;; -- ) # Stop option processing shift; break ;; - ) # Use stdin as input. break ;; -* ) echo "$me: invalid option $1$help" exit 1 ;; *local*) # First pass through any local machine types. echo $1 exit ;; * ) break ;; esac done case $# in 0) echo "$me: missing argument$help" >&2 exit 1;; 1) ;; *) echo "$me: too many arguments$help" >&2 exit 1;; esac # Separate what the user gave into CPU-COMPANY and OS or KERNEL-OS (if any). # Here we must recognize all the valid KERNEL-OS combinations. maybe_os=`echo $1 | sed 's/^\(.*\)-\([^-]*-[^-]*\)$/\2/'` case $maybe_os in nto-qnx* | linux-gnu* | linux-android* | linux-dietlibc | linux-newlib* | \ linux-musl* | linux-uclibc* | uclinux-uclibc* | uclinux-gnu* | kfreebsd*-gnu* | \ knetbsd*-gnu* | netbsd*-gnu* | \ kopensolaris*-gnu* | \ storm-chaos* | os2-emx* | rtmk-nova*) os=-$maybe_os basic_machine=`echo $1 | sed 's/^\(.*\)-\([^-]*-[^-]*\)$/\1/'` ;; android-linux) os=-linux-android basic_machine=`echo $1 | sed 's/^\(.*\)-\([^-]*-[^-]*\)$/\1/'`-unknown ;; *) basic_machine=`echo $1 | sed 's/-[^-]*$//'` if [ $basic_machine != $1 ] then os=`echo $1 | sed 's/.*-/-/'` else os=; fi ;; esac ### Let's recognize common machines as not being operating systems so ### that things like config.sub decstation-3100 work. We also ### recognize some manufacturers as not being operating systems, so we ### can provide default operating systems below. case $os in -sun*os*) # Prevent following clause from handling this invalid input. ;; -dec* | -mips* | -sequent* | -encore* | -pc532* | -sgi* | -sony* | \ -att* | -7300* | -3300* | -delta* | -motorola* | -sun[234]* | \ -unicom* | -ibm* | -next | -hp | -isi* | -apollo | -altos* | \ -convergent* | -ncr* | -news | -32* | -3600* | -3100* | -hitachi* |\ -c[123]* | -convex* | -sun | -crds | -omron* | -dg | -ultra | -tti* | \ -harris | -dolphin | -highlevel | -gould | -cbm | -ns | -masscomp | \ -apple | -axis | -knuth | -cray | -microblaze*) os= basic_machine=$1 ;; -bluegene*) os=-cnk ;; -sim | -cisco | -oki | -wec | -winbond) os= basic_machine=$1 ;; -scout) ;; -wrs) os=-vxworks basic_machine=$1 ;; -chorusos*) os=-chorusos basic_machine=$1 ;; -chorusrdb) os=-chorusrdb basic_machine=$1 ;; -hiux*) os=-hiuxwe2 ;; -sco6) os=-sco5v6 basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` ;; -sco5) os=-sco3.2v5 basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` ;; -sco4) os=-sco3.2v4 basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` ;; -sco3.2.[4-9]*) os=`echo $os | sed -e 's/sco3.2./sco3.2v/'` basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` ;; -sco3.2v[4-9]*) # Don't forget version if it is 3.2v4 or newer. basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` ;; -sco5v6*) # Don't forget version if it is 3.2v4 or newer. basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` ;; -sco*) os=-sco3.2v2 basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` ;; -udk*) basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` ;; -isc) os=-isc2.2 basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` ;; -clix*) basic_machine=clipper-intergraph ;; -isc*) basic_machine=`echo $1 | sed -e 's/86-.*/86-pc/'` ;; -lynx*178) os=-lynxos178 ;; -lynx*5) os=-lynxos5 ;; -lynx*) os=-lynxos ;; -ptx*) basic_machine=`echo $1 | sed -e 's/86-.*/86-sequent/'` ;; -windowsnt*) os=`echo $os | sed -e 's/windowsnt/winnt/'` ;; -psos*) os=-psos ;; -mint | -mint[0-9]*) basic_machine=m68k-atari os=-mint ;; esac # Decode aliases for certain CPU-COMPANY combinations. case $basic_machine in # Recognize the basic CPU types without company name. # Some are omitted here because they have special meanings below. 1750a | 580 \ | a29k \ | aarch64 | aarch64_be \ | alpha | alphaev[4-8] | alphaev56 | alphaev6[78] | alphapca5[67] \ | alpha64 | alpha64ev[4-8] | alpha64ev56 | alpha64ev6[78] | alpha64pca5[67] \ | am33_2.0 \ | arc | arceb \ | arm | arm[bl]e | arme[lb] | armv[2-8] | armv[3-8][lb] | armv7[arm] \ | avr | avr32 \ | be32 | be64 \ | bfin \ | c4x | c8051 | clipper \ | d10v | d30v | dlx | dsp16xx \ | epiphany \ | fido | fr30 | frv \ | h8300 | h8500 | hppa | hppa1.[01] | hppa2.0 | hppa2.0[nw] | hppa64 \ | hexagon \ | i370 | i860 | i960 | ia64 \ | ip2k | iq2000 \ | k1om \ | le32 | le64 \ | lm32 \ | m32c | m32r | m32rle | m68000 | m68k | m88k \ | maxq | mb | microblaze | microblazeel | mcore | mep | metag \ | mips | mipsbe | mipseb | mipsel | mipsle \ | mips16 \ | mips64 | mips64el \ | mips64octeon | mips64octeonel \ | mips64orion | mips64orionel \ | mips64r5900 | mips64r5900el \ | mips64vr | mips64vrel \ | mips64vr4100 | mips64vr4100el \ | mips64vr4300 | mips64vr4300el \ | mips64vr5000 | mips64vr5000el \ | mips64vr5900 | mips64vr5900el \ | mipsisa32 | mipsisa32el \ | mipsisa32r2 | mipsisa32r2el \ | mipsisa64 | mipsisa64el \ | mipsisa64r2 | mipsisa64r2el \ | mipsisa64sb1 | mipsisa64sb1el \ | mipsisa64sr71k | mipsisa64sr71kel \ | mipsr5900 | mipsr5900el \ | mipstx39 | mipstx39el \ | mn10200 | mn10300 \ | moxie \ | mt \ | msp430 \ | nds32 | nds32le | nds32be \ | nios | nios2 | nios2eb | nios2el \ | ns16k | ns32k \ | open8 \ | or1k | or32 \ | pdp10 | pdp11 | pj | pjl \ | powerpc | powerpc64 | powerpc64le | powerpcle \ | pyramid \ | rl78 | rx \ | score \ | sh | sh[1234] | sh[24]a | sh[24]aeb | sh[23]e | sh[34]eb | sheb | shbe | shle | sh[1234]le | sh3ele \ | sh64 | sh64le \ | sparc | sparc64 | sparc64b | sparc64v | sparc86x | sparclet | sparclite \ | sparcv8 | sparcv9 | sparcv9b | sparcv9v \ | spu \ | tahoe | tic4x | tic54x | tic55x | tic6x | tic80 | tron \ | ubicom32 \ | v850 | v850e | v850e1 | v850e2 | v850es | v850e2v3 \ | we32k \ | x86 | xc16x | xstormy16 | xtensa \ | z8k | z80) basic_machine=$basic_machine-unknown ;; c54x) basic_machine=tic54x-unknown ;; c55x) basic_machine=tic55x-unknown ;; c6x) basic_machine=tic6x-unknown ;; m6811 | m68hc11 | m6812 | m68hc12 | m68hcs12x | nvptx | picochip) basic_machine=$basic_machine-unknown os=-none ;; m88110 | m680[12346]0 | m683?2 | m68360 | m5200 | v70 | w65 | z8k) ;; ms1) basic_machine=mt-unknown ;; strongarm | thumb | xscale) basic_machine=arm-unknown ;; xgate) basic_machine=$basic_machine-unknown os=-none ;; xscaleeb) basic_machine=armeb-unknown ;; xscaleel) basic_machine=armel-unknown ;; # We use `pc' rather than `unknown' # because (1) that's what they normally are, and # (2) the word "unknown" tends to confuse beginning users. i*86 | x86_64) basic_machine=$basic_machine-pc ;; # Object if more than one company name word. *-*-*) echo Invalid configuration \`$1\': machine \`$basic_machine\' not recognized 1>&2 exit 1 ;; # Recognize the basic CPU types with company name. 580-* \ | a29k-* \ | aarch64-* | aarch64_be-* \ | alpha-* | alphaev[4-8]-* | alphaev56-* | alphaev6[78]-* \ | alpha64-* | alpha64ev[4-8]-* | alpha64ev56-* | alpha64ev6[78]-* \ | alphapca5[67]-* | alpha64pca5[67]-* | arc-* | arceb-* \ | arm-* | armbe-* | armle-* | armeb-* | armv*-* \ | avr-* | avr32-* \ | be32-* | be64-* \ | bfin-* | bs2000-* \ | c[123]* | c30-* | [cjt]90-* | c4x-* \ | c8051-* | clipper-* | craynv-* | cydra-* \ | d10v-* | d30v-* | dlx-* \ | elxsi-* \ | f30[01]-* | f700-* | fido-* | fr30-* | frv-* | fx80-* \ | h8300-* | h8500-* \ | hppa-* | hppa1.[01]-* | hppa2.0-* | hppa2.0[nw]-* | hppa64-* \ | hexagon-* \ | i*86-* | i860-* | i960-* | ia64-* \ | ip2k-* | iq2000-* \ | k1om-* \ | le32-* | le64-* \ | lm32-* \ | m32c-* | m32r-* | m32rle-* \ | m68000-* | m680[012346]0-* | m68360-* | m683?2-* | m68k-* \ | m88110-* | m88k-* | maxq-* | mcore-* | metag-* \ | microblaze-* | microblazeel-* \ | mips-* | mipsbe-* | mipseb-* | mipsel-* | mipsle-* \ | mips16-* \ | mips64-* | mips64el-* \ | mips64octeon-* | mips64octeonel-* \ | mips64orion-* | mips64orionel-* \ | mips64r5900-* | mips64r5900el-* \ | mips64vr-* | mips64vrel-* \ | mips64vr4100-* | mips64vr4100el-* \ | mips64vr4300-* | mips64vr4300el-* \ | mips64vr5000-* | mips64vr5000el-* \ | mips64vr5900-* | mips64vr5900el-* \ | mipsisa32-* | mipsisa32el-* \ | mipsisa32r2-* | mipsisa32r2el-* \ | mipsisa64-* | mipsisa64el-* \ | mipsisa64r2-* | mipsisa64r2el-* \ | mipsisa64sb1-* | mipsisa64sb1el-* \ | mipsisa64sr71k-* | mipsisa64sr71kel-* \ | mipsr5900-* | mipsr5900el-* \ | mipstx39-* | mipstx39el-* \ | mmix-* \ | mt-* \ | msp430-* \ | nds32-* | nds32le-* | nds32be-* \ | nios-* | nios2-* | nios2eb-* | nios2el-* \ | none-* | np1-* | ns16k-* | ns32k-* \ | open8-* \ | orion-* \ | pdp10-* | pdp11-* | pj-* | pjl-* | pn-* | power-* \ | powerpc-* | powerpc64-* | powerpc64le-* | powerpcle-* \ | pyramid-* \ | rl78-* | romp-* | rs6000-* | rx-* \ | sh-* | sh[1234]-* | sh[24]a-* | sh[24]aeb-* | sh[23]e-* | sh[34]eb-* | sheb-* | shbe-* \ | shle-* | sh[1234]le-* | sh3ele-* | sh64-* | sh64le-* \ | sparc-* | sparc64-* | sparc64b-* | sparc64v-* | sparc86x-* | sparclet-* \ | sparclite-* \ | sparcv8-* | sparcv9-* | sparcv9b-* | sparcv9v-* | sv1-* | sx?-* \ | tahoe-* \ | tic30-* | tic4x-* | tic54x-* | tic55x-* | tic6x-* | tic80-* \ | tile*-* \ | tron-* \ | ubicom32-* \ | v850-* | v850e-* | v850e1-* | v850es-* | v850e2-* | v850e2v3-* \ | vax-* \ | we32k-* \ | x86-* | x86_64-* | xc16x-* | xps100-* \ | xstormy16-* | xtensa*-* \ | ymp-* \ | z8k-* | z80-*) ;; # Recognize the basic CPU types without company name, with glob match. xtensa*) basic_machine=$basic_machine-unknown ;; # Recognize the various machine names and aliases which stand # for a CPU type and a company and sometimes even an OS. 386bsd) basic_machine=i386-unknown os=-bsd ;; 3b1 | 7300 | 7300-att | att-7300 | pc7300 | safari | unixpc) basic_machine=m68000-att ;; 3b*) basic_machine=we32k-att ;; a29khif) basic_machine=a29k-amd os=-udi ;; abacus) basic_machine=abacus-unknown ;; adobe68k) basic_machine=m68010-adobe os=-scout ;; alliant | fx80) basic_machine=fx80-alliant ;; altos | altos3068) basic_machine=m68k-altos ;; am29k) basic_machine=a29k-none os=-bsd ;; amd64) basic_machine=x86_64-pc ;; amd64-*) basic_machine=x86_64-`echo $basic_machine | sed 's/^[^-]*-//'` ;; amdahl) basic_machine=580-amdahl os=-sysv ;; amiga | amiga-*) basic_machine=m68k-unknown ;; amigaos | amigados) basic_machine=m68k-unknown os=-amigaos ;; amigaunix | amix) basic_machine=m68k-unknown os=-sysv4 ;; apollo68) basic_machine=m68k-apollo os=-sysv ;; apollo68bsd) basic_machine=m68k-apollo os=-bsd ;; aros) basic_machine=i386-pc os=-aros ;; aux) basic_machine=m68k-apple os=-aux ;; balance) basic_machine=ns32k-sequent os=-dynix ;; blackfin) basic_machine=bfin-unknown os=-linux ;; blackfin-*) basic_machine=bfin-`echo $basic_machine | sed 's/^[^-]*-//'` os=-linux ;; bluegene*) basic_machine=powerpc-ibm os=-cnk ;; c54x-*) basic_machine=tic54x-`echo $basic_machine | sed 's/^[^-]*-//'` ;; c55x-*) basic_machine=tic55x-`echo $basic_machine | sed 's/^[^-]*-//'` ;; c6x-*) basic_machine=tic6x-`echo $basic_machine | sed 's/^[^-]*-//'` ;; c90) basic_machine=c90-cray os=-unicos ;; cegcc) basic_machine=arm-unknown os=-cegcc ;; convex-c1) basic_machine=c1-convex os=-bsd ;; convex-c2) basic_machine=c2-convex os=-bsd ;; convex-c32) basic_machine=c32-convex os=-bsd ;; convex-c34) basic_machine=c34-convex os=-bsd ;; convex-c38) basic_machine=c38-convex os=-bsd ;; cray | j90) basic_machine=j90-cray os=-unicos ;; craynv) basic_machine=craynv-cray os=-unicosmp ;; cr16 | cr16-*) basic_machine=cr16-unknown os=-elf ;; crds | unos) basic_machine=m68k-crds ;; crisv32 | crisv32-* | etraxfs*) basic_machine=crisv32-axis ;; cris | cris-* | etrax*) basic_machine=cris-axis ;; crx) basic_machine=crx-unknown os=-elf ;; da30 | da30-*) basic_machine=m68k-da30 ;; decstation | decstation-3100 | pmax | pmax-* | pmin | dec3100 | decstatn) basic_machine=mips-dec ;; decsystem10* | dec10*) basic_machine=pdp10-dec os=-tops10 ;; decsystem20* | dec20*) basic_machine=pdp10-dec os=-tops20 ;; delta | 3300 | motorola-3300 | motorola-delta \ | 3300-motorola | delta-motorola) basic_machine=m68k-motorola ;; delta88) basic_machine=m88k-motorola os=-sysv3 ;; dicos) basic_machine=i686-pc os=-dicos ;; djgpp) basic_machine=i586-pc os=-msdosdjgpp ;; dpx20 | dpx20-*) basic_machine=rs6000-bull os=-bosx ;; dpx2* | dpx2*-bull) basic_machine=m68k-bull os=-sysv3 ;; ebmon29k) basic_machine=a29k-amd os=-ebmon ;; elxsi) basic_machine=elxsi-elxsi os=-bsd ;; encore | umax | mmax) basic_machine=ns32k-encore ;; es1800 | OSE68k | ose68k | ose | OSE) basic_machine=m68k-ericsson os=-ose ;; fx2800) basic_machine=i860-alliant ;; genix) basic_machine=ns32k-ns ;; gmicro) basic_machine=tron-gmicro os=-sysv ;; go32) basic_machine=i386-pc os=-go32 ;; h3050r* | hiux*) basic_machine=hppa1.1-hitachi os=-hiuxwe2 ;; h8300hms) basic_machine=h8300-hitachi os=-hms ;; h8300xray) basic_machine=h8300-hitachi os=-xray ;; h8500hms) basic_machine=h8500-hitachi os=-hms ;; harris) basic_machine=m88k-harris os=-sysv3 ;; hp300-*) basic_machine=m68k-hp ;; hp300bsd) basic_machine=m68k-hp os=-bsd ;; hp300hpux) basic_machine=m68k-hp os=-hpux ;; hp3k9[0-9][0-9] | hp9[0-9][0-9]) basic_machine=hppa1.0-hp ;; hp9k2[0-9][0-9] | hp9k31[0-9]) basic_machine=m68000-hp ;; hp9k3[2-9][0-9]) basic_machine=m68k-hp ;; hp9k6[0-9][0-9] | hp6[0-9][0-9]) basic_machine=hppa1.0-hp ;; hp9k7[0-79][0-9] | hp7[0-79][0-9]) basic_machine=hppa1.1-hp ;; hp9k78[0-9] | hp78[0-9]) # FIXME: really hppa2.0-hp basic_machine=hppa1.1-hp ;; hp9k8[67]1 | hp8[67]1 | hp9k80[24] | hp80[24] | hp9k8[78]9 | hp8[78]9 | hp9k893 | hp893) # FIXME: really hppa2.0-hp basic_machine=hppa1.1-hp ;; hp9k8[0-9][13679] | hp8[0-9][13679]) basic_machine=hppa1.1-hp ;; hp9k8[0-9][0-9] | hp8[0-9][0-9]) basic_machine=hppa1.0-hp ;; hppa-next) os=-nextstep3 ;; hppaosf) basic_machine=hppa1.1-hp os=-osf ;; hppro) basic_machine=hppa1.1-hp os=-proelf ;; i370-ibm* | ibm*) basic_machine=i370-ibm ;; i*86v32) basic_machine=`echo $1 | sed -e 's/86.*/86-pc/'` os=-sysv32 ;; i*86v4*) basic_machine=`echo $1 | sed -e 's/86.*/86-pc/'` os=-sysv4 ;; i*86v) basic_machine=`echo $1 | sed -e 's/86.*/86-pc/'` os=-sysv ;; i*86sol2) basic_machine=`echo $1 | sed -e 's/86.*/86-pc/'` os=-solaris2 ;; i386mach) basic_machine=i386-mach os=-mach ;; i386-vsta | vsta) basic_machine=i386-unknown os=-vsta ;; iris | iris4d) basic_machine=mips-sgi case $os in -irix*) ;; *) os=-irix4 ;; esac ;; isi68 | isi) basic_machine=m68k-isi os=-sysv ;; m68knommu) basic_machine=m68k-unknown os=-linux ;; m68knommu-*) basic_machine=m68k-`echo $basic_machine | sed 's/^[^-]*-//'` os=-linux ;; m88k-omron*) basic_machine=m88k-omron ;; magnum | m3230) basic_machine=mips-mips os=-sysv ;; merlin) basic_machine=ns32k-utek os=-sysv ;; microblaze*) basic_machine=microblaze-xilinx ;; mingw64) basic_machine=x86_64-pc os=-mingw64 ;; mingw32) basic_machine=i686-pc os=-mingw32 ;; mingw32ce) basic_machine=arm-unknown os=-mingw32ce ;; miniframe) basic_machine=m68000-convergent ;; *mint | -mint[0-9]* | *MiNT | *MiNT[0-9]*) basic_machine=m68k-atari os=-mint ;; mips3*-*) basic_machine=`echo $basic_machine | sed -e 's/mips3/mips64/'` ;; mips3*) basic_machine=`echo $basic_machine | sed -e 's/mips3/mips64/'`-unknown ;; monitor) basic_machine=m68k-rom68k os=-coff ;; morphos) basic_machine=powerpc-unknown os=-morphos ;; msdos) basic_machine=i386-pc os=-msdos ;; ms1-*) basic_machine=`echo $basic_machine | sed -e 's/ms1-/mt-/'` ;; msys) basic_machine=i686-pc os=-msys ;; mvs) basic_machine=i370-ibm os=-mvs ;; nacl) basic_machine=le32-unknown os=-nacl ;; ncr3000) basic_machine=i486-ncr os=-sysv4 ;; netbsd386) basic_machine=i386-unknown os=-netbsd ;; netwinder) basic_machine=armv4l-rebel os=-linux ;; news | news700 | news800 | news900) basic_machine=m68k-sony os=-newsos ;; news1000) basic_machine=m68030-sony os=-newsos ;; news-3600 | risc-news) basic_machine=mips-sony os=-newsos ;; necv70) basic_machine=v70-nec os=-sysv ;; next | m*-next ) basic_machine=m68k-next case $os in -nextstep* ) ;; -ns2*) os=-nextstep2 ;; *) os=-nextstep3 ;; esac ;; nh3000) basic_machine=m68k-harris os=-cxux ;; nh[45]000) basic_machine=m88k-harris os=-cxux ;; nindy960) basic_machine=i960-intel os=-nindy ;; mon960) basic_machine=i960-intel os=-mon960 ;; nonstopux) basic_machine=mips-compaq os=-nonstopux ;; np1) basic_machine=np1-gould ;; neo-tandem) basic_machine=neo-tandem ;; nse-tandem) basic_machine=nse-tandem ;; nsr-tandem) basic_machine=nsr-tandem ;; op50n-* | op60c-*) basic_machine=hppa1.1-oki os=-proelf ;; openrisc | openrisc-*) basic_machine=or32-unknown ;; os400) basic_machine=powerpc-ibm os=-os400 ;; OSE68000 | ose68000) basic_machine=m68000-ericsson os=-ose ;; os68k) basic_machine=m68k-none os=-os68k ;; pa-hitachi) basic_machine=hppa1.1-hitachi os=-hiuxwe2 ;; paragon) basic_machine=i860-intel os=-osf ;; parisc) basic_machine=hppa-unknown os=-linux ;; parisc-*) basic_machine=hppa-`echo $basic_machine | sed 's/^[^-]*-//'` os=-linux ;; pbd) basic_machine=sparc-tti ;; pbb) basic_machine=m68k-tti ;; pc532 | pc532-*) basic_machine=ns32k-pc532 ;; pc98) basic_machine=i386-pc ;; pc98-*) basic_machine=i386-`echo $basic_machine | sed 's/^[^-]*-//'` ;; pentium | p5 | k5 | k6 | nexgen | viac3) basic_machine=i586-pc ;; pentiumpro | p6 | 6x86 | athlon | athlon_*) basic_machine=i686-pc ;; pentiumii | pentium2 | pentiumiii | pentium3) basic_machine=i686-pc ;; pentium4) basic_machine=i786-pc ;; pentium-* | p5-* | k5-* | k6-* | nexgen-* | viac3-*) basic_machine=i586-`echo $basic_machine | sed 's/^[^-]*-//'` ;; pentiumpro-* | p6-* | 6x86-* | athlon-*) basic_machine=i686-`echo $basic_machine | sed 's/^[^-]*-//'` ;; pentiumii-* | pentium2-* | pentiumiii-* | pentium3-*) basic_machine=i686-`echo $basic_machine | sed 's/^[^-]*-//'` ;; pentium4-*) basic_machine=i786-`echo $basic_machine | sed 's/^[^-]*-//'` ;; pn) basic_machine=pn-gould ;; power) basic_machine=power-ibm ;; ppc | ppcbe) basic_machine=powerpc-unknown ;; ppc-* | ppcbe-*) basic_machine=powerpc-`echo $basic_machine | sed 's/^[^-]*-//'` ;; ppcle | powerpclittle | ppc-le | powerpc-little) basic_machine=powerpcle-unknown ;; ppcle-* | powerpclittle-*) basic_machine=powerpcle-`echo $basic_machine | sed 's/^[^-]*-//'` ;; ppc64) basic_machine=powerpc64-unknown ;; ppc64-*) basic_machine=powerpc64-`echo $basic_machine | sed 's/^[^-]*-//'` ;; ppc64le | powerpc64little | ppc64-le | powerpc64-little) basic_machine=powerpc64le-unknown ;; ppc64le-* | powerpc64little-*) basic_machine=powerpc64le-`echo $basic_machine | sed 's/^[^-]*-//'` ;; ps2) basic_machine=i386-ibm ;; pw32) basic_machine=i586-unknown os=-pw32 ;; rdos | rdos64) basic_machine=x86_64-pc os=-rdos ;; rdos32) basic_machine=i386-pc os=-rdos ;; rom68k) basic_machine=m68k-rom68k os=-coff ;; rm[46]00) basic_machine=mips-siemens ;; rtpc | rtpc-*) basic_machine=romp-ibm ;; s390 | s390-*) basic_machine=s390-ibm ;; s390x | s390x-*) basic_machine=s390x-ibm ;; sa29200) basic_machine=a29k-amd os=-udi ;; sb1) basic_machine=mipsisa64sb1-unknown ;; sb1el) basic_machine=mipsisa64sb1el-unknown ;; sde) basic_machine=mipsisa32-sde os=-elf ;; sei) basic_machine=mips-sei os=-seiux ;; sequent) basic_machine=i386-sequent ;; sh) basic_machine=sh-hitachi os=-hms ;; sh5el) basic_machine=sh5le-unknown ;; sh64) basic_machine=sh64-unknown ;; sparclite-wrs | simso-wrs) basic_machine=sparclite-wrs os=-vxworks ;; sps7) basic_machine=m68k-bull os=-sysv2 ;; spur) basic_machine=spur-unknown ;; st2000) basic_machine=m68k-tandem ;; stratus) basic_machine=i860-stratus os=-sysv4 ;; strongarm-* | thumb-*) basic_machine=arm-`echo $basic_machine | sed 's/^[^-]*-//'` ;; sun2) basic_machine=m68000-sun ;; sun2os3) basic_machine=m68000-sun os=-sunos3 ;; sun2os4) basic_machine=m68000-sun os=-sunos4 ;; sun3os3) basic_machine=m68k-sun os=-sunos3 ;; sun3os4) basic_machine=m68k-sun os=-sunos4 ;; sun4os3) basic_machine=sparc-sun os=-sunos3 ;; sun4os4) basic_machine=sparc-sun os=-sunos4 ;; sun4sol2) basic_machine=sparc-sun os=-solaris2 ;; sun3 | sun3-*) basic_machine=m68k-sun ;; sun4) basic_machine=sparc-sun ;; sun386 | sun386i | roadrunner) basic_machine=i386-sun ;; sv1) basic_machine=sv1-cray os=-unicos ;; symmetry) basic_machine=i386-sequent os=-dynix ;; t3e) basic_machine=alphaev5-cray os=-unicos ;; t90) basic_machine=t90-cray os=-unicos ;; tile*) basic_machine=$basic_machine-unknown os=-linux-gnu ;; tx39) basic_machine=mipstx39-unknown ;; tx39el) basic_machine=mipstx39el-unknown ;; toad1) basic_machine=pdp10-xkl os=-tops20 ;; tower | tower-32) basic_machine=m68k-ncr ;; tpf) basic_machine=s390x-ibm os=-tpf ;; udi29k) basic_machine=a29k-amd os=-udi ;; ultra3) basic_machine=a29k-nyu os=-sym1 ;; v810 | necv810) basic_machine=v810-nec os=-none ;; vaxv) basic_machine=vax-dec os=-sysv ;; vms) basic_machine=vax-dec os=-vms ;; vpp*|vx|vx-*) basic_machine=f301-fujitsu ;; vxworks960) basic_machine=i960-wrs os=-vxworks ;; vxworks68) basic_machine=m68k-wrs os=-vxworks ;; vxworks29k) basic_machine=a29k-wrs os=-vxworks ;; w65*) basic_machine=w65-wdc os=-none ;; w89k-*) basic_machine=hppa1.1-winbond os=-proelf ;; xbox) basic_machine=i686-pc os=-mingw32 ;; xps | xps100) basic_machine=xps100-honeywell ;; xscale-* | xscalee[bl]-*) basic_machine=`echo $basic_machine | sed 's/^xscale/arm/'` ;; ymp) basic_machine=ymp-cray os=-unicos ;; z8k-*-coff) basic_machine=z8k-unknown os=-sim ;; z80-*-coff) basic_machine=z80-unknown os=-sim ;; none) basic_machine=none-none os=-none ;; # Here we handle the default manufacturer of certain CPU types. It is in # some cases the only manufacturer, in others, it is the most popular. w89k) basic_machine=hppa1.1-winbond ;; op50n) basic_machine=hppa1.1-oki ;; op60c) basic_machine=hppa1.1-oki ;; romp) basic_machine=romp-ibm ;; mmix) basic_machine=mmix-knuth ;; rs6000) basic_machine=rs6000-ibm ;; vax) basic_machine=vax-dec ;; pdp10) # there are many clones, so DEC is not a safe bet basic_machine=pdp10-unknown ;; pdp11) basic_machine=pdp11-dec ;; we32k) basic_machine=we32k-att ;; sh[1234] | sh[24]a | sh[24]aeb | sh[34]eb | sh[1234]le | sh[23]ele) basic_machine=sh-unknown ;; sparc | sparcv8 | sparcv9 | sparcv9b | sparcv9v) basic_machine=sparc-sun ;; cydra) basic_machine=cydra-cydrome ;; orion) basic_machine=orion-highlevel ;; orion105) basic_machine=clipper-highlevel ;; mac | mpw | mac-mpw) basic_machine=m68k-apple ;; pmac | pmac-mpw) basic_machine=powerpc-apple ;; *-unknown) # Make sure to match an already-canonicalized machine name. ;; *) echo Invalid configuration \`$1\': machine \`$basic_machine\' not recognized 1>&2 exit 1 ;; esac # Here we canonicalize certain aliases for manufacturers. case $basic_machine in *-digital*) basic_machine=`echo $basic_machine | sed 's/digital.*/dec/'` ;; *-commodore*) basic_machine=`echo $basic_machine | sed 's/commodore.*/cbm/'` ;; *) ;; esac # Decode manufacturer-specific aliases for certain operating systems. if [ x"$os" != x"" ] then case $os in # First match some system type aliases # that might get confused with valid system types. # -solaris* is a basic system type, with this one exception. -auroraux) os=-auroraux ;; -solaris1 | -solaris1.*) os=`echo $os | sed -e 's|solaris1|sunos4|'` ;; -solaris) os=-solaris2 ;; -svr4*) os=-sysv4 ;; -unixware*) os=-sysv4.2uw ;; -gnu/linux*) os=`echo $os | sed -e 's|gnu/linux|linux-gnu|'` ;; # First accept the basic system types. # The portable systems comes first. # Each alternative MUST END IN A *, to match a version number. # -sysv* is not here because it comes later, after sysvr4. -gnu* | -bsd* | -mach* | -minix* | -genix* | -ultrix* | -irix* \ | -*vms* | -sco* | -esix* | -isc* | -aix* | -cnk* | -sunos | -sunos[34]*\ | -hpux* | -unos* | -osf* | -luna* | -dgux* | -auroraux* | -solaris* \ | -sym* | -kopensolaris* | -plan9* \ | -amigaos* | -amigados* | -msdos* | -newsos* | -unicos* | -aof* \ | -aos* | -aros* \ | -nindy* | -vxsim* | -vxworks* | -ebmon* | -hms* | -mvs* \ | -clix* | -riscos* | -uniplus* | -iris* | -rtu* | -xenix* \ | -hiux* | -386bsd* | -knetbsd* | -mirbsd* | -netbsd* \ | -bitrig* | -openbsd* | -solidbsd* \ | -ekkobsd* | -kfreebsd* | -freebsd* | -riscix* | -lynxos* \ | -bosx* | -nextstep* | -cxux* | -aout* | -elf* | -oabi* \ | -ptx* | -coff* | -ecoff* | -winnt* | -domain* | -vsta* \ | -udi* | -eabi* | -lites* | -ieee* | -go32* | -aux* \ | -chorusos* | -chorusrdb* | -cegcc* \ | -cygwin* | -msys* | -pe* | -psos* | -moss* | -proelf* | -rtems* \ | -mingw32* | -mingw64* | -linux-gnu* | -linux-android* \ | -linux-newlib* | -linux-musl* | -linux-uclibc* \ | -uxpv* | -beos* | -mpeix* | -udk* \ | -interix* | -uwin* | -mks* | -rhapsody* | -darwin* | -opened* \ | -openstep* | -oskit* | -conix* | -pw32* | -nonstopux* \ | -storm-chaos* | -tops10* | -tenex* | -tops20* | -its* \ | -os2* | -vos* | -palmos* | -uclinux* | -nucleus* \ | -morphos* | -superux* | -rtmk* | -rtmk-nova* | -windiss* \ | -powermax* | -dnix* | -nx6 | -nx7 | -sei* | -dragonfly* \ | -skyos* | -haiku* | -rdos* | -toppers* | -drops* | -es*) # Remember, each alternative MUST END IN *, to match a version number. ;; -qnx*) case $basic_machine in x86-* | i*86-*) ;; *) os=-nto$os ;; esac ;; -nto-qnx*) ;; -nto*) os=`echo $os | sed -e 's|nto|nto-qnx|'` ;; -sim | -es1800* | -hms* | -xray | -os68k* | -none* | -v88r* \ | -windows* | -osx | -abug | -netware* | -os9* | -beos* | -haiku* \ | -macos* | -mpw* | -magic* | -mmixware* | -mon960* | -lnews*) ;; -mac*) os=`echo $os | sed -e 's|mac|macos|'` ;; -linux-dietlibc) os=-linux-dietlibc ;; -linux*) os=`echo $os | sed -e 's|linux|linux-gnu|'` ;; -sunos5*) os=`echo $os | sed -e 's|sunos5|solaris2|'` ;; -sunos6*) os=`echo $os | sed -e 's|sunos6|solaris3|'` ;; -opened*) os=-openedition ;; -os400*) os=-os400 ;; -wince*) os=-wince ;; -osfrose*) os=-osfrose ;; -osf*) os=-osf ;; -utek*) os=-bsd ;; -dynix*) os=-bsd ;; -acis*) os=-aos ;; -atheos*) os=-atheos ;; -syllable*) os=-syllable ;; -386bsd) os=-bsd ;; -ctix* | -uts*) os=-sysv ;; -nova*) os=-rtmk-nova ;; -ns2 ) os=-nextstep2 ;; -nsk*) os=-nsk ;; # Preserve the version number of sinix5. -sinix5.*) os=`echo $os | sed -e 's|sinix|sysv|'` ;; -sinix*) os=-sysv4 ;; -tpf*) os=-tpf ;; -triton*) os=-sysv3 ;; -oss*) os=-sysv3 ;; -svr4) os=-sysv4 ;; -svr3) os=-sysv3 ;; -sysvr4) os=-sysv4 ;; # This must come after -sysvr4. -sysv*) ;; -ose*) os=-ose ;; -es1800*) os=-ose ;; -xenix) os=-xenix ;; -*mint | -mint[0-9]* | -*MiNT | -MiNT[0-9]*) os=-mint ;; -aros*) os=-aros ;; -zvmoe) os=-zvmoe ;; -dicos*) os=-dicos ;; -nacl*) ;; -none) ;; *) # Get rid of the `-' at the beginning of $os. os=`echo $os | sed 's/[^-]*-//'` echo Invalid configuration \`$1\': system \`$os\' not recognized 1>&2 exit 1 ;; esac else # Here we handle the default operating systems that come with various machines. # The value should be what the vendor currently ships out the door with their # machine or put another way, the most popular os provided with the machine. # Note that if you're going to try to match "-MANUFACTURER" here (say, # "-sun"), then you have to tell the case statement up towards the top # that MANUFACTURER isn't an operating system. Otherwise, code above # will signal an error saying that MANUFACTURER isn't an operating # system, and we'll never get to this point. case $basic_machine in score-*) os=-elf ;; spu-*) os=-elf ;; *-acorn) os=-riscix1.2 ;; arm*-rebel) os=-linux ;; arm*-semi) os=-aout ;; c4x-* | tic4x-*) os=-coff ;; c8051-*) os=-elf ;; hexagon-*) os=-elf ;; tic54x-*) os=-coff ;; tic55x-*) os=-coff ;; tic6x-*) os=-coff ;; # This must come before the *-dec entry. pdp10-*) os=-tops20 ;; pdp11-*) os=-none ;; *-dec | vax-*) os=-ultrix4.2 ;; m68*-apollo) os=-domain ;; i386-sun) os=-sunos4.0.2 ;; m68000-sun) os=-sunos3 ;; m68*-cisco) os=-aout ;; mep-*) os=-elf ;; mips*-cisco) os=-elf ;; mips*-*) os=-elf ;; or1k-*) os=-elf ;; or32-*) os=-coff ;; *-tti) # must be before sparc entry or we get the wrong os. os=-sysv3 ;; sparc-* | *-sun) os=-sunos4.1.1 ;; *-be) os=-beos ;; *-haiku) os=-haiku ;; *-ibm) os=-aix ;; *-knuth) os=-mmixware ;; *-wec) os=-proelf ;; *-winbond) os=-proelf ;; *-oki) os=-proelf ;; *-hp) os=-hpux ;; *-hitachi) os=-hiux ;; i860-* | *-att | *-ncr | *-altos | *-motorola | *-convergent) os=-sysv ;; *-cbm) os=-amigaos ;; *-dg) os=-dgux ;; *-dolphin) os=-sysv3 ;; m68k-ccur) os=-rtu ;; m88k-omron*) os=-luna ;; *-next ) os=-nextstep ;; *-sequent) os=-ptx ;; *-crds) os=-unos ;; *-ns) os=-genix ;; i370-*) os=-mvs ;; *-next) os=-nextstep3 ;; *-gould) os=-sysv ;; *-highlevel) os=-bsd ;; *-encore) os=-bsd ;; *-sgi) os=-irix ;; *-siemens) os=-sysv4 ;; *-masscomp) os=-rtu ;; f30[01]-fujitsu | f700-fujitsu) os=-uxpv ;; *-rom68k) os=-coff ;; *-*bug) os=-coff ;; *-apple) os=-macos ;; *-atari*) os=-mint ;; *) os=-none ;; esac fi # Here we handle the case where we know the os, and the CPU type, but not the # manufacturer. We pick the logical manufacturer. vendor=unknown case $basic_machine in *-unknown) case $os in -riscix*) vendor=acorn ;; -sunos*) vendor=sun ;; -cnk*|-aix*) vendor=ibm ;; -beos*) vendor=be ;; -hpux*) vendor=hp ;; -mpeix*) vendor=hp ;; -hiux*) vendor=hitachi ;; -unos*) vendor=crds ;; -dgux*) vendor=dg ;; -luna*) vendor=omron ;; -genix*) vendor=ns ;; -mvs* | -opened*) vendor=ibm ;; -os400*) vendor=ibm ;; -ptx*) vendor=sequent ;; -tpf*) vendor=ibm ;; -vxsim* | -vxworks* | -windiss*) vendor=wrs ;; -aux*) vendor=apple ;; -hms*) vendor=hitachi ;; -mpw* | -macos*) vendor=apple ;; -*mint | -mint[0-9]* | -*MiNT | -MiNT[0-9]*) vendor=atari ;; -vos*) vendor=stratus ;; esac basic_machine=`echo $basic_machine | sed "s/unknown/$vendor/"` ;; esac echo $basic_machine$os exit # Local variables: # eval: (add-hook 'write-file-hooks 'time-stamp) # time-stamp-start: "timestamp='" # time-stamp-format: "%:y-%02m-%02d" # time-stamp-end: "'" # End: festival/config.guess0000755000175000017500000013135512254116520013325 0ustar awbawb#! /bin/sh # Attempt to guess a canonical system name. # Copyright 1992-2013 Free Software Foundation, Inc. timestamp='2013-11-29' # This file is free software; you can redistribute it and/or modify it # under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 3 of the License, or # (at your option) any later version. # # This program is distributed in the hope that it will be useful, but # WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU # General Public License for more details. # # You should have received a copy of the GNU General Public License # along with this program; if not, see . # # As a special exception to the GNU General Public License, if you # distribute this file as part of a program that contains a # configuration script generated by Autoconf, you may include it under # the same distribution terms that you use for the rest of that # program. This Exception is an additional permission under section 7 # of the GNU General Public License, version 3 ("GPLv3"). # # Originally written by Per Bothner. # # You can get the latest version of this script from: # http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD # # Please send patches with a ChangeLog entry to config-patches@gnu.org. me=`echo "$0" | sed -e 's,.*/,,'` usage="\ Usage: $0 [OPTION] Output the configuration name of the system \`$me' is run on. Operation modes: -h, --help print this help, then exit -t, --time-stamp print date of last modification, then exit -v, --version print version number, then exit Report bugs and patches to ." version="\ GNU config.guess ($timestamp) Originally written by Per Bothner. Copyright 1992-2013 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE." help=" Try \`$me --help' for more information." # Parse command line while test $# -gt 0 ; do case $1 in --time-stamp | --time* | -t ) echo "$timestamp" ; exit ;; --version | -v ) echo "$version" ; exit ;; --help | --h* | -h ) echo "$usage"; exit ;; -- ) # Stop option processing shift; break ;; - ) # Use stdin as input. break ;; -* ) echo "$me: invalid option $1$help" >&2 exit 1 ;; * ) break ;; esac done if test $# != 0; then echo "$me: too many arguments$help" >&2 exit 1 fi trap 'exit 1' 1 2 15 # CC_FOR_BUILD -- compiler used by this script. Note that the use of a # compiler to aid in system detection is discouraged as it requires # temporary files to be created and, as you can see below, it is a # headache to deal with in a portable fashion. # Historically, `CC_FOR_BUILD' used to be named `HOST_CC'. We still # use `HOST_CC' if defined, but it is deprecated. # Portable tmp directory creation inspired by the Autoconf team. set_cc_for_build=' trap "exitcode=\$?; (rm -f \$tmpfiles 2>/dev/null; rmdir \$tmp 2>/dev/null) && exit \$exitcode" 0 ; trap "rm -f \$tmpfiles 2>/dev/null; rmdir \$tmp 2>/dev/null; exit 1" 1 2 13 15 ; : ${TMPDIR=/tmp} ; { tmp=`(umask 077 && mktemp -d "$TMPDIR/cgXXXXXX") 2>/dev/null` && test -n "$tmp" && test -d "$tmp" ; } || { test -n "$RANDOM" && tmp=$TMPDIR/cg$$-$RANDOM && (umask 077 && mkdir $tmp) ; } || { tmp=$TMPDIR/cg-$$ && (umask 077 && mkdir $tmp) && echo "Warning: creating insecure temp directory" >&2 ; } || { echo "$me: cannot create a temporary directory in $TMPDIR" >&2 ; exit 1 ; } ; dummy=$tmp/dummy ; tmpfiles="$dummy.c $dummy.o $dummy.rel $dummy" ; case $CC_FOR_BUILD,$HOST_CC,$CC in ,,) echo "int x;" > $dummy.c ; for c in cc gcc c89 c99 ; do if ($c -c -o $dummy.o $dummy.c) >/dev/null 2>&1 ; then CC_FOR_BUILD="$c"; break ; fi ; done ; if test x"$CC_FOR_BUILD" = x ; then CC_FOR_BUILD=no_compiler_found ; fi ;; ,,*) CC_FOR_BUILD=$CC ;; ,*,*) CC_FOR_BUILD=$HOST_CC ;; esac ; set_cc_for_build= ;' # This is needed to find uname on a Pyramid OSx when run in the BSD universe. # (ghazi@noc.rutgers.edu 1994-08-24) if (test -f /.attbin/uname) >/dev/null 2>&1 ; then PATH=$PATH:/.attbin ; export PATH fi UNAME_MACHINE=`(uname -m) 2>/dev/null` || UNAME_MACHINE=unknown UNAME_RELEASE=`(uname -r) 2>/dev/null` || UNAME_RELEASE=unknown UNAME_SYSTEM=`(uname -s) 2>/dev/null` || UNAME_SYSTEM=unknown UNAME_VERSION=`(uname -v) 2>/dev/null` || UNAME_VERSION=unknown case "${UNAME_SYSTEM}" in Linux|GNU|GNU/*) # If the system lacks a compiler, then just pick glibc. # We could probably try harder. LIBC=gnu eval $set_cc_for_build cat <<-EOF > $dummy.c #include #if defined(__UCLIBC__) LIBC=uclibc #elif defined(__dietlibc__) LIBC=dietlibc #else LIBC=gnu #endif EOF eval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep '^LIBC'` ;; esac # Note: order is significant - the case branches are not exclusive. case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in *:NetBSD:*:*) # NetBSD (nbsd) targets should (where applicable) match one or # more of the tuples: *-*-netbsdelf*, *-*-netbsdaout*, # *-*-netbsdecoff* and *-*-netbsd*. For targets that recently # switched to ELF, *-*-netbsd* would select the old # object file format. This provides both forward # compatibility and a consistent mechanism for selecting the # object file format. # # Note: NetBSD doesn't particularly care about the vendor # portion of the name. We always set it to "unknown". sysctl="sysctl -n hw.machine_arch" UNAME_MACHINE_ARCH=`(/sbin/$sysctl 2>/dev/null || \ /usr/sbin/$sysctl 2>/dev/null || echo unknown)` case "${UNAME_MACHINE_ARCH}" in armeb) machine=armeb-unknown ;; arm*) machine=arm-unknown ;; sh3el) machine=shl-unknown ;; sh3eb) machine=sh-unknown ;; sh5el) machine=sh5le-unknown ;; *) machine=${UNAME_MACHINE_ARCH}-unknown ;; esac # The Operating System including object format, if it has switched # to ELF recently, or will in the future. case "${UNAME_MACHINE_ARCH}" in arm*|i386|m68k|ns32k|sh3*|sparc|vax) eval $set_cc_for_build if echo __ELF__ | $CC_FOR_BUILD -E - 2>/dev/null \ | grep -q __ELF__ then # Once all utilities can be ECOFF (netbsdecoff) or a.out (netbsdaout). # Return netbsd for either. FIX? os=netbsd else os=netbsdelf fi ;; *) os=netbsd ;; esac # The OS release # Debian GNU/NetBSD machines have a different userland, and # thus, need a distinct triplet. However, they do not need # kernel version information, so it can be replaced with a # suitable tag, in the style of linux-gnu. case "${UNAME_VERSION}" in Debian*) release='-gnu' ;; *) release=`echo ${UNAME_RELEASE}|sed -e 's/[-_].*/\./'` ;; esac # Since CPU_TYPE-MANUFACTURER-KERNEL-OPERATING_SYSTEM: # contains redundant information, the shorter form: # CPU_TYPE-MANUFACTURER-OPERATING_SYSTEM is used. echo "${machine}-${os}${release}" exit ;; *:Bitrig:*:*) UNAME_MACHINE_ARCH=`arch | sed 's/Bitrig.//'` echo ${UNAME_MACHINE_ARCH}-unknown-bitrig${UNAME_RELEASE} exit ;; *:OpenBSD:*:*) UNAME_MACHINE_ARCH=`arch | sed 's/OpenBSD.//'` echo ${UNAME_MACHINE_ARCH}-unknown-openbsd${UNAME_RELEASE} exit ;; *:ekkoBSD:*:*) echo ${UNAME_MACHINE}-unknown-ekkobsd${UNAME_RELEASE} exit ;; *:SolidBSD:*:*) echo ${UNAME_MACHINE}-unknown-solidbsd${UNAME_RELEASE} exit ;; macppc:MirBSD:*:*) echo powerpc-unknown-mirbsd${UNAME_RELEASE} exit ;; *:MirBSD:*:*) echo ${UNAME_MACHINE}-unknown-mirbsd${UNAME_RELEASE} exit ;; alpha:OSF1:*:*) case $UNAME_RELEASE in *4.0) UNAME_RELEASE=`/usr/sbin/sizer -v | awk '{print $3}'` ;; *5.*) UNAME_RELEASE=`/usr/sbin/sizer -v | awk '{print $4}'` ;; esac # According to Compaq, /usr/sbin/psrinfo has been available on # OSF/1 and Tru64 systems produced since 1995. I hope that # covers most systems running today. This code pipes the CPU # types through head -n 1, so we only detect the type of CPU 0. ALPHA_CPU_TYPE=`/usr/sbin/psrinfo -v | sed -n -e 's/^ The alpha \(.*\) processor.*$/\1/p' | head -n 1` case "$ALPHA_CPU_TYPE" in "EV4 (21064)") UNAME_MACHINE="alpha" ;; "EV4.5 (21064)") UNAME_MACHINE="alpha" ;; "LCA4 (21066/21068)") UNAME_MACHINE="alpha" ;; "EV5 (21164)") UNAME_MACHINE="alphaev5" ;; "EV5.6 (21164A)") UNAME_MACHINE="alphaev56" ;; "EV5.6 (21164PC)") UNAME_MACHINE="alphapca56" ;; "EV5.7 (21164PC)") UNAME_MACHINE="alphapca57" ;; "EV6 (21264)") UNAME_MACHINE="alphaev6" ;; "EV6.7 (21264A)") UNAME_MACHINE="alphaev67" ;; "EV6.8CB (21264C)") UNAME_MACHINE="alphaev68" ;; "EV6.8AL (21264B)") UNAME_MACHINE="alphaev68" ;; "EV6.8CX (21264D)") UNAME_MACHINE="alphaev68" ;; "EV6.9A (21264/EV69A)") UNAME_MACHINE="alphaev69" ;; "EV7 (21364)") UNAME_MACHINE="alphaev7" ;; "EV7.9 (21364A)") UNAME_MACHINE="alphaev79" ;; esac # A Pn.n version is a patched version. # A Vn.n version is a released version. # A Tn.n version is a released field test version. # A Xn.n version is an unreleased experimental baselevel. # 1.2 uses "1.2" for uname -r. echo ${UNAME_MACHINE}-dec-osf`echo ${UNAME_RELEASE} | sed -e 's/^[PVTX]//' | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz'` # Reset EXIT trap before exiting to avoid spurious non-zero exit code. exitcode=$? trap '' 0 exit $exitcode ;; Alpha\ *:Windows_NT*:*) # How do we know it's Interix rather than the generic POSIX subsystem? # Should we change UNAME_MACHINE based on the output of uname instead # of the specific Alpha model? echo alpha-pc-interix exit ;; 21064:Windows_NT:50:3) echo alpha-dec-winnt3.5 exit ;; Amiga*:UNIX_System_V:4.0:*) echo m68k-unknown-sysv4 exit ;; *:[Aa]miga[Oo][Ss]:*:*) echo ${UNAME_MACHINE}-unknown-amigaos exit ;; *:[Mm]orph[Oo][Ss]:*:*) echo ${UNAME_MACHINE}-unknown-morphos exit ;; *:OS/390:*:*) echo i370-ibm-openedition exit ;; *:z/VM:*:*) echo s390-ibm-zvmoe exit ;; *:OS400:*:*) echo powerpc-ibm-os400 exit ;; arm:RISC*:1.[012]*:*|arm:riscix:1.[012]*:*) echo arm-acorn-riscix${UNAME_RELEASE} exit ;; arm*:riscos:*:*|arm*:RISCOS:*:*) echo arm-unknown-riscos exit ;; SR2?01:HI-UX/MPP:*:* | SR8000:HI-UX/MPP:*:*) echo hppa1.1-hitachi-hiuxmpp exit ;; Pyramid*:OSx*:*:* | MIS*:OSx*:*:* | MIS*:SMP_DC-OSx*:*:*) # akee@wpdis03.wpafb.af.mil (Earle F. Ake) contributed MIS and NILE. if test "`(/bin/universe) 2>/dev/null`" = att ; then echo pyramid-pyramid-sysv3 else echo pyramid-pyramid-bsd fi exit ;; NILE*:*:*:dcosx) echo pyramid-pyramid-svr4 exit ;; DRS?6000:unix:4.0:6*) echo sparc-icl-nx6 exit ;; DRS?6000:UNIX_SV:4.2*:7* | DRS?6000:isis:4.2*:7*) case `/usr/bin/uname -p` in sparc) echo sparc-icl-nx7; exit ;; esac ;; s390x:SunOS:*:*) echo ${UNAME_MACHINE}-ibm-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'` exit ;; sun4H:SunOS:5.*:*) echo sparc-hal-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'` exit ;; sun4*:SunOS:5.*:* | tadpole*:SunOS:5.*:*) echo sparc-sun-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'` exit ;; i86pc:AuroraUX:5.*:* | i86xen:AuroraUX:5.*:*) echo i386-pc-auroraux${UNAME_RELEASE} exit ;; i86pc:SunOS:5.*:* | i86xen:SunOS:5.*:*) eval $set_cc_for_build SUN_ARCH="i386" # If there is a compiler, see if it is configured for 64-bit objects. # Note that the Sun cc does not turn __LP64__ into 1 like gcc does. # This test works for both compilers. if [ "$CC_FOR_BUILD" != 'no_compiler_found' ]; then if (echo '#ifdef __amd64'; echo IS_64BIT_ARCH; echo '#endif') | \ (CCOPTS= $CC_FOR_BUILD -E - 2>/dev/null) | \ grep IS_64BIT_ARCH >/dev/null then SUN_ARCH="x86_64" fi fi echo ${SUN_ARCH}-pc-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'` exit ;; sun4*:SunOS:6*:*) # According to config.sub, this is the proper way to canonicalize # SunOS6. Hard to guess exactly what SunOS6 will be like, but # it's likely to be more like Solaris than SunOS4. echo sparc-sun-solaris3`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'` exit ;; sun4*:SunOS:*:*) case "`/usr/bin/arch -k`" in Series*|S4*) UNAME_RELEASE=`uname -v` ;; esac # Japanese Language versions have a version number like `4.1.3-JL'. echo sparc-sun-sunos`echo ${UNAME_RELEASE}|sed -e 's/-/_/'` exit ;; sun3*:SunOS:*:*) echo m68k-sun-sunos${UNAME_RELEASE} exit ;; sun*:*:4.2BSD:*) UNAME_RELEASE=`(sed 1q /etc/motd | awk '{print substr($5,1,3)}') 2>/dev/null` test "x${UNAME_RELEASE}" = "x" && UNAME_RELEASE=3 case "`/bin/arch`" in sun3) echo m68k-sun-sunos${UNAME_RELEASE} ;; sun4) echo sparc-sun-sunos${UNAME_RELEASE} ;; esac exit ;; aushp:SunOS:*:*) echo sparc-auspex-sunos${UNAME_RELEASE} exit ;; # The situation for MiNT is a little confusing. The machine name # can be virtually everything (everything which is not # "atarist" or "atariste" at least should have a processor # > m68000). The system name ranges from "MiNT" over "FreeMiNT" # to the lowercase version "mint" (or "freemint"). Finally # the system name "TOS" denotes a system which is actually not # MiNT. But MiNT is downward compatible to TOS, so this should # be no problem. atarist[e]:*MiNT:*:* | atarist[e]:*mint:*:* | atarist[e]:*TOS:*:*) echo m68k-atari-mint${UNAME_RELEASE} exit ;; atari*:*MiNT:*:* | atari*:*mint:*:* | atarist[e]:*TOS:*:*) echo m68k-atari-mint${UNAME_RELEASE} exit ;; *falcon*:*MiNT:*:* | *falcon*:*mint:*:* | *falcon*:*TOS:*:*) echo m68k-atari-mint${UNAME_RELEASE} exit ;; milan*:*MiNT:*:* | milan*:*mint:*:* | *milan*:*TOS:*:*) echo m68k-milan-mint${UNAME_RELEASE} exit ;; hades*:*MiNT:*:* | hades*:*mint:*:* | *hades*:*TOS:*:*) echo m68k-hades-mint${UNAME_RELEASE} exit ;; *:*MiNT:*:* | *:*mint:*:* | *:*TOS:*:*) echo m68k-unknown-mint${UNAME_RELEASE} exit ;; m68k:machten:*:*) echo m68k-apple-machten${UNAME_RELEASE} exit ;; powerpc:machten:*:*) echo powerpc-apple-machten${UNAME_RELEASE} exit ;; RISC*:Mach:*:*) echo mips-dec-mach_bsd4.3 exit ;; RISC*:ULTRIX:*:*) echo mips-dec-ultrix${UNAME_RELEASE} exit ;; VAX*:ULTRIX*:*:*) echo vax-dec-ultrix${UNAME_RELEASE} exit ;; 2020:CLIX:*:* | 2430:CLIX:*:*) echo clipper-intergraph-clix${UNAME_RELEASE} exit ;; mips:*:*:UMIPS | mips:*:*:RISCos) eval $set_cc_for_build sed 's/^ //' << EOF >$dummy.c #ifdef __cplusplus #include /* for printf() prototype */ int main (int argc, char *argv[]) { #else int main (argc, argv) int argc; char *argv[]; { #endif #if defined (host_mips) && defined (MIPSEB) #if defined (SYSTYPE_SYSV) printf ("mips-mips-riscos%ssysv\n", argv[1]); exit (0); #endif #if defined (SYSTYPE_SVR4) printf ("mips-mips-riscos%ssvr4\n", argv[1]); exit (0); #endif #if defined (SYSTYPE_BSD43) || defined(SYSTYPE_BSD) printf ("mips-mips-riscos%sbsd\n", argv[1]); exit (0); #endif #endif exit (-1); } EOF $CC_FOR_BUILD -o $dummy $dummy.c && dummyarg=`echo "${UNAME_RELEASE}" | sed -n 's/\([0-9]*\).*/\1/p'` && SYSTEM_NAME=`$dummy $dummyarg` && { echo "$SYSTEM_NAME"; exit; } echo mips-mips-riscos${UNAME_RELEASE} exit ;; Motorola:PowerMAX_OS:*:*) echo powerpc-motorola-powermax exit ;; Motorola:*:4.3:PL8-*) echo powerpc-harris-powermax exit ;; Night_Hawk:*:*:PowerMAX_OS | Synergy:PowerMAX_OS:*:*) echo powerpc-harris-powermax exit ;; Night_Hawk:Power_UNIX:*:*) echo powerpc-harris-powerunix exit ;; m88k:CX/UX:7*:*) echo m88k-harris-cxux7 exit ;; m88k:*:4*:R4*) echo m88k-motorola-sysv4 exit ;; m88k:*:3*:R3*) echo m88k-motorola-sysv3 exit ;; AViiON:dgux:*:*) # DG/UX returns AViiON for all architectures UNAME_PROCESSOR=`/usr/bin/uname -p` if [ $UNAME_PROCESSOR = mc88100 ] || [ $UNAME_PROCESSOR = mc88110 ] then if [ ${TARGET_BINARY_INTERFACE}x = m88kdguxelfx ] || \ [ ${TARGET_BINARY_INTERFACE}x = x ] then echo m88k-dg-dgux${UNAME_RELEASE} else echo m88k-dg-dguxbcs${UNAME_RELEASE} fi else echo i586-dg-dgux${UNAME_RELEASE} fi exit ;; M88*:DolphinOS:*:*) # DolphinOS (SVR3) echo m88k-dolphin-sysv3 exit ;; M88*:*:R3*:*) # Delta 88k system running SVR3 echo m88k-motorola-sysv3 exit ;; XD88*:*:*:*) # Tektronix XD88 system running UTekV (SVR3) echo m88k-tektronix-sysv3 exit ;; Tek43[0-9][0-9]:UTek:*:*) # Tektronix 4300 system running UTek (BSD) echo m68k-tektronix-bsd exit ;; *:IRIX*:*:*) echo mips-sgi-irix`echo ${UNAME_RELEASE}|sed -e 's/-/_/g'` exit ;; ????????:AIX?:[12].1:2) # AIX 2.2.1 or AIX 2.1.1 is RT/PC AIX. echo romp-ibm-aix # uname -m gives an 8 hex-code CPU id exit ;; # Note that: echo "'`uname -s`'" gives 'AIX ' i*86:AIX:*:*) echo i386-ibm-aix exit ;; ia64:AIX:*:*) if [ -x /usr/bin/oslevel ] ; then IBM_REV=`/usr/bin/oslevel` else IBM_REV=${UNAME_VERSION}.${UNAME_RELEASE} fi echo ${UNAME_MACHINE}-ibm-aix${IBM_REV} exit ;; *:AIX:2:3) if grep bos325 /usr/include/stdio.h >/dev/null 2>&1; then eval $set_cc_for_build sed 's/^ //' << EOF >$dummy.c #include main() { if (!__power_pc()) exit(1); puts("powerpc-ibm-aix3.2.5"); exit(0); } EOF if $CC_FOR_BUILD -o $dummy $dummy.c && SYSTEM_NAME=`$dummy` then echo "$SYSTEM_NAME" else echo rs6000-ibm-aix3.2.5 fi elif grep bos324 /usr/include/stdio.h >/dev/null 2>&1; then echo rs6000-ibm-aix3.2.4 else echo rs6000-ibm-aix3.2 fi exit ;; *:AIX:*:[4567]) IBM_CPU_ID=`/usr/sbin/lsdev -C -c processor -S available | sed 1q | awk '{ print $1 }'` if /usr/sbin/lsattr -El ${IBM_CPU_ID} | grep ' POWER' >/dev/null 2>&1; then IBM_ARCH=rs6000 else IBM_ARCH=powerpc fi if [ -x /usr/bin/oslevel ] ; then IBM_REV=`/usr/bin/oslevel` else IBM_REV=${UNAME_VERSION}.${UNAME_RELEASE} fi echo ${IBM_ARCH}-ibm-aix${IBM_REV} exit ;; *:AIX:*:*) echo rs6000-ibm-aix exit ;; ibmrt:4.4BSD:*|romp-ibm:BSD:*) echo romp-ibm-bsd4.4 exit ;; ibmrt:*BSD:*|romp-ibm:BSD:*) # covers RT/PC BSD and echo romp-ibm-bsd${UNAME_RELEASE} # 4.3 with uname added to exit ;; # report: romp-ibm BSD 4.3 *:BOSX:*:*) echo rs6000-bull-bosx exit ;; DPX/2?00:B.O.S.:*:*) echo m68k-bull-sysv3 exit ;; 9000/[34]??:4.3bsd:1.*:*) echo m68k-hp-bsd exit ;; hp300:4.4BSD:*:* | 9000/[34]??:4.3bsd:2.*:*) echo m68k-hp-bsd4.4 exit ;; 9000/[34678]??:HP-UX:*:*) HPUX_REV=`echo ${UNAME_RELEASE}|sed -e 's/[^.]*.[0B]*//'` case "${UNAME_MACHINE}" in 9000/31? ) HP_ARCH=m68000 ;; 9000/[34]?? ) HP_ARCH=m68k ;; 9000/[678][0-9][0-9]) if [ -x /usr/bin/getconf ]; then sc_cpu_version=`/usr/bin/getconf SC_CPU_VERSION 2>/dev/null` sc_kernel_bits=`/usr/bin/getconf SC_KERNEL_BITS 2>/dev/null` case "${sc_cpu_version}" in 523) HP_ARCH="hppa1.0" ;; # CPU_PA_RISC1_0 528) HP_ARCH="hppa1.1" ;; # CPU_PA_RISC1_1 532) # CPU_PA_RISC2_0 case "${sc_kernel_bits}" in 32) HP_ARCH="hppa2.0n" ;; 64) HP_ARCH="hppa2.0w" ;; '') HP_ARCH="hppa2.0" ;; # HP-UX 10.20 esac ;; esac fi if [ "${HP_ARCH}" = "" ]; then eval $set_cc_for_build sed 's/^ //' << EOF >$dummy.c #define _HPUX_SOURCE #include #include int main () { #if defined(_SC_KERNEL_BITS) long bits = sysconf(_SC_KERNEL_BITS); #endif long cpu = sysconf (_SC_CPU_VERSION); switch (cpu) { case CPU_PA_RISC1_0: puts ("hppa1.0"); break; case CPU_PA_RISC1_1: puts ("hppa1.1"); break; case CPU_PA_RISC2_0: #if defined(_SC_KERNEL_BITS) switch (bits) { case 64: puts ("hppa2.0w"); break; case 32: puts ("hppa2.0n"); break; default: puts ("hppa2.0"); break; } break; #else /* !defined(_SC_KERNEL_BITS) */ puts ("hppa2.0"); break; #endif default: puts ("hppa1.0"); break; } exit (0); } EOF (CCOPTS= $CC_FOR_BUILD -o $dummy $dummy.c 2>/dev/null) && HP_ARCH=`$dummy` test -z "$HP_ARCH" && HP_ARCH=hppa fi ;; esac if [ ${HP_ARCH} = "hppa2.0w" ] then eval $set_cc_for_build # hppa2.0w-hp-hpux* has a 64-bit kernel and a compiler generating # 32-bit code. hppa64-hp-hpux* has the same kernel and a compiler # generating 64-bit code. GNU and HP use different nomenclature: # # $ CC_FOR_BUILD=cc ./config.guess # => hppa2.0w-hp-hpux11.23 # $ CC_FOR_BUILD="cc +DA2.0w" ./config.guess # => hppa64-hp-hpux11.23 if echo __LP64__ | (CCOPTS= $CC_FOR_BUILD -E - 2>/dev/null) | grep -q __LP64__ then HP_ARCH="hppa2.0w" else HP_ARCH="hppa64" fi fi echo ${HP_ARCH}-hp-hpux${HPUX_REV} exit ;; ia64:HP-UX:*:*) HPUX_REV=`echo ${UNAME_RELEASE}|sed -e 's/[^.]*.[0B]*//'` echo ia64-hp-hpux${HPUX_REV} exit ;; 3050*:HI-UX:*:*) eval $set_cc_for_build sed 's/^ //' << EOF >$dummy.c #include int main () { long cpu = sysconf (_SC_CPU_VERSION); /* The order matters, because CPU_IS_HP_MC68K erroneously returns true for CPU_PA_RISC1_0. CPU_IS_PA_RISC returns correct results, however. */ if (CPU_IS_PA_RISC (cpu)) { switch (cpu) { case CPU_PA_RISC1_0: puts ("hppa1.0-hitachi-hiuxwe2"); break; case CPU_PA_RISC1_1: puts ("hppa1.1-hitachi-hiuxwe2"); break; case CPU_PA_RISC2_0: puts ("hppa2.0-hitachi-hiuxwe2"); break; default: puts ("hppa-hitachi-hiuxwe2"); break; } } else if (CPU_IS_HP_MC68K (cpu)) puts ("m68k-hitachi-hiuxwe2"); else puts ("unknown-hitachi-hiuxwe2"); exit (0); } EOF $CC_FOR_BUILD -o $dummy $dummy.c && SYSTEM_NAME=`$dummy` && { echo "$SYSTEM_NAME"; exit; } echo unknown-hitachi-hiuxwe2 exit ;; 9000/7??:4.3bsd:*:* | 9000/8?[79]:4.3bsd:*:* ) echo hppa1.1-hp-bsd exit ;; 9000/8??:4.3bsd:*:*) echo hppa1.0-hp-bsd exit ;; *9??*:MPE/iX:*:* | *3000*:MPE/iX:*:*) echo hppa1.0-hp-mpeix exit ;; hp7??:OSF1:*:* | hp8?[79]:OSF1:*:* ) echo hppa1.1-hp-osf exit ;; hp8??:OSF1:*:*) echo hppa1.0-hp-osf exit ;; i*86:OSF1:*:*) if [ -x /usr/sbin/sysversion ] ; then echo ${UNAME_MACHINE}-unknown-osf1mk else echo ${UNAME_MACHINE}-unknown-osf1 fi exit ;; parisc*:Lites*:*:*) echo hppa1.1-hp-lites exit ;; C1*:ConvexOS:*:* | convex:ConvexOS:C1*:*) echo c1-convex-bsd exit ;; C2*:ConvexOS:*:* | convex:ConvexOS:C2*:*) if getsysinfo -f scalar_acc then echo c32-convex-bsd else echo c2-convex-bsd fi exit ;; C34*:ConvexOS:*:* | convex:ConvexOS:C34*:*) echo c34-convex-bsd exit ;; C38*:ConvexOS:*:* | convex:ConvexOS:C38*:*) echo c38-convex-bsd exit ;; C4*:ConvexOS:*:* | convex:ConvexOS:C4*:*) echo c4-convex-bsd exit ;; CRAY*Y-MP:*:*:*) echo ymp-cray-unicos${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/' exit ;; CRAY*[A-Z]90:*:*:*) echo ${UNAME_MACHINE}-cray-unicos${UNAME_RELEASE} \ | sed -e 's/CRAY.*\([A-Z]90\)/\1/' \ -e y/ABCDEFGHIJKLMNOPQRSTUVWXYZ/abcdefghijklmnopqrstuvwxyz/ \ -e 's/\.[^.]*$/.X/' exit ;; CRAY*TS:*:*:*) echo t90-cray-unicos${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/' exit ;; CRAY*T3E:*:*:*) echo alphaev5-cray-unicosmk${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/' exit ;; CRAY*SV1:*:*:*) echo sv1-cray-unicos${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/' exit ;; *:UNICOS/mp:*:*) echo craynv-cray-unicosmp${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/' exit ;; F30[01]:UNIX_System_V:*:* | F700:UNIX_System_V:*:*) FUJITSU_PROC=`uname -m | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz'` FUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\///'` FUJITSU_REL=`echo ${UNAME_RELEASE} | sed -e 's/ /_/'` echo "${FUJITSU_PROC}-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}" exit ;; 5000:UNIX_System_V:4.*:*) FUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\///'` FUJITSU_REL=`echo ${UNAME_RELEASE} | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/ /_/'` echo "sparc-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}" exit ;; i*86:BSD/386:*:* | i*86:BSD/OS:*:* | *:Ascend\ Embedded/OS:*:*) echo ${UNAME_MACHINE}-pc-bsdi${UNAME_RELEASE} exit ;; sparc*:BSD/OS:*:*) echo sparc-unknown-bsdi${UNAME_RELEASE} exit ;; *:BSD/OS:*:*) echo ${UNAME_MACHINE}-unknown-bsdi${UNAME_RELEASE} exit ;; *:FreeBSD:*:*) UNAME_PROCESSOR=`/usr/bin/uname -p` case ${UNAME_PROCESSOR} in amd64) echo x86_64-unknown-freebsd`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'` ;; *) echo ${UNAME_PROCESSOR}-unknown-freebsd`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'` ;; esac exit ;; i*:CYGWIN*:*) echo ${UNAME_MACHINE}-pc-cygwin exit ;; *:MINGW64*:*) echo ${UNAME_MACHINE}-pc-mingw64 exit ;; *:MINGW*:*) echo ${UNAME_MACHINE}-pc-mingw32 exit ;; i*:MSYS*:*) echo ${UNAME_MACHINE}-pc-msys exit ;; i*:windows32*:*) # uname -m includes "-pc" on this system. echo ${UNAME_MACHINE}-mingw32 exit ;; i*:PW*:*) echo ${UNAME_MACHINE}-pc-pw32 exit ;; *:Interix*:*) case ${UNAME_MACHINE} in x86) echo i586-pc-interix${UNAME_RELEASE} exit ;; authenticamd | genuineintel | EM64T) echo x86_64-unknown-interix${UNAME_RELEASE} exit ;; IA64) echo ia64-unknown-interix${UNAME_RELEASE} exit ;; esac ;; [345]86:Windows_95:* | [345]86:Windows_98:* | [345]86:Windows_NT:*) echo i${UNAME_MACHINE}-pc-mks exit ;; 8664:Windows_NT:*) echo x86_64-pc-mks exit ;; i*:Windows_NT*:* | Pentium*:Windows_NT*:*) # How do we know it's Interix rather than the generic POSIX subsystem? # It also conflicts with pre-2.0 versions of AT&T UWIN. Should we # UNAME_MACHINE based on the output of uname instead of i386? echo i586-pc-interix exit ;; i*:UWIN*:*) echo ${UNAME_MACHINE}-pc-uwin exit ;; amd64:CYGWIN*:*:* | x86_64:CYGWIN*:*:*) echo x86_64-unknown-cygwin exit ;; p*:CYGWIN*:*) echo powerpcle-unknown-cygwin exit ;; prep*:SunOS:5.*:*) echo powerpcle-unknown-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'` exit ;; *:GNU:*:*) # the GNU system echo `echo ${UNAME_MACHINE}|sed -e 's,[-/].*$,,'`-unknown-${LIBC}`echo ${UNAME_RELEASE}|sed -e 's,/.*$,,'` exit ;; *:GNU/*:*:*) # other systems with GNU libc and userland echo ${UNAME_MACHINE}-unknown-`echo ${UNAME_SYSTEM} | sed 's,^[^/]*/,,' | tr '[A-Z]' '[a-z]'``echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'`-${LIBC} exit ;; i*86:Minix:*:*) echo ${UNAME_MACHINE}-pc-minix exit ;; aarch64:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; aarch64_be:Linux:*:*) UNAME_MACHINE=aarch64_be echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; alpha:Linux:*:*) case `sed -n '/^cpu model/s/^.*: \(.*\)/\1/p' < /proc/cpuinfo` in EV5) UNAME_MACHINE=alphaev5 ;; EV56) UNAME_MACHINE=alphaev56 ;; PCA56) UNAME_MACHINE=alphapca56 ;; PCA57) UNAME_MACHINE=alphapca56 ;; EV6) UNAME_MACHINE=alphaev6 ;; EV67) UNAME_MACHINE=alphaev67 ;; EV68*) UNAME_MACHINE=alphaev68 ;; esac objdump --private-headers /bin/sh | grep -q ld.so.1 if test "$?" = 0 ; then LIBC="gnulibc1" ; fi echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; arc:Linux:*:* | arceb:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; arm*:Linux:*:*) eval $set_cc_for_build if echo __ARM_EABI__ | $CC_FOR_BUILD -E - 2>/dev/null \ | grep -q __ARM_EABI__ then echo ${UNAME_MACHINE}-unknown-linux-${LIBC} else if echo __ARM_PCS_VFP | $CC_FOR_BUILD -E - 2>/dev/null \ | grep -q __ARM_PCS_VFP then echo ${UNAME_MACHINE}-unknown-linux-${LIBC}eabi else echo ${UNAME_MACHINE}-unknown-linux-${LIBC}eabihf fi fi exit ;; avr32*:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; cris:Linux:*:*) echo ${UNAME_MACHINE}-axis-linux-${LIBC} exit ;; crisv32:Linux:*:*) echo ${UNAME_MACHINE}-axis-linux-${LIBC} exit ;; frv:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; hexagon:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; i*86:Linux:*:*) echo ${UNAME_MACHINE}-pc-linux-${LIBC} exit ;; ia64:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; m32r*:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; m68*:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; mips:Linux:*:* | mips64:Linux:*:*) eval $set_cc_for_build sed 's/^ //' << EOF >$dummy.c #undef CPU #undef ${UNAME_MACHINE} #undef ${UNAME_MACHINE}el #if defined(__MIPSEL__) || defined(__MIPSEL) || defined(_MIPSEL) || defined(MIPSEL) CPU=${UNAME_MACHINE}el #else #if defined(__MIPSEB__) || defined(__MIPSEB) || defined(_MIPSEB) || defined(MIPSEB) CPU=${UNAME_MACHINE} #else CPU= #endif #endif EOF eval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep '^CPU'` test x"${CPU}" != x && { echo "${CPU}-unknown-linux-${LIBC}"; exit; } ;; or1k:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; or32:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; padre:Linux:*:*) echo sparc-unknown-linux-${LIBC} exit ;; parisc64:Linux:*:* | hppa64:Linux:*:*) echo hppa64-unknown-linux-${LIBC} exit ;; parisc:Linux:*:* | hppa:Linux:*:*) # Look for CPU level case `grep '^cpu[^a-z]*:' /proc/cpuinfo 2>/dev/null | cut -d' ' -f2` in PA7*) echo hppa1.1-unknown-linux-${LIBC} ;; PA8*) echo hppa2.0-unknown-linux-${LIBC} ;; *) echo hppa-unknown-linux-${LIBC} ;; esac exit ;; ppc64:Linux:*:*) echo powerpc64-unknown-linux-${LIBC} exit ;; ppc:Linux:*:*) echo powerpc-unknown-linux-${LIBC} exit ;; ppc64le:Linux:*:*) echo powerpc64le-unknown-linux-${LIBC} exit ;; ppcle:Linux:*:*) echo powerpcle-unknown-linux-${LIBC} exit ;; s390:Linux:*:* | s390x:Linux:*:*) echo ${UNAME_MACHINE}-ibm-linux-${LIBC} exit ;; sh64*:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; sh*:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; sparc:Linux:*:* | sparc64:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; tile*:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; vax:Linux:*:*) echo ${UNAME_MACHINE}-dec-linux-${LIBC} exit ;; x86_64:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; xtensa*:Linux:*:*) echo ${UNAME_MACHINE}-unknown-linux-${LIBC} exit ;; i*86:DYNIX/ptx:4*:*) # ptx 4.0 does uname -s correctly, with DYNIX/ptx in there. # earlier versions are messed up and put the nodename in both # sysname and nodename. echo i386-sequent-sysv4 exit ;; i*86:UNIX_SV:4.2MP:2.*) # Unixware is an offshoot of SVR4, but it has its own version # number series starting with 2... # I am not positive that other SVR4 systems won't match this, # I just have to hope. -- rms. # Use sysv4.2uw... so that sysv4* matches it. echo ${UNAME_MACHINE}-pc-sysv4.2uw${UNAME_VERSION} exit ;; i*86:OS/2:*:*) # If we were able to find `uname', then EMX Unix compatibility # is probably installed. echo ${UNAME_MACHINE}-pc-os2-emx exit ;; i*86:XTS-300:*:STOP) echo ${UNAME_MACHINE}-unknown-stop exit ;; i*86:atheos:*:*) echo ${UNAME_MACHINE}-unknown-atheos exit ;; i*86:syllable:*:*) echo ${UNAME_MACHINE}-pc-syllable exit ;; i*86:LynxOS:2.*:* | i*86:LynxOS:3.[01]*:* | i*86:LynxOS:4.[02]*:*) echo i386-unknown-lynxos${UNAME_RELEASE} exit ;; i*86:*DOS:*:*) echo ${UNAME_MACHINE}-pc-msdosdjgpp exit ;; i*86:*:4.*:* | i*86:SYSTEM_V:4.*:*) UNAME_REL=`echo ${UNAME_RELEASE} | sed 's/\/MP$//'` if grep Novell /usr/include/link.h >/dev/null 2>/dev/null; then echo ${UNAME_MACHINE}-univel-sysv${UNAME_REL} else echo ${UNAME_MACHINE}-pc-sysv${UNAME_REL} fi exit ;; i*86:*:5:[678]*) # UnixWare 7.x, OpenUNIX and OpenServer 6. case `/bin/uname -X | grep "^Machine"` in *486*) UNAME_MACHINE=i486 ;; *Pentium) UNAME_MACHINE=i586 ;; *Pent*|*Celeron) UNAME_MACHINE=i686 ;; esac echo ${UNAME_MACHINE}-unknown-sysv${UNAME_RELEASE}${UNAME_SYSTEM}${UNAME_VERSION} exit ;; i*86:*:3.2:*) if test -f /usr/options/cb.name; then UNAME_REL=`sed -n 's/.*Version //p' /dev/null >/dev/null ; then UNAME_REL=`(/bin/uname -X|grep Release|sed -e 's/.*= //')` (/bin/uname -X|grep i80486 >/dev/null) && UNAME_MACHINE=i486 (/bin/uname -X|grep '^Machine.*Pentium' >/dev/null) \ && UNAME_MACHINE=i586 (/bin/uname -X|grep '^Machine.*Pent *II' >/dev/null) \ && UNAME_MACHINE=i686 (/bin/uname -X|grep '^Machine.*Pentium Pro' >/dev/null) \ && UNAME_MACHINE=i686 echo ${UNAME_MACHINE}-pc-sco$UNAME_REL else echo ${UNAME_MACHINE}-pc-sysv32 fi exit ;; pc:*:*:*) # Left here for compatibility: # uname -m prints for DJGPP always 'pc', but it prints nothing about # the processor, so we play safe by assuming i586. # Note: whatever this is, it MUST be the same as what config.sub # prints for the "djgpp" host, or else GDB configury will decide that # this is a cross-build. echo i586-pc-msdosdjgpp exit ;; Intel:Mach:3*:*) echo i386-pc-mach3 exit ;; paragon:*:*:*) echo i860-intel-osf1 exit ;; i860:*:4.*:*) # i860-SVR4 if grep Stardent /usr/include/sys/uadmin.h >/dev/null 2>&1 ; then echo i860-stardent-sysv${UNAME_RELEASE} # Stardent Vistra i860-SVR4 else # Add other i860-SVR4 vendors below as they are discovered. echo i860-unknown-sysv${UNAME_RELEASE} # Unknown i860-SVR4 fi exit ;; mini*:CTIX:SYS*5:*) # "miniframe" echo m68010-convergent-sysv exit ;; mc68k:UNIX:SYSTEM5:3.51m) echo m68k-convergent-sysv exit ;; M680?0:D-NIX:5.3:*) echo m68k-diab-dnix exit ;; M68*:*:R3V[5678]*:*) test -r /sysV68 && { echo 'm68k-motorola-sysv'; exit; } ;; 3[345]??:*:4.0:3.0 | 3[34]??A:*:4.0:3.0 | 3[34]??,*:*:4.0:3.0 | 3[34]??/*:*:4.0:3.0 | 4400:*:4.0:3.0 | 4850:*:4.0:3.0 | SKA40:*:4.0:3.0 | SDS2:*:4.0:3.0 | SHG2:*:4.0:3.0 | S7501*:*:4.0:3.0) OS_REL='' test -r /etc/.relid \ && OS_REL=.`sed -n 's/[^ ]* [^ ]* \([0-9][0-9]\).*/\1/p' < /etc/.relid` /bin/uname -p 2>/dev/null | grep 86 >/dev/null \ && { echo i486-ncr-sysv4.3${OS_REL}; exit; } /bin/uname -p 2>/dev/null | /bin/grep entium >/dev/null \ && { echo i586-ncr-sysv4.3${OS_REL}; exit; } ;; 3[34]??:*:4.0:* | 3[34]??,*:*:4.0:*) /bin/uname -p 2>/dev/null | grep 86 >/dev/null \ && { echo i486-ncr-sysv4; exit; } ;; NCR*:*:4.2:* | MPRAS*:*:4.2:*) OS_REL='.3' test -r /etc/.relid \ && OS_REL=.`sed -n 's/[^ ]* [^ ]* \([0-9][0-9]\).*/\1/p' < /etc/.relid` /bin/uname -p 2>/dev/null | grep 86 >/dev/null \ && { echo i486-ncr-sysv4.3${OS_REL}; exit; } /bin/uname -p 2>/dev/null | /bin/grep entium >/dev/null \ && { echo i586-ncr-sysv4.3${OS_REL}; exit; } /bin/uname -p 2>/dev/null | /bin/grep pteron >/dev/null \ && { echo i586-ncr-sysv4.3${OS_REL}; exit; } ;; m68*:LynxOS:2.*:* | m68*:LynxOS:3.0*:*) echo m68k-unknown-lynxos${UNAME_RELEASE} exit ;; mc68030:UNIX_System_V:4.*:*) echo m68k-atari-sysv4 exit ;; TSUNAMI:LynxOS:2.*:*) echo sparc-unknown-lynxos${UNAME_RELEASE} exit ;; rs6000:LynxOS:2.*:*) echo rs6000-unknown-lynxos${UNAME_RELEASE} exit ;; PowerPC:LynxOS:2.*:* | PowerPC:LynxOS:3.[01]*:* | PowerPC:LynxOS:4.[02]*:*) echo powerpc-unknown-lynxos${UNAME_RELEASE} exit ;; SM[BE]S:UNIX_SV:*:*) echo mips-dde-sysv${UNAME_RELEASE} exit ;; RM*:ReliantUNIX-*:*:*) echo mips-sni-sysv4 exit ;; RM*:SINIX-*:*:*) echo mips-sni-sysv4 exit ;; *:SINIX-*:*:*) if uname -p 2>/dev/null >/dev/null ; then UNAME_MACHINE=`(uname -p) 2>/dev/null` echo ${UNAME_MACHINE}-sni-sysv4 else echo ns32k-sni-sysv fi exit ;; PENTIUM:*:4.0*:*) # Unisys `ClearPath HMP IX 4000' SVR4/MP effort # says echo i586-unisys-sysv4 exit ;; *:UNIX_System_V:4*:FTX*) # From Gerald Hewes . # How about differentiating between stratus architectures? -djm echo hppa1.1-stratus-sysv4 exit ;; *:*:*:FTX*) # From seanf@swdc.stratus.com. echo i860-stratus-sysv4 exit ;; i*86:VOS:*:*) # From Paul.Green@stratus.com. echo ${UNAME_MACHINE}-stratus-vos exit ;; *:VOS:*:*) # From Paul.Green@stratus.com. echo hppa1.1-stratus-vos exit ;; mc68*:A/UX:*:*) echo m68k-apple-aux${UNAME_RELEASE} exit ;; news*:NEWS-OS:6*:*) echo mips-sony-newsos6 exit ;; R[34]000:*System_V*:*:* | R4000:UNIX_SYSV:*:* | R*000:UNIX_SV:*:*) if [ -d /usr/nec ]; then echo mips-nec-sysv${UNAME_RELEASE} else echo mips-unknown-sysv${UNAME_RELEASE} fi exit ;; BeBox:BeOS:*:*) # BeOS running on hardware made by Be, PPC only. echo powerpc-be-beos exit ;; BeMac:BeOS:*:*) # BeOS running on Mac or Mac clone, PPC only. echo powerpc-apple-beos exit ;; BePC:BeOS:*:*) # BeOS running on Intel PC compatible. echo i586-pc-beos exit ;; BePC:Haiku:*:*) # Haiku running on Intel PC compatible. echo i586-pc-haiku exit ;; x86_64:Haiku:*:*) echo x86_64-unknown-haiku exit ;; SX-4:SUPER-UX:*:*) echo sx4-nec-superux${UNAME_RELEASE} exit ;; SX-5:SUPER-UX:*:*) echo sx5-nec-superux${UNAME_RELEASE} exit ;; SX-6:SUPER-UX:*:*) echo sx6-nec-superux${UNAME_RELEASE} exit ;; SX-7:SUPER-UX:*:*) echo sx7-nec-superux${UNAME_RELEASE} exit ;; SX-8:SUPER-UX:*:*) echo sx8-nec-superux${UNAME_RELEASE} exit ;; SX-8R:SUPER-UX:*:*) echo sx8r-nec-superux${UNAME_RELEASE} exit ;; Power*:Rhapsody:*:*) echo powerpc-apple-rhapsody${UNAME_RELEASE} exit ;; *:Rhapsody:*:*) echo ${UNAME_MACHINE}-apple-rhapsody${UNAME_RELEASE} exit ;; *:Darwin:*:*) UNAME_PROCESSOR=`uname -p` || UNAME_PROCESSOR=unknown eval $set_cc_for_build if test "$UNAME_PROCESSOR" = unknown ; then UNAME_PROCESSOR=powerpc fi if test `echo "$UNAME_RELEASE" | sed -e 's/\..*//'` -le 10 ; then if [ "$CC_FOR_BUILD" != 'no_compiler_found' ]; then if (echo '#ifdef __LP64__'; echo IS_64BIT_ARCH; echo '#endif') | \ (CCOPTS= $CC_FOR_BUILD -E - 2>/dev/null) | \ grep IS_64BIT_ARCH >/dev/null then case $UNAME_PROCESSOR in i386) UNAME_PROCESSOR=x86_64 ;; powerpc) UNAME_PROCESSOR=powerpc64 ;; esac fi fi elif test "$UNAME_PROCESSOR" = i386 ; then # Avoid executing cc on OS X 10.9, as it ships with a stub # that puts up a graphical alert prompting to install # developer tools. Any system running Mac OS X 10.7 or # later (Darwin 11 and later) is required to have a 64-bit # processor. This is not true of the ARM version of Darwin # that Apple uses in portable devices. UNAME_PROCESSOR=x86_64 fi echo ${UNAME_PROCESSOR}-apple-darwin${UNAME_RELEASE} exit ;; *:procnto*:*:* | *:QNX:[0123456789]*:*) UNAME_PROCESSOR=`uname -p` if test "$UNAME_PROCESSOR" = "x86"; then UNAME_PROCESSOR=i386 UNAME_MACHINE=pc fi echo ${UNAME_PROCESSOR}-${UNAME_MACHINE}-nto-qnx${UNAME_RELEASE} exit ;; *:QNX:*:4*) echo i386-pc-qnx exit ;; NEO-?:NONSTOP_KERNEL:*:*) echo neo-tandem-nsk${UNAME_RELEASE} exit ;; NSE-*:NONSTOP_KERNEL:*:*) echo nse-tandem-nsk${UNAME_RELEASE} exit ;; NSR-?:NONSTOP_KERNEL:*:*) echo nsr-tandem-nsk${UNAME_RELEASE} exit ;; *:NonStop-UX:*:*) echo mips-compaq-nonstopux exit ;; BS2000:POSIX*:*:*) echo bs2000-siemens-sysv exit ;; DS/*:UNIX_System_V:*:*) echo ${UNAME_MACHINE}-${UNAME_SYSTEM}-${UNAME_RELEASE} exit ;; *:Plan9:*:*) # "uname -m" is not consistent, so use $cputype instead. 386 # is converted to i386 for consistency with other x86 # operating systems. if test "$cputype" = "386"; then UNAME_MACHINE=i386 else UNAME_MACHINE="$cputype" fi echo ${UNAME_MACHINE}-unknown-plan9 exit ;; *:TOPS-10:*:*) echo pdp10-unknown-tops10 exit ;; *:TENEX:*:*) echo pdp10-unknown-tenex exit ;; KS10:TOPS-20:*:* | KL10:TOPS-20:*:* | TYPE4:TOPS-20:*:*) echo pdp10-dec-tops20 exit ;; XKL-1:TOPS-20:*:* | TYPE5:TOPS-20:*:*) echo pdp10-xkl-tops20 exit ;; *:TOPS-20:*:*) echo pdp10-unknown-tops20 exit ;; *:ITS:*:*) echo pdp10-unknown-its exit ;; SEI:*:*:SEIUX) echo mips-sei-seiux${UNAME_RELEASE} exit ;; *:DragonFly:*:*) echo ${UNAME_MACHINE}-unknown-dragonfly`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'` exit ;; *:*VMS:*:*) UNAME_MACHINE=`(uname -p) 2>/dev/null` case "${UNAME_MACHINE}" in A*) echo alpha-dec-vms ; exit ;; I*) echo ia64-dec-vms ; exit ;; V*) echo vax-dec-vms ; exit ;; esac ;; *:XENIX:*:SysV) echo i386-pc-xenix exit ;; i*86:skyos:*:*) echo ${UNAME_MACHINE}-pc-skyos`echo ${UNAME_RELEASE}` | sed -e 's/ .*$//' exit ;; i*86:rdos:*:*) echo ${UNAME_MACHINE}-pc-rdos exit ;; i*86:AROS:*:*) echo ${UNAME_MACHINE}-pc-aros exit ;; x86_64:VMkernel:*:*) echo ${UNAME_MACHINE}-unknown-esx exit ;; esac eval $set_cc_for_build cat >$dummy.c < # include #endif main () { #if defined (sony) #if defined (MIPSEB) /* BFD wants "bsd" instead of "newsos". Perhaps BFD should be changed, I don't know.... */ printf ("mips-sony-bsd\n"); exit (0); #else #include printf ("m68k-sony-newsos%s\n", #ifdef NEWSOS4 "4" #else "" #endif ); exit (0); #endif #endif #if defined (__arm) && defined (__acorn) && defined (__unix) printf ("arm-acorn-riscix\n"); exit (0); #endif #if defined (hp300) && !defined (hpux) printf ("m68k-hp-bsd\n"); exit (0); #endif #if defined (NeXT) #if !defined (__ARCHITECTURE__) #define __ARCHITECTURE__ "m68k" #endif int version; version=`(hostinfo | sed -n 's/.*NeXT Mach \([0-9]*\).*/\1/p') 2>/dev/null`; if (version < 4) printf ("%s-next-nextstep%d\n", __ARCHITECTURE__, version); else printf ("%s-next-openstep%d\n", __ARCHITECTURE__, version); exit (0); #endif #if defined (MULTIMAX) || defined (n16) #if defined (UMAXV) printf ("ns32k-encore-sysv\n"); exit (0); #else #if defined (CMU) printf ("ns32k-encore-mach\n"); exit (0); #else printf ("ns32k-encore-bsd\n"); exit (0); #endif #endif #endif #if defined (__386BSD__) printf ("i386-pc-bsd\n"); exit (0); #endif #if defined (sequent) #if defined (i386) printf ("i386-sequent-dynix\n"); exit (0); #endif #if defined (ns32000) printf ("ns32k-sequent-dynix\n"); exit (0); #endif #endif #if defined (_SEQUENT_) struct utsname un; uname(&un); if (strncmp(un.version, "V2", 2) == 0) { printf ("i386-sequent-ptx2\n"); exit (0); } if (strncmp(un.version, "V1", 2) == 0) { /* XXX is V1 correct? */ printf ("i386-sequent-ptx1\n"); exit (0); } printf ("i386-sequent-ptx\n"); exit (0); #endif #if defined (vax) # if !defined (ultrix) # include # if defined (BSD) # if BSD == 43 printf ("vax-dec-bsd4.3\n"); exit (0); # else # if BSD == 199006 printf ("vax-dec-bsd4.3reno\n"); exit (0); # else printf ("vax-dec-bsd\n"); exit (0); # endif # endif # else printf ("vax-dec-bsd\n"); exit (0); # endif # else printf ("vax-dec-ultrix\n"); exit (0); # endif #endif #if defined (alliant) && defined (i860) printf ("i860-alliant-bsd\n"); exit (0); #endif exit (1); } EOF $CC_FOR_BUILD -o $dummy $dummy.c 2>/dev/null && SYSTEM_NAME=`$dummy` && { echo "$SYSTEM_NAME"; exit; } # Apollos put the system type in the environment. test -d /usr/apollo && { echo ${ISP}-apollo-${SYSTYPE}; exit; } # Convex versions that predate uname can use getsysinfo(1) if [ -x /usr/convex/getsysinfo ] then case `getsysinfo -f cpu_type` in c1*) echo c1-convex-bsd exit ;; c2*) if getsysinfo -f scalar_acc then echo c32-convex-bsd else echo c2-convex-bsd fi exit ;; c34*) echo c34-convex-bsd exit ;; c38*) echo c38-convex-bsd exit ;; c4*) echo c4-convex-bsd exit ;; esac fi cat >&2 < in order to provide the needed information to handle your system. config.guess timestamp = $timestamp uname -m = `(uname -m) 2>/dev/null || echo unknown` uname -r = `(uname -r) 2>/dev/null || echo unknown` uname -s = `(uname -s) 2>/dev/null || echo unknown` uname -v = `(uname -v) 2>/dev/null || echo unknown` /usr/bin/uname -p = `(/usr/bin/uname -p) 2>/dev/null` /bin/uname -X = `(/bin/uname -X) 2>/dev/null` hostinfo = `(hostinfo) 2>/dev/null` /bin/universe = `(/bin/universe) 2>/dev/null` /usr/bin/arch -k = `(/usr/bin/arch -k) 2>/dev/null` /bin/arch = `(/bin/arch) 2>/dev/null` /usr/bin/oslevel = `(/usr/bin/oslevel) 2>/dev/null` /usr/convex/getsysinfo = `(/usr/convex/getsysinfo) 2>/dev/null` UNAME_MACHINE = ${UNAME_MACHINE} UNAME_RELEASE = ${UNAME_RELEASE} UNAME_SYSTEM = ${UNAME_SYSTEM} UNAME_VERSION = ${UNAME_VERSION} EOF exit 1 # Local variables: # eval: (add-hook 'write-file-hooks 'time-stamp) # time-stamp-start: "timestamp='" # time-stamp-format: "%:y-%02m-%02d" # time-stamp-end: "'" # End: festival/missing0000755000175000017500000001421307262616703012410 0ustar awbawb#! /bin/sh # Common stub for a few missing GNU programs while installing. # Copyright (C) 1996, 1997 Free Software Foundation, Inc. # Franc,ois Pinard , 1996. # This program is free software; you can redistribute it and/or modify # it under the terms of the GNU General Public License as published by # the Free Software Foundation; either version 2, or (at your option) # any later version. # This program is distributed in the hope that it will be useful, # but WITHOUT ANY WARRANTY; without even the implied warranty of # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the # GNU General Public License for more details. # You should have received a copy of the GNU General Public License # along with this program; if not, write to the Free Software # Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA # 02111-1307, USA. if test $# -eq 0; then echo 1>&2 "Try \`$0 --help' for more information" exit 1 fi case "$1" in -h|--h|--he|--hel|--help) echo "\ $0 [OPTION]... PROGRAM [ARGUMENT]... Handle \`PROGRAM [ARGUMENT]...' for when PROGRAM is missing, or return an error status if there is no known handling for PROGRAM. Options: -h, --help display this help and exit -v, --version output version information and exit Supported PROGRAM values: aclocal touch file \`aclocal.m4' autoconf touch file \`configure' autoheader touch file \`config.h.in' automake touch all \`Makefile.in' files bison create \`y.tab.[ch]', if possible, from existing .[ch] flex create \`lex.yy.c', if possible, from existing .c lex create \`lex.yy.c', if possible, from existing .c makeinfo touch the output file yacc create \`y.tab.[ch]', if possible, from existing .[ch]" ;; -v|--v|--ve|--ver|--vers|--versi|--versio|--version) echo "missing - GNU libit 0.0" ;; -*) echo 1>&2 "$0: Unknown \`$1' option" echo 1>&2 "Try \`$0 --help' for more information" exit 1 ;; aclocal) echo 1>&2 "\ WARNING: \`$1' is missing on your system. You should only need it if you modified \`acinclude.m4' or \`configure.in'. You might want to install the \`Automake' and \`Perl' packages. Grab them from any GNU archive site." touch aclocal.m4 ;; autoconf) echo 1>&2 "\ WARNING: \`$1' is missing on your system. You should only need it if you modified \`configure.in'. You might want to install the \`Autoconf' and \`GNU m4' packages. Grab them from any GNU archive site." touch configure ;; autoheader) echo 1>&2 "\ WARNING: \`$1' is missing on your system. You should only need it if you modified \`acconfig.h' or \`configure.in'. You might want to install the \`Autoconf' and \`GNU m4' packages. Grab them from any GNU archive site." files=`sed -n 's/^[ ]*A[CM]_CONFIG_HEADER(\([^)]*\)).*/\1/p' configure.in` test -z "$files" && files="config.h" touch_files= for f in $files; do case "$f" in *:*) touch_files="$touch_files "`echo "$f" | sed -e 's/^[^:]*://' -e 's/:.*//'`;; *) touch_files="$touch_files $f.in";; esac done touch $touch_files ;; automake) echo 1>&2 "\ WARNING: \`$1' is missing on your system. You should only need it if you modified \`Makefile.am', \`acinclude.m4' or \`configure.in'. You might want to install the \`Automake' and \`Perl' packages. Grab them from any GNU archive site." find . -type f -name Makefile.am -print | sed 's/\.am$/.in/' | while read f; do touch "$f"; done ;; bison|yacc) echo 1>&2 "\ WARNING: \`$1' is missing on your system. You should only need it if you modified a \`.y' file. You may need the \`Bison' package in order for those modifications to take effect. You can get \`Bison' from any GNU archive site." rm -f y.tab.c y.tab.h if [ $# -ne 1 ]; then eval LASTARG="\${$#}" case "$LASTARG" in *.y) SRCFILE=`echo "$LASTARG" | sed 's/y$/c/'` if [ -f "$SRCFILE" ]; then cp "$SRCFILE" y.tab.c fi SRCFILE=`echo "$LASTARG" | sed 's/y$/h/'` if [ -f "$SRCFILE" ]; then cp "$SRCFILE" y.tab.h fi ;; esac fi if [ ! -f y.tab.h ]; then echo >y.tab.h fi if [ ! -f y.tab.c ]; then echo 'main() { return 0; }' >y.tab.c fi ;; lex|flex) echo 1>&2 "\ WARNING: \`$1' is missing on your system. You should only need it if you modified a \`.l' file. You may need the \`Flex' package in order for those modifications to take effect. You can get \`Flex' from any GNU archive site." rm -f lex.yy.c if [ $# -ne 1 ]; then eval LASTARG="\${$#}" case "$LASTARG" in *.l) SRCFILE=`echo "$LASTARG" | sed 's/l$/c/'` if [ -f "$SRCFILE" ]; then cp "$SRCFILE" lex.yy.c fi ;; esac fi if [ ! -f lex.yy.c ]; then echo 'main() { return 0; }' >lex.yy.c fi ;; makeinfo) echo 1>&2 "\ WARNING: \`$1' is missing on your system. You should only need it if you modified a \`.texi' or \`.texinfo' file, or any other file indirectly affecting the aspect of the manual. The spurious call might also be the consequence of using a buggy \`make' (AIX, DU, IRIX). You might want to install the \`Texinfo' package or the \`GNU make' package. Grab either from any GNU archive site." file=`echo "$*" | sed -n 's/.*-o \([^ ]*\).*/\1/p'` if test -z "$file"; then file=`echo "$*" | sed 's/.* \([^ ]*\) *$/\1/'` file=`sed -n '/^@setfilename/ { s/.* \([^ ]*\) *$/\1/; p; q; }' $file` fi touch $file ;; *) echo 1>&2 "\ WARNING: \`$1' is needed, and you do not seem to have it handy on your system. You might have modified some files without having the proper tools for further handling them. Check the \`README' file, it often tells you about the needed prerequirements for installing this package. You may also peek at any GNU archive site, in case some other package would contain this missing \`$1' program." exit 1 ;; esac exit 0 festival/install-sh0000755000175000017500000001273607262616703013025 0ustar awbawb#!/bin/sh # # install - install a program, script, or datafile # This comes from X11R5 (mit/util/scripts/install.sh). # # Copyright 1991 by the Massachusetts Institute of Technology # # Permission to use, copy, modify, distribute, and sell this software and its # documentation for any purpose is hereby granted without fee, provided that # the above copyright notice appear in all copies and that both that # copyright notice and this permission notice appear in supporting # documentation, and that the name of M.I.T. not be used in advertising or # publicity pertaining to distribution of the software without specific, # written prior permission. M.I.T. makes no representations about the # suitability of this software for any purpose. It is provided "as is" # without express or implied warranty. # # Calling this script install-sh is preferred over install.sh, to prevent # `make' implicit rules from creating a file called install from it # when there is no Makefile. # # This script is compatible with the BSD install script, but was written # from scratch. It can only install one file at a time, a restriction # shared with many OS's install programs. # set DOITPROG to echo to test this script # Don't use :- since 4.3BSD and earlier shells don't like it. doit="${DOITPROG-}" # put in absolute paths if you don't have them in your path; or use env. vars. mvprog="${MVPROG-mv}" cpprog="${CPPROG-cp}" chmodprog="${CHMODPROG-chmod}" chownprog="${CHOWNPROG-chown}" chgrpprog="${CHGRPPROG-chgrp}" stripprog="${STRIPPROG-strip}" rmprog="${RMPROG-rm}" mkdirprog="${MKDIRPROG-mkdir}" transformbasename="" transform_arg="" instcmd="$mvprog" chmodcmd="$chmodprog 0755" chowncmd="" chgrpcmd="" stripcmd="" rmcmd="$rmprog -f" mvcmd="$mvprog" src="" dst="" dir_arg="" while [ x"$1" != x ]; do case $1 in -c) instcmd="$cpprog" shift continue;; -d) dir_arg=true shift continue;; -m) chmodcmd="$chmodprog $2" shift shift continue;; -o) chowncmd="$chownprog $2" shift shift continue;; -g) chgrpcmd="$chgrpprog $2" shift shift continue;; -s) stripcmd="$stripprog" shift continue;; -t=*) transformarg=`echo $1 | sed 's/-t=//'` shift continue;; -b=*) transformbasename=`echo $1 | sed 's/-b=//'` shift continue;; *) if [ x"$src" = x ] then src=$1 else # this colon is to work around a 386BSD /bin/sh bug : dst=$1 fi shift continue;; esac done if [ x"$src" = x ] then echo "install: no input file specified" exit 1 else true fi if [ x"$dir_arg" != x ]; then dst=$src src="" if [ -d $dst ]; then instcmd=: chmodcmd="" else instcmd=mkdir fi else # Waiting for this to be detected by the "$instcmd $src $dsttmp" command # might cause directories to be created, which would be especially bad # if $src (and thus $dsttmp) contains '*'. if [ -f $src -o -d $src ] then true else echo "install: $src does not exist" exit 1 fi if [ x"$dst" = x ] then echo "install: no destination specified" exit 1 else true fi # If destination is a directory, append the input filename; if your system # does not like double slashes in filenames, you may need to add some logic if [ -d $dst ] then dst="$dst"/`basename $src` else true fi fi ## this sed command emulates the dirname command dstdir=`echo $dst | sed -e 's,[^/]*$,,;s,/$,,;s,^$,.,'` # Make sure that the destination directory exists. # this part is taken from Noah Friedman's mkinstalldirs script # Skip lots of stat calls in the usual case. if [ ! -d "$dstdir" ]; then defaultIFS=' ' IFS="${IFS-${defaultIFS}}" oIFS="${IFS}" # Some sh's can't handle IFS=/ for some reason. IFS='%' set - `echo ${dstdir} | sed -e 's@/@%@g' -e 's@^%@/@'` IFS="${oIFS}" pathcomp='' while [ $# -ne 0 ] ; do pathcomp="${pathcomp}${1}" shift if [ ! -d "${pathcomp}" ] ; then $mkdirprog "${pathcomp}" else true fi pathcomp="${pathcomp}/" done fi if [ x"$dir_arg" != x ] then $doit $instcmd $dst && if [ x"$chowncmd" != x ]; then $doit $chowncmd $dst; else true ; fi && if [ x"$chgrpcmd" != x ]; then $doit $chgrpcmd $dst; else true ; fi && if [ x"$stripcmd" != x ]; then $doit $stripcmd $dst; else true ; fi && if [ x"$chmodcmd" != x ]; then $doit $chmodcmd $dst; else true ; fi else # If we're going to rename the final executable, determine the name now. if [ x"$transformarg" = x ] then dstfile=`basename $dst` else dstfile=`basename $dst $transformbasename | sed $transformarg`$transformbasename fi # don't allow the sed command to completely eliminate the filename if [ x"$dstfile" = x ] then dstfile=`basename $dst` else true fi # Make a temp file name in the proper directory. dsttmp=$dstdir/#inst.$$# # Move or copy the file name to the temp name $doit $instcmd $src $dsttmp && trap "rm -f ${dsttmp}" 0 && # and set any options; do chmod last to preserve setuid bits # If any of these fail, we abort the whole thing. If we want to # ignore errors from any of these, just make sure not to ignore # errors from the above "$doit $instcmd $src $dsttmp" command. if [ x"$chowncmd" != x ]; then $doit $chowncmd $dsttmp; else true;fi && if [ x"$chgrpcmd" != x ]; then $doit $chgrpcmd $dsttmp; else true;fi && if [ x"$stripcmd" != x ]; then $doit $stripcmd $dsttmp; else true;fi && if [ x"$chmodcmd" != x ]; then $doit $chmodcmd $dsttmp; else true;fi && # Now rename the file to the real destination. $doit $rmcmd -f $dstdir/$dstfile && $doit $mvcmd $dsttmp $dstdir/$dstfile fi && exit 0 festival/mkinstalldirs0000755000175000017500000000132107262616703013613 0ustar awbawb#! /bin/sh # mkinstalldirs --- make directory hierarchy # Author: Noah Friedman # Created: 1993-05-16 # Public domain # $Id: mkinstalldirs,v 1.1 2001/04/04 13:12:35 awb Exp $ errstatus=0 for file do set fnord `echo ":$file" | sed -ne 's/^:\//#/;s/^://;s/\// /g;s/^#/\//;p'` shift pathcomp= for d do pathcomp="$pathcomp$d" case "$pathcomp" in -* ) pathcomp=./$pathcomp ;; esac if test ! -d "$pathcomp"; then echo "mkdir $pathcomp" mkdir "$pathcomp" || lasterr=$? if test ! -d "$pathcomp"; then errstatus=$lasterr fi fi pathcomp="$pathcomp/" done done exit $errstatus # mkinstalldirs ends here festival/config/Makefile0000644000175000017500000000574207304237411013715 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Makefile for config directory ## ## ## ########################################################################### TOP=.. DIRNAME=config CONFIGS = config.in vc_config_make_rules-dist FILES=Makefile common_make_rules vc_common_make_rules \ test_make_rules project.mak project_config_check.mak \ system.sh make_system.mak \ $(CONFIGS) ALL_DIRS = modules systems include $(TOP)/config/common_make_rules festival/config/common_make_rules0000644000175000017500000000707007262605650015702 0ustar awbawb ########################################################-*-mode:Makefile-*- ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Various People ## ## : Reorganised (and probably broken) ## ## : by Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: June 1997 ## ## --------------------------------------------------------------------- ## ## Default Makefile rules includede everywhere. ## ## ## ########################################################################### # This is the default rule all: $(ALL) .sub_directories @ : Do nothing but shut up make # Include generated system description include $(TOP)/config/system.mak ifeq ($(SYSTEM_LOADED),) MACHINETYPE=unknown OSTYPE=unknown OSREV= endif # Include project specific rules -include $(TOP)/config/project.mak # Include installation specific information -include $(TOP)/config/config # indirect to shared rule sets directory include $(EST)/config/rules/common_make_rules.mak festival/config/vc_common_make_rules0000644000175000017500000000331307262605650016366 0ustar awbawb ###################################################################### # # # Make rules for MicroCruft Visual C++ # # # ###################################################################### !include $(TOP)\config\vc_config_make_rules !include $(TOP)\config\project.mak .SUFFIXES: .cc .obj CPP=cl /nologo /DSYSTEM_IS_WIN32=1 /DINSTANTIATE_TEMPLATES=1 $(MODULEFLAGS) CC=cl /nologo /DSYSTEM_IS_WIN32=1 $(MODULEFLAGS) default_target: $(DIRS) all @echo done $(DIRNAME) all: $(ALL) $(DIRS) x1 : FORCE @echo building in $(DIRNAME)\$@ @cd $@ @nmake /nologo /fVCMakefile @cd .. FORCE: .vcbuildlib: $(OBJS) @echo add to $(INLIB) $(OBJS) @if EXIST $(INLIB) lib/nologo $(INLIB) $(OBJS) @if NOT EXIST $(INLIB) lib/nologo /out:$(INLIB) $(OBJS) @echo built > .vcbuildlib .vc_add_to_lib: $(TOADD) lib/nologo $(ADDLIB) $(TOADD) @echo built > .vc_add_to_lib .libraries: @echo Libraries not touched for VC++ relink: @echo Links not made for Visual C++ .vc_build_scripts: @echo Scripts not built for Visual C++ .vc_build_manpages: @echo manual pages not built for Visual C++ .config_error:: @echo Config OK .sub_directories: $(BUILD_DIRS) .remove_links: @echo Links not made for Visual C++ .process_scripts: @echo Scripts not created for VC .process_docs: @echo Documentation not created for VC .link_main: @echo Links not made for Visual C++ .cc.obj: $(CPP) $(CFLAGS) /c /Tp$*.cc /Fo$*.obj .c.obj: $(CC) $(CFLAGS) /c /Tc$*.c /Fo$*.obj # this dummy rule stops the comment in make.depend getting # interpreted as a command... YEUCH hack_dummy_target: festival/config/test_make_rules0000644000175000017500000001117207304555116015365 0ustar awbawb ########################################################-*-mode:Makefile-*- ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### # Makefile rules for testing things # ########################################################################### TEST_PROGRAMS = $(TEST_MODULES:%=%_example) $(TEST_MODULES:%=%_regression) #TSRCS = $(TEST_PROGRAMS:%=%.C) #SRCS = $(TSRCS) #OBJS = $(SRCS:%.C=%.o) include $(TOP)/config/common_make_rules test_scripts: $(TEST_SCRIPTS:%=%_script_test) test_modules: $(TEST_MODULES:%=%_module_build_and_test) $(TEST_MODULES:%=%_module_build_and_test) : %_module_build_and_test : %_module_rebuild %_module_test $(TEST_MODULES:%=%_module_rebuild) : %_module_rebuild : @echo 'build $* (module)' @/bin/rm -f $(OBJS) @if $(MAKE) --no-print-directory OPTIMISE=$(TEST_OPTIMISE) WARN=1 $*_example $*_regression ;\ then \ : ;\ else \ echo $* example status: FAILED ; exit 1 ;\ fi $(TEST_MODULES:%=%_module_test) : %_module_test : correct/%_example.out correct/%_regression.out @echo 'test $* (module)' @if ./$*_example $($(*:=_example_args)) > $*_example.out ;\ then \ echo $*_example completed ;\ if [ ! -f $*_example.out ] || diff $*_example.out correct/$*_example.out ;\ then \ echo $* example status: CORRECT ;\ else \ echo $* example status: INCORRECT ;\ fi ;\ else \ echo $* example status: FAILED ;\ fi @if ./$*_regression $($(*:=_regression_args)) > $*_regression.out ;\ then \ echo $*_regression completed ;\ if [ ! -f $*_regression.out ] || diff $*_regression.out correct/$*_regression.out ;\ then \ echo $* regression status: CORRECT ;\ else \ echo $* regression status: INCORRECT ;\ fi ;\ else \ echo $* regression status: FAILED ;\ fi @echo @echo $(TEST_SCRIPTS:%=%_script_test) : %_script_test : %.sh correct/%_script.out @echo 'test $* (script)' @OUTPUT='$*_script.out' ;\ TOP='$(TOP)' ;\ export TOP OUTPUT ;\ if /bin/sh $*.sh $($(*:=_script_args)) ;\ then \ echo $* script completed ;\ if [ ! -f $*_script.out ] || diff $*_script.out correct/$*_script.out ;\ then \ echo $* script status: CORRECT ;\ else \ echo $* script status: INCORRECT ;\ fi ;\ else \ echo $* script status: FAILED ;\ fi @echo @echo $(SRCS:%.C=%.o) : %.o : %.C % : %.o $(CXX) $(CXXFLAGS) $(TEMPLATES) -o $@ $@.o $(ESTLIB) $($(@:=_LIBS)) $(LIBS) festival/config/project.mak0000644000175000017500000001140012445313067014405 0ustar awbawb ########################################################-*-mode:Makefile-*- ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Tue Oct 7 1997 ## ## -------------------------------------------------------------------- ## ## Description of festival. ## ## ## ########################################################################### PROJECT_NAME = Festival Speech Synthesis System PROJECT_PREFIX = FESTIVAL PROJECT_VERSION = 2.4 PROJECT_DATE = December 2014 PROJECT_STATE = release # config files of projects we depend on PROJECT_OTHER_CONFIGS = $(EST)/config/config # Place to find the optional modules for this project. MODULE_DIRECTORY = $(TOP)/src/modules DISTRIBUTED_MODULES = \ JAVA DEVELOPMENT_MODULES = RJC_SYNTHESIS \ UNISYN_PHONOLOGY UTILITY_MODULES = ALL_REAL_MODULES = \ $(DISTRIBUTED_MODULES) \ $(DEVELOPMENT_MODULES) ALL_MODULES = \ $(ALL_REAL_MODULES) \ $(UTILITY_MODULES) # Place where programs are compiled PROJECT_MAIN_DIR=$(FESTIVAL_HOME)/src/main PROJECT_SCRIPTS_DIR=$(FESTIVAL_HOME)/src/scripts # Where the main RCS tree is, probably only used within CSTR CENTRAL_DIR = $(LOCAL_REPOSITORY)/festival/code_base/festival # Libraries defined in this project PROJECT_LIBRARIES = Festival PROJECT_LIBRARY_DIR_Festival = $(TOP)/src/lib PROJECT_DEFAULT_LIBRARY = Festival # Libraries used from other projects REQUIRED_LIBRARIES = estools estbase eststring REQUIRED_LIBRARY_DIR_estools = $(EST)/lib REQUIRED_LIBRARY_DIR_estbase = $(EST)/lib REQUIRED_LIBRARY_DIR_eststring = $(EST)/lib REQUIRED_MAKE_INCLUDE = $(EST)/make.include # Includes for this and related projects PROJECT_INCLUDES = -I$(TOP)/src/include -I$(EST)/include PROJECT_TEMPLATE_DIRS = src/arch/festival PROJECT_TEMPLATE_DBS = $(TOP) $(EST) LIBRARY_TEMPLATE_DIRS_estools = $(LIBRARY_TEMPLATE_DIRS:%=$(EST)/%) JAVA_CLASS_LIBRARY = $(TOP)/src/lib/festival.jar JAVA_CLASSPATH= $(TOP)/lib/festival.jar:$(EST_HOME)/lib/est_$(EST_JAVA_VERSION).jar:$(SYSTEM_JAVA_CLASSPATH) PROJECT_JAVA_ROOT=$(TOP)/src/modules/java # Places to look for documentation DOCXX_DIRS = $(TOP)/src MODULE_TO_DOCXX = perl $(TOP)/src/modules/utilities/extract_module_doc++.prl FTLIBDIR = $(FESTIVAL_HOME)/lib festival/config/project_config_check.mak0000644000175000017500000000515407262605650017103 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ifndef INCLUDE_SIOD .config_error:: FORCE @echo "+--------------------------------------------------" @echo "| Must compile speech tools with SIOD support" @echo "+--------------------------------------------------" @exit 1 endif festival/config/system.sh0000644000175000017500000001165607603316046014142 0ustar awbawb#!/bin/sh ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Guess what kind of system we are on. ## ## ## ########################################################################### # Where the Makefile fragments live SYSTEMS=$1 # Drop and _xxx from the end OSTYPE=`uname -s | sed -e '/^\([^_]*\).*/s//\1/' -e '/\//s///g'` # CPU, downcased, /s and some uninteresting details eliminated MACHINETYPE=`{ mach || uname -m || echo unknown ; } 2>/dev/null | tr ABCDEFGHIJKLMNOPQRSTUVWXYZ/ abcdefghijklmnopqrstuvwxyz_ | sed -e 's/i[0-9]86/ix86/' \ -e 's/sun4/sparc/' \ -e 's/ip[0-9]*/ip/'\ -e 's/ /_/g'\ -e 's/9000_7../hp9000/' ` # OR revision, only take first two numbers. OSREV=`{ uname -r || echo ""; } 2> /dev/null | sed -e 's/^\([^.]*\)\(\.[^-. ]*\).*/\1\2/'` # Sort out various flavours of Linux if [ "$OSTYPE" = Linux ] then if [ -f "/etc/redhat-release" ] then OSTYPE=RedHatLinux OSREV=`cat /etc/redhat-release | sed -e 's/[^0-9]*\([0-9.]*\).*/\1/'` elif [ -f "/etc/debian_version" ] then OSTYPE=DebianGNULinux OSREV=`cat /etc/debian_version` else # Generic unknown GNU/Linux system. OSTYPE=Linux fi fi # Make sure we actually have a .mak file for it, otherwise fall back # to sensible defaults (for example, kernel version and architecture # are completely irrelevant on Linux) if [ ! -f "${SYSTEMS}/${MACHINETYPE}_${OSTYPE}${OSREV}.mak" ]; then if [ -f "${SYSTEMS}/${OSTYPE}${OSREV}.mak" ]; then MACHINETYPE=unknown elif [ -f "${SYSTEMS}/${MACHINETYPE}_${OSTYPE}.mak" ]; then OSREV= elif [ -f "${SYSTEMS}/unknown_${OSTYPE}.mak" ]; then MACHINETYPE=unknown OSREV= elif [ "$OSTYPE" = "RedHatLinux" -o "$OSTYPE" = "DebianGNULinux" ]; then MACHINETYPE=unknown OSTYPE=Linux OSREV= elif [ "$OSTYPE" = "Darwin" ]; then OSREV= else OSTYPE=unknown OSREV= fi fi echo ' ###########################################################################' echo ' ## This file is created automatically from your config file.' echo ' ## Do not hand edit.' echo ' ## Created:'`date` echo ' ###########################################################################' echo '' echo "OSTYPE:=$OSTYPE" echo "MACHINETYPE:=$MACHINETYPE" echo "OSREV:=$OSREV" echo "SYSTEM_LOADED:=1" exit 0 festival/config/make_system.mak0000644000175000017500000000606407304237411015266 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Thu Oct 2 1997 ## ## -------------------------------------------------------------------- ## ## Guess what kind of system we are on. ## ## ## ########################################################################### system.mak : config @echo Check system type >&2 @/bin/sh $(TOP)/config/system.sh $(TOP)/config/systems > system.mak festival/config/config.in0000644000175000017500000000525511337413435014054 0ustar awbawb########################################################-*-mode:Makefile-*- ## ## ## Festival: local configuration file ## ## ## ########################################################################### ## ## Specific config file for local installation ## ########################################################################### ## Which speech tools to use EST=$(TOP)/../speech_tools ########################################################################### ## Where the festival tree will be installed. ## ## The default is that festival will remain where it is compiled. ## ## You may need to set this explicitly if automounter or NFS ## side effects cause problems FESTIVAL_HOME := $(shell (cd $(TOP); pwd)) ########################################################################### ## Feature selection. ## ## Select modules to include. ## Non Free PSOLA synthesis. This isn't distributed with festival because ## of a patent, if you have src/modules/diphone/di_psolaTM.cc you can ## include this feature. # INCLUDE_PSOLA_TM=1 ## Support for TCL. So that festival may eval TCL commands and TCL may eval ## festival commands. This was added to support the CSLU toolkit but ## others may want it too. # INCLUDE_TCL=1 ########################################################################### ## Take most settings from speech tools. include $(EST)/config/config ########################################################################### ## Add any extra modules you wish to include ## These sub modules are *optional* and unless you know what they are ## you probabaly don't want them or need them. They are typically ## new code that isn't yet stable yet and being used for research or ## old code left in for compatibility for some users ## Experimental UniSyn, metrical tree, phonological structure matching ## code # ALSO_INCLUDE += UniSyn_phonology UniSyn_selection ## Cluster unit selection code as described in "Building Voices in ## Festival", again experimental and suitable for research purposes only. ALSO_INCLUDE += clunits clustergen MultiSyn ## NITECH and Tokyo Institute of Technologies HTS support ALSO_INCLUDE += hts_engine ## Old diphone code that will be delete, left in only for some ## compatibility # ALSO_INCLUDE += diphone ## Other (non-Edinburgh) modules may also be specified here (e.g. OGI code), ALSO_INCLUDE += ########################################################################### ## ## Describe your local system below by redefining things defined ## in config/configs/default.mak. festival/config/vc_config_make_rules-dist0000644000175000017500000000210511366534447017307 0ustar awbawb ###################################################################### # # # Configuration variable make settings for MicroCruft Visual C++ # # # ###################################################################### EST=$(TOP)\..\speech_tools SYSTEM_LIB=c:\\festival\\lib MODULEFLAGS=/DSUPPORT_EDITLINE=1 AUDIOFLAGS= DEBUGFLAGS= /Zi LINKDEBUGFLAGS = /debug OPTFLAGS= /EHsc /wd4675 /GR DEFINES=/DFTLIBDIRC=$(SYSTEM_LIB) "/DFTOSTYPEC=win32_vc" "/DFTNAME=$(PROJECT_NAME)" "/DFTVERSION=$(PROJECT_VERSION)" "/DFTSTATE=$(PROJECT_STATE)" "/DFTDATE=$(PROJECT_DATE)" INCLUDEFLAGS= /I$(TOP)/src/include /I$(EST)\include $(LOCAL_INCLUDES) /DNO_READLINE=1 /DNO_SPOOLER=1 $(DEFINES) LINKFLAGS=$(LINKDEBUGFLAGS) LIB_DIR=src\lib FESTLIBS = $(TOP)\src\lib\libfestival.lib ESTLIBS = $(EST)\lib\libestools.lib $(EST)\lib\libestbase.lib $(EST)\lib\libeststring.lib WINLIBS = wsock32.lib winmm.lib !ifndef VCLIBS VCLIBS = $(FESTLIBS) $(ESTLIBS) !endif festival/config/modules/Makefile0000644000175000017500000000613410077563027015367 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Makefile for config directory ## ## ## ########################################################################### TOP=../.. DIRNAME=config/modules RULESETS = efence.mak dmalloc.mak \ psola_tm.mak editline.mak tcl.mak \ freebsd16_audio.mak irix_audio.mak linux16_audio.mak \ sun16_audio.mak win32_audio.mak macosx_audio.mak \ mplayer_audio.mak nas_audio.mak esd_audio.mak native_audio.mak \ siod.mak wagon.mak scfg.mak wfst.mak ols.mak debugging.mak FILES = Makefile descriptions $(RULESETS) include $(TOP)/config/common_make_rules festival/config/modules/descriptions0000644000175000017500000000136107262605650016356 0ustar awbawb ########################################################################### ## ## ## Descriptions of uninteresting modules. ## ## ## ########################################################################### desc_asr="(From EST) Speech recognition code" desc_native_audio="(from EST) Native audio module for your system" desc_ols="(from EST) Ordinary Least Squares support" desc_scfg="(from EST) Stochastic context free grammars" desc_siod="(from EST) Scheme In One Defun" desc_wagon="(from EST) Wagon CART tree system" desc_wfst="(from EST) Weighted Finite State Automata" festival/config/modules/efence.mak0000644000175000017500000000607207262605650015650 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Defenitions for whan compiling with efence. ## ## ## ########################################################################### INCLUDE_EFENCE=1 MOD_DESC_EFENCE=Compile with efence memory error detection package DEBUG_LIBS += -L$(EFENCE_LIB) -lefence festival/config/modules/dmalloc.mak0000644000175000017500000000627607262605650016044 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed Jun 3 1998 ## ## -------------------------------------------------------------------- ## ## Link in the dmalloc library and include the header. ## ## ## ########################################################################### ifndef INCLUDE_DMALLOC INCLUDE_DMALLOC=1 endif MOD_DESC_DMALLOC=Compile with debugging malloc library DEBUG_LIBS += -L$(DMALLOC_LIB) -ldmalloc ifdef DMALLOC_INCLUDE DEBUG_DEFINES += -I $(DMALLOC_INCLUDE) endif DEBUG_DEFINES += -DINCLUDE_DMALLOC festival/config/modules/psola_tm.mak0000644000175000017500000000614007262605650016235 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Defenitions for psola(tm) audio support. ## ## ## ########################################################################### INCLUDE_PSOLA_TM=1 MOD_DESC_PSOLA_TM=Include PSOLA(tm) synthesis code. MODULE_DIPHONE_DEFINES += -DSUPPORT_PSOLA_TM $(MODULE_PSOLA_TM_OPTIONS:%=-DPSOLA_TM_%) festival/config/modules/editline.mak0000644000175000017500000000633507603316046016217 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## Date: December 1998 ## ## -------------------------------------------------------------------- ## ## Command line editor based on editline ## ## ## ########################################################################### INCLUDE_EDITLINE=1 MOD_DESC_EDITLINE=Use editline for command line editing and history IO_DEFINES += -DSUPPORT_EDITLINE $(MODULE_EDITLINE_OPTIONS:%=-DEDITLINE_%) MODULE_LIBS += $(TERMCAPLIB) ifeq ($(DIRNAME),siod) CSRCS := $(CSRCS) el_complete.c editline.c el_sys_unix.c endif festival/config/modules/tcl.mak0000644000175000017500000000616007262605650015203 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Defenitions for TCL audio support. ## ## ## ########################################################################### INCLUDE_TCL=1 MOD_DESC_TCL=(from EST) Include TCL interface FESTIVAL_DEFINES += -DSUPPORT_TCL FESTIVAL_INCLUDES += -I$(TCL_INCLUDE) MODULE_LIBS += -L$(TCL_LIB) -ltcl7.6 festival/config/modules/freebsd16_audio.mak0000644000175000017500000000611207262605650017360 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Defenitions for Freebsd 16 bit audio support. ## ## ## ########################################################################### INCLUDE_FREEBSD16_AUDIO=1 MOD_DESC_FREEBSD16_AUDIO=(from EST) Native audio module for FreeBSD systems AUDIO_DEFINES += -DSUPPORT_FREEBSD16 festival/config/modules/irix_audio.mak0000644000175000017500000000607007262605650016555 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Defenitions for Irix audio support. ## ## ## ########################################################################### INCLUDE_IRIX_AUDIO=1 MOD_DESC_IRIX_AUDIO=(from EST) Native audio module for Irix systems AUDIO_DEFINES += -DSUPPORT_IRIX festival/config/modules/linux16_audio.mak0000644000175000017500000000640411110017036017067 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Defenitions for Linux 16 bit audio support. ## ## ## ########################################################################### INCLUDE_LINUX16_AUDIO=1 MOD_DESC_LINUX16_AUDIO=(from EST) Native audio module for Linux systems ifeq ($(LINUXAUDIO),alsa) AUDIO_DEFINES += -DSUPPORT_ALSALINUX MODULE_LIBS += -lasound endif ifeq ($(LINUXAUDIO),none) AUDIO_DEFINES += -DSUPPORT_VOXWARE endif ifdef INCLUDE_JAVA_CPP MODULE_LIBS += -lpthread endif festival/config/modules/sun16_audio.mak0000644000175000017500000000607607262605650016564 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Defenitions for Sun 16 bit audio support. ## ## ## ########################################################################### INCLUDE_SUN16_AUDIO=1 MOD_DESC_SUN16_AUDIO=(from EST) Native audio module for Solaris systems AUDIO_DEFINES += -DSUPPORT_SUN16 festival/config/modules/win32_audio.mak0000644000175000017500000000606307262605650016546 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Defenitions for Win32 audio support. ## ## ## ########################################################################### INCLUDE_WIN32_AUDIO=1 MOD_DESC_WIN32_AUDIO=(from EST) Audio for Win32 systems AUDIO_DEFINES += -DSUPPORT_WIN32AUDIO festival/config/modules/macosx_audio.mak0000644000175000017500000000153210401517371017061 0ustar awbawb ########################################################################### ## ## ## Author: Brian Foley (bfoley@compsoc.nuigalway.ie) ## ## Date: Wed Feb 17 2004 ## ## -------------------------------------------------------------------- ## ## Definitions for MacOS X audio support. ## ## ## ########################################################################### INCLUDE_MACOSX_AUDIO=1 MOD_DESC_MACOSX_AUDIO=(from EST) CoreAudio audio module for MacOS X systems AUDIO_DEFINES += -DSUPPORT_MACOSX_AUDIO MODULE_LIBS += -framework CoreAudio -framework AudioUnit -framework AudioToolbox -framework Carbon festival/config/modules/mplayer_audio.mak0000644000175000017500000000610507262605650017252 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Defenitions for Mplayer audio support. ## ## ## ########################################################################### INCLUDE_MPLAYER_AUDIO=1 MOD_DESC_MPLAYER_AUDIO=(from EST) Audio module for calling windows mplayer AUDIO_DEFINES += -DSUPPORT_MPLAYER festival/config/modules/nas_audio.mak0000644000175000017500000000620207262605650016360 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Defenitions for NAS audio support. ## ## ## ########################################################################### INCLUDE_NAS_AUDIO=1 MOD_DESC_NAS_AUDIO=(from EST) Use Network Audio AUDIO_DEFINES += -DSUPPORT_NAS AUDIO_INCLUDES += -I$(NAS_INCLUDE) MODULE_LIBS += -L$(NAS_LIB) -laudio -L$(X11_LIB) -lX11 -lXt festival/config/modules/esd_audio.mak0000644000175000017500000000616007262605650016355 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Defenitions for ESD audio support. ## ## ## ########################################################################### INCLUDE_ESD_AUDIO=1 MOD_DESC_ESD_AUDIO=(from EST) Use ESD Audio AUDIO_DEFINES += -DSUPPORT_ESD AUDIO_INCLUDES += -I$(ESD_INCLUDE) MODULE_LIBS += -L$(ESD_LIB) -lesd -laudiofile festival/config/modules/native_audio.mak0000644000175000017500000000571707262605650017077 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Dummy module to document native audio selection. ## ## ## ########################################################################### INCLUDE_NATIVE_AUDIO=1 MOD_DESC_NATIVE_AUDIO=(from EST) Native audio module for your system festival/config/modules/siod.mak0000644000175000017500000000577207262605650015367 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Mon Jun 1 1998 ## ## -------------------------------------------------------------------- ## ## NOthing special to do. ## ## ## ########################################################################### INCLUDE_SIOD=1 MOD_DESC_SIOD=(from EST) Scheme In One Defun festival/config/modules/wagon.mak0000644000175000017500000000600107262605650015526 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Mon Jun 1 1998 ## ## -------------------------------------------------------------------- ## ## NOthing special to do. ## ## ## ########################################################################### INCLUDE_WAGON=1 MOD_DESC_WAGON=(from EST) Wagon CART tree system festival/config/modules/scfg.mak0000644000175000017500000000600607262605650015342 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Mon Jun 1 1998 ## ## -------------------------------------------------------------------- ## ## NOthing special to do. ## ## ## ########################################################################### INCLUDE_SCFG=1 MOD_DESC_SCFG=(from EST) Stochastic context free grammars festival/config/modules/wfst.mak0000644000175000017500000000600507262605650015402 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Mon Jun 1 1998 ## ## -------------------------------------------------------------------- ## ## NOthing special to do. ## ## ## ########################################################################### INCLUDE_WFST=1 MOD_DESC_WFST=(from EST) Weighted Finite State Automata festival/config/modules/ols.mak0000644000175000017500000000600507262605650015214 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Mon Jun 1 1998 ## ## -------------------------------------------------------------------- ## ## NOthing special to do. ## ## ## ########################################################################### INCLUDE_OLS=1 MOD_DESC_OLS=(from EST) Ordinary Least Squares support festival/config/modules/debugging.mak0000644000175000017500000000600607262605650016353 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Turn on various debugging facilities. ## ## ## ########################################################################### ifndef INCLUDE_DEBUGGING INCLUDE_DEBUGGING=1 endif MOD_DESC_DEBUGGING=Compile in some debugging facilities MODULE_DEFINES += -DEST_DEBUGGING festival/config/systems/Makefile0000644000175000017500000001042011363323252015410 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Makefile for config directory ## ## ## ########################################################################### TOP=../.. DIRNAME=config/systems SYSTEMS = \ Linux.mak \ RedHatLinux.mak \ DebianGNULinux.mak \ alpha_Linux.mak \ alpha_OSF1V4.0.mak \ alpha_RedHatLinux.mak \ hp9000_HP-UX.mak \ hp9000_HP-UXB.10.mak \ ip_IRIX.mak \ ip_IRIX5.3.mak \ ip_IRIX6.3.mak \ ip_IRIX6.4.mak \ ip_IRIX646.4.mak \ ix86_CYGWIN1.0.mak \ ix86_CYGWIN1.1.mak \ ix86_CYGWIN1.3.mak \ ix86_CYGWIN1.4.mak \ ix86_CYGWIN1.5.mak \ ix86_CYGWIN1.7.mak \ ix86_CYGWIN20.1.mak \ ix86_CYGWIN32.mak \ ix86_CYGWIN324.0.mak \ ix86_Darwin.mak \ ix86_FreeBSD.mak \ ix86_FreeBSD2.1.mak \ ix86_FreeBSD2.2.mak \ ix86_FreeBSD3.0.mak \ ix86_FreeBSD3.1.mak \ ix86_FreeBSD3.2.mak \ ix86_FreeBSD3.3.mak \ ix86_FreeBSD4.0.mak \ ix86_OS22.mak \ ix86_RedHatLinux4.0.mak \ ix86_RedHatLinux4.1.mak \ ix86_RedHatLinux4.2.mak \ ix86_RedHatLinux5.0.mak \ ix86_RedHatLinux5.1.mak \ ix86_RedHatLinux5.2.mak \ ix86_RedHatLinux6.0.mak \ ix86_RedHatLinux6.1.mak \ ix86_RedHatLinux6.2.mak \ ix86_RedHatLinux7.0.mak \ ix86_SunOS5.5.mak \ ix86_SunOS5.6.mak \ ix86_SunOS5.7.mak \ ix86_SunOS5.8.mak \ ix86_SunOS5.mak \ rs6000_AIX4.1.mak \ sparc_SunOS4.1.mak \ sparc_SunOS4.mak \ sparc_SunOS5.5.mak \ sparc_SunOS5.6.mak \ sparc_SunOS5.7.mak \ sparc_SunOS5.8.mak \ sparc_SunOS5.mak \ unknown_DebianGNULinux.mak \ unknown_Linux.mak \ unknown_RedHatLinux.mak \ power_macintosh_Darwin.mak \ unknown_unknown.mak \ x86_64_Darwin.mak \ FILES = Makefile default.mak $(SYSTEMS) include $(TOP)/config/common_make_rules festival/config/systems/default.mak0000644000175000017500000001253307304237411016076 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Very common settings to avoid repetition. ## ## ## ########################################################################### ########################################################################### ## Installation directories INSTALL_PREFIX=/usr/local BINDIR=$(INSTALL_PREFIX)/bin LIBDIR=$(INSTALL_PREFIX)/lib INCDIR=$(INSTALL_PREFIX)/include MANDIR=$(INSTALL_PREFIX)/man ########################################################################### ## Where the central RCS masters are stored. ## ## Used for development at CSTR, you can probably ignore it. LOCAL_REPOSITORY = ########################################################################### ## Where to find Network Audio NAS_INCLUDE = /usr/X11R6/include NAS_LIB = /usr/X11R6/lib ########################################################################### ## Where to find Enlightenment Speech Demon ESD_INCLUDE = /usr/local/include ESD_LIB = /usr/local/lib ########################################################################### ## Where to find X11 X11_INCLUDE = /usr/X11R6/include X11_LIB = /usr/X11R6/lib ########################################################################### ## TCL support TCL_INCLUDE = /usr/local/include TCL_LIB = /usr/local/lib TCL_LIBRARY = -ltcl7.6 ########################################################################### ## Efence library for malloc debugging EFENCE_LIB = /usr/local/lib ########################################################################### ## Commands. ## Must support -nt GNUTEST = gnutest ## INSTALL_PROG = install ## Used to index libraries RANLIB = ranlib ## echo without a newline ECHO_N = echo -n ## make depend for when we haven't specified a compiler MAKE_DEPEND = makedepend $(INCLUDES) $(TEMPLATES) $(TEMPLATE_SPECIFIC) ## Generic library building BUILD_LIB =$(AR) cruv ## generic library indexing INDEX_LIB = $(RANLIB) ## shrink executables STRIP = strip ## Useful sloth DO_NOTHING = true DO_NOTHING_ARGS = : ## different types of awk. For our purposes gawk can be used for nawk AWK = awk NAWK = nawk ## Perl. Not used in build, but we have some perl scripts. PERL=/usr/bin/perl ## Just in case someone has a broken test TEST = test ## Must understand -nt GNUTEST = gnutest ## Avoid clever RMs people may have on their path RM = /bin/rm ########################################################################### ## Arguments for DOC++ for creating documentation DOCXX = doc++_sane DOCXX_ARGS = -a -f -B banner.inc -M sane -D 'SYSTEM "$(EST_HOME)/doc/sane.dtd"' COMPILER_VERSION_COMMAND=true JAVA_COMPILER_VERSION_COMMAND=true JAVA_SYSTEM_INCLUDES = -I$(JAVA_HOME)/include/genunix festival/config/systems/Linux.mak0000644000175000017500000000652007304237411015550 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Thu Oct 2 1997 ## ## -------------------------------------------------------------------- ## ## Settings for Linux. ## ## ## ########################################################################### include $(EST)/config/systems/default.mak DEFAULT_JAVA_HOME=/usr/lib/jdk-1.1.6 JAVA=/usr/bin/java JAVAC=/usr/bin/javac JAVAH=/usr/bin/javah TCL_LIBRARY = -ltcl OS_LIBS = -ldl ## the native audio module for this type of system NATIVE_AUDIO_MODULE = LINUX16 ## echo -n doesn't work ECHO_N = /usr/bin/printf "%s" GNUTEST=test ## awk is gawk, so it does all we could desire and then more. NAWK=awk festival/config/systems/RedHatLinux.mak0000644000175000017500000000556207304237411016645 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Red Hat Linux distributions. ## ## ## ########################################################################### festival/config/systems/DebianGNULinux.mak0000644000175000017500000000564607304237411017235 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: David Huggins-Daines ## ## -------------------------------------------------------------------- ## ## Settings for Debian GNU/Linux distributions. ## ## ## ########################################################################### # Debian does not use termcap OS_LIBS = -ldl -lncurses festival/config/systems/alpha_Linux.mak0000644000175000017500000000565507304237411016725 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Red Hat Linux distributions. ## ## ## ########################################################################### include $(EST)/config/systems/Linux.mak CFLAGS += -mieee festival/config/systems/alpha_OSF1V4.0.mak0000644000175000017500000000606507304237411016702 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## Date: Sat Oct 18 1997 ## ## -------------------------------------------------------------------- ## ## Settings for DEC Alpha OSF1 V4.0 ## ## ## ########################################################################### include $(EST)/config/systems/default.mak ## echo -n doesn't work (well only sometimes ?) ECHO_N = /bin/printf "%s" festival/config/systems/alpha_RedHatLinux.mak0000644000175000017500000000572707304237411020015 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Red Hat Linux on Alpha. ## ## ## ########################################################################### include $(EST)/config/systems/alpha_Linux.mak ifndef GCC GCC=egcs endif EGCS_CC=gcc EGCS_CXX=g++ festival/config/systems/hp9000_HP-UX.mak0000644000175000017500000000651707304237411016320 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## HP 9000 series. ## ## ## ########################################################################### include $(EST)/config/systems/default.mak ## Just guesses for what people are likely to have GCC=gcc27 ## Libraries needed for sockets based programs. OS_LIBS = -lsocket -lnsl ## the native audio module for this type of system NATIVE_AUDIO_MODULE = ## Official location for java DEFAULT_JAVA_HOME=/usr/java1.1 GCC_SYSTEM_OPTIONS = ## specific java files JAVA_SYSTEM_INCLUDES = -I$(JAVA_HOME)/include/SOMETHING_HP_HERE ## echo -n doesn't work ECHO_N = /bin/printf "%s" festival/config/systems/hp9000_HP-UXB.10.mak0000644000175000017500000000563707304237411016643 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Specifics for this version of HP-UX. ## ## ## ########################################################################### include $(EST)/config/systems/hp9000_HP-UX.mak festival/config/systems/ip_IRIX.mak0000644000175000017500000000634107304237411015715 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh,UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black ## ## Date: Nov 1997 ## ########################################################################### ## Settings for Irix ## ## ## ########################################################################### include $(EST)/config/systems/default.mak ## the native audio module for this type of system NATIVE_AUDIO_MODULE = IRIX ## echo -n doesn't work (well only sometimes ?) ECHO_N = /bin/printf "%s" ## Doesn't have or need RANLIB RANLIB = true ## IRIX specific java include files. JAVA_SYSTEM_INCLUDES = -I$(JAVA_HOME)/include/irix festival/config/systems/ip_IRIX5.3.mak0000644000175000017500000000602207304237411016137 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh,UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black ## ## Date: Nov 1997 ## ########################################################################### ## Settings for Irix 5.3. ## ## ## ########################################################################### include $(EST)/config/systems/ip_IRIX.mak ## the native audio module for this type of system NATIVE_AUDIO_MODULE = IRIX53 festival/config/systems/ip_IRIX6.3.mak0000644000175000017500000000570207304237411016144 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh,UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black ## ## Date: Nov 1997 ## ########################################################################### ## Settings for Irix 6.3. ## ## ## ########################################################################### include $(EST)/config/systems/ip_IRIX.mak festival/config/systems/ip_IRIX6.4.mak0000644000175000017500000000570207304237411016145 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh,UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black ## ## Date: Nov 1997 ## ########################################################################### ## Settings for Irix 6.4. ## ## ## ########################################################################### include $(EST)/config/systems/ip_IRIX.mak festival/config/systems/ip_IRIX646.4.mak0000644000175000017500000000570207304237411016317 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh,UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black ## ## Date: Nov 1997 ## ########################################################################### ## Settings for Irix 6.4. ## ## ## ########################################################################### include $(EST)/config/systems/ip_IRIX.mak festival/config/systems/ix86_CYGWIN1.0.mak0000644000175000017500000000571407304237411016552 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Cygnus gnuwin1.0 ## ## ########################################################################### include $(EST)/config/systems/ix86_CYGWIN32.mak ## Cygwin version of egcs has optimisatin problems with some files. HONOUR_NOOPT=1 festival/config/systems/ix86_CYGWIN1.1.mak0000644000175000017500000000571407304237411016553 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Cygnus gnuwin1.1 ## ## ########################################################################### include $(EST)/config/systems/ix86_CYGWIN32.mak ## Cygwin version of egcs has optimisatin problems with some files. HONOUR_NOOPT=1 festival/config/systems/ix86_CYGWIN1.3.mak0000644000175000017500000000571407327551242016563 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Cygnus gnuwin1.3 ## ## ########################################################################### include $(EST)/config/systems/ix86_CYGWIN32.mak ## Cygwin version of egcs has optimisatin problems with some files. HONOUR_NOOPT=1 festival/config/systems/ix86_CYGWIN1.4.mak0000644000175000017500000000564310006206360016547 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Cygnus gnuwin32 v 4.0 (b19). ## ## ## ########################################################################### include $(EST)/config/systems/ix86_CYGWIN32.mak festival/config/systems/ix86_CYGWIN1.5.mak0000644000175000017500000000564310006206360016550 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Cygnus gnuwin32 v 4.0 (b19). ## ## ## ########################################################################### include $(EST)/config/systems/ix86_CYGWIN32.mak festival/config/systems/ix86_CYGWIN1.7.mak0000644000175000017500000000564311363322453016563 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Cygnus gnuwin32 v 4.0 (b19). ## ## ## ########################################################################### include $(EST)/config/systems/ix86_CYGWIN32.mak festival/config/systems/ix86_CYGWIN20.1.mak0000644000175000017500000000576707304237411016644 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Cygnus gnuwin32 v 4.0 (b19). ## ## ## ########################################################################### include $(EST)/config/systems/ix86_CYGWIN32.mak ## Cygwin version of egcs has optimisatin problems with some files. HONOUR_NOOPT=1 festival/config/systems/ix86_CYGWIN32.mak0000644000175000017500000000630607304237411016476 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Cygwin32. ## ## ## ########################################################################### include $(EST)/config/systems/default.mak DEFAULT_JAVA_HOME=/usr/lib/jdk-1.1.6 gcc=gcc27 ## the native audio module for this type of system NATIVE_AUDIO_MODULE = WIN32 GNUTEST=test RM=rm ## awk is gawk, so it does all we could desire and then more. NAWK=awk ## EGCS installs as gcc EGCS_CC=gcc EGCS_CXX=gcc OS_LIBS = -lwinmm -luser32 festival/config/systems/ix86_CYGWIN324.0.mak0000644000175000017500000000564307304237411016723 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Cygnus gnuwin32 v 4.0 (b19). ## ## ## ########################################################################### include $(EST)/config/systems/ix86_CYGWIN32.mak festival/config/systems/ix86_Darwin.mak0000644000175000017500000000575211363322453016562 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## Date: Aug 3 2006 ## ## -------------------------------------------------------------------- ## ## Settings for Apple Darwin. ## ## Thanks to Brian West ## ########################################################################### include $(EST)/config/systems/ix86_Darwin.mak festival/config/systems/ix86_FreeBSD.mak0000644000175000017500000000654307304237411016546 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## Date: Fri Oct 3 1997 ## ## -------------------------------------------------------------------- ## ## Settings for FreeBSD. ## ## ## ########################################################################### include $(EST)/config/systems/default.mak ## the native audio module for this type of system NATIVE_AUDIO_MODULE = FREEBSD16 ## echo -n doesn't work (well only sometimes ?) ECHO_N = /usr/bin/printf "%s" NAWK=awk # GCC_MAKE_SHARED_LIB = ld -Bshareable -x -o XXX DEFAULT_JAVA_HOME=/usr/local/jdk JAVA=$(JAVA_HOME)/bin/java JAVAC=$(JAVA_HOME)/bin/javac JAVAH=$(JAVA_HOME)/bin/javah -jni JAR=$(JAVA_HOME)/bin/jar cf0v festival/config/systems/ix86_FreeBSD2.1.mak0000644000175000017500000000577207304237411016772 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## Date: Sat Oct 11 1997 ## ## -------------------------------------------------------------------- ## ## Settings for FreeBSD 2.2 ## ## ## ########################################################################### include $(EST)/config/systems/ix86_FreeBSD.mak GCC=gcc26 festival/config/systems/ix86_FreeBSD2.2.mak0000644000175000017500000000577207304237411016773 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## Date: Sat Oct 11 1997 ## ## -------------------------------------------------------------------- ## ## Settings for FreeBSD 2.2 ## ## ## ########################################################################### include $(EST)/config/systems/ix86_FreeBSD.mak GCC=gcc27 festival/config/systems/ix86_FreeBSD3.0.mak0000644000175000017500000000631207304237411016761 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## Date: Wed Mar 10 1999 ## ## -------------------------------------------------------------------- ## ## Settings for FreeBSD 3.0 ## ## ## ########################################################################### include $(EST)/config/systems/ix86_FreeBSD.mak GCC=gcc27 JAVA_SYSTEM_INCLUDES = -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/freebsd GCC=gcc27 JAVA=$(JAVA_HOME)/bin/java JAVAC=$(JAVA_HOME)/bin/javac JAVAH=$(JAVA_HOME)/bin/javah -jni JAR=$(JAVA_HOME)/bin/jar cf0v festival/config/systems/ix86_FreeBSD3.1.mak0000644000175000017500000000645407304237411016771 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## Date: Wed Mar 10 1999 ## ## -------------------------------------------------------------------- ## ## Settings for FreeBSD 3.1 ## ## ## ########################################################################### # They went to ELF at this point, but everything seems fine as is include $(EST)/config/systems/ix86_FreeBSD.mak JAVA_SYSTEM_INCLUDES = -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/freebsd GCC=gcc27 JAVA=$(JAVA_HOME)/bin/java JAVAC=$(JAVA_HOME)/bin/javac JAVAH=$(JAVA_HOME)/bin/javah -jni JAR=$(JAVA_HOME)/bin/jar cf0v GCC_MAKE_SHARED_LIB = gcc -shared -o XXX festival/config/systems/ix86_FreeBSD3.2.mak0000644000175000017500000000575007304237411016770 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for FreeBSD 3.2 ## ## ## ########################################################################### # They went to ELF at this point, but everything seems fine as is include $(EST)/config/systems/ix86_FreeBSD3.1.mak festival/config/systems/ix86_FreeBSD3.3.mak0000644000175000017500000000575007304237411016771 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for FreeBSD 3.3 ## ## ## ########################################################################### # They went to ELF at this point, but everything seems fine as is include $(EST)/config/systems/ix86_FreeBSD3.1.mak festival/config/systems/ix86_FreeBSD4.0.mak0000644000175000017500000000607507304237411016770 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for FreeBSD 4.0 (not actually checked yet) ## ## ## ########################################################################### # They went to ELF at this point, but everything seems fine as is include $(EST)/config/systems/ix86_FreeBSD.mak JAVA_SYSTEM_INCLUDES = -I$(JAVA_HOME)/include -I$(JAVA_HOME)/include/freebsd GCC=gcc27 festival/config/systems/ix86_OS22.mak0000644000175000017500000000672707304237411016025 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Samuel Audet ## ## Date: Sat Aug 1 1998 ## ## -------------------------------------------------------------------- ## ## Settings for OS/2 Warp 3 and 4 using EMX/GCC ## ## ## ########################################################################### include $(TOP)/config/systems/default.mak ## custom settings needed GCC=gcc27emx ## Libraries needed for sockets based programs. OS_LIBS = -lsocket ## uses the path RM = rm ## the native audio module for this type of system NATIVE_AUDIO_MODULE = OS2 ## Must support -nt GNUTEST = test ## echo -n doesn't work (well only sometimes ?) ECHO_N = echo -n ## awk is gawk, so it does all we could desire and then more. NAWK=awk ## Used to index libraries RANLIB = ar s  festival/config/systems/ix86_RedHatLinux4.0.mak0000644000175000017500000000572307304237411017744 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Red Hat Linux 4.0. ## ## ## ########################################################################### include $(EST)/config/systems/Linux.mak include $(EST)/config/systems/RedHatLinux.mak GCC=gcc27 festival/config/systems/ix86_RedHatLinux4.1.mak0000644000175000017500000000572307304237411017745 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Red Hat Linux 4.1. ## ## ## ########################################################################### include $(EST)/config/systems/Linux.mak include $(EST)/config/systems/RedHatLinux.mak GCC=gcc27 festival/config/systems/ix86_RedHatLinux4.2.mak0000644000175000017500000000572307304237411017746 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Red Hat Linux 4.1. ## ## ## ########################################################################### include $(EST)/config/systems/Linux.mak include $(EST)/config/systems/RedHatLinux.mak GCC=gcc27 festival/config/systems/ix86_RedHatLinux5.0.mak0000644000175000017500000000572307304237411017745 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Red Hat Linux 5.1. ## ## ## ########################################################################### include $(EST)/config/systems/Linux.mak include $(EST)/config/systems/RedHatLinux.mak GCC=gcc27 festival/config/systems/ix86_RedHatLinux5.1.mak0000644000175000017500000000577107304237411017751 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Red Hat Linux 5.1. ## ## ## ########################################################################### include $(EST)/config/systems/Linux.mak include $(EST)/config/systems/RedHatLinux.mak ifndef GCC GCC=egcs endif EGCS_CC=gcc EGCS_CXX=g++ festival/config/systems/ix86_RedHatLinux5.2.mak0000644000175000017500000000610507304237411017742 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Red Hat Linux 5.1. ## ## ## ########################################################################### include $(EST)/config/systems/ix86_RedHatLinux5.1.mak DEFAULT_JAVA_HOME=/usr/local/jdk-1.1.6 JAVA=$(JAVA_HOME)/bin/java JAVAC=$(JAVA_HOME)/bin/javac JAVAH=$(JAVA_HOME)/bin/javah -jni JAR=$(JAVA_HOME)/bin/jar cf0v festival/config/systems/ix86_RedHatLinux6.0.mak0000644000175000017500000000623007304237411017740 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Red Hat Linux 6.0. ## ## ## ########################################################################### include $(EST)/config/systems/Linux.mak include $(EST)/config/systems/RedHatLinux.mak ifndef GCC GCC=egcs endif EGCS_CC=gcc EGCS_CXX=g++ DEFAULT_JAVA_HOME=/usr/local/jdk1.1.6 JAVA=$(JAVA_HOME)/bin/java JAVAC=$(JAVA_HOME)/bin/javac JAVAH=$(JAVA_HOME)/bin/javah -jni JAR=$(JAVA_HOME)/bin/jar cf0v festival/config/systems/ix86_RedHatLinux6.1.mak0000644000175000017500000000623107304237411017742 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Red Hat Linux 6.1. (guess) ## ## ## ########################################################################### include $(EST)/config/systems/Linux.mak include $(EST)/config/systems/RedHatLinux.mak ifndef GCC GCC=egcs endif EGCS_CC=gcc EGCS_CXX=g++ DEFAULT_JAVA_HOME=/usr/local/jdk1.1.6 JAVA=$(JAVA_HOME)/bin/java JAVAC=$(JAVA_HOME)/bin/javac JAVAH=$(JAVA_HOME)/bin/javah -jni JAR=$(JAVA_HOME)/bin/jar cf0v festival/config/systems/ix86_RedHatLinux6.2.mak0000644000175000017500000000564607304237411017754 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Red Hat Linux 6.2. (guess) ## ## ## ########################################################################### include $(EST)/config/systems/ix86_RedHatLinux6.1.mak festival/config/systems/ix86_RedHatLinux7.0.mak0000644000175000017500000000566207304237411017751 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for Red Hat Linux 7.0. (guess) ## ## ## ########################################################################### include $(EST)/config/systems/ix86_RedHatLinux6.1.mak GCC=gcc296 festival/config/systems/ix86_SunOS5.5.mak0000644000175000017500000000564007304237411016570 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Settings for i386 SunOS 5.5 ## ## ## ## Like sparc except no mv8 ## ## ## ########################################################################### include $(EST)/config/systems/ix86_SunOS5.mak festival/config/systems/ix86_SunOS5.6.mak0000644000175000017500000000564007304237411016571 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Settings for i386 SunOS 5.6 ## ## ## ## Like sparc but no mv8 ## ## ## ########################################################################### include $(EST)/config/systems/ix86_SunOS5.mak festival/config/systems/ix86_SunOS5.7.mak0000644000175000017500000000564007304237411016572 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Settings for i386 SunOS 5.7 ## ## ## ## Like sparc but no mv8 ## ## ## ########################################################################### include $(EST)/config/systems/ix86_SunOS5.mak festival/config/systems/ix86_SunOS5.8.mak0000644000175000017500000000563607624146540016607 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for SunOS 5.8 ## ## ## ########################################################################### include $(EST)/config/systems/ix86_SunOS5.mak festival/config/systems/ix86_SunOS5.mak0000644000175000017500000000644607304237411016432 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Thu Oct 2 1997 ## ## -------------------------------------------------------------------- ## ## Settings for SunOS 5 on Intel platform. ## ## ## ########################################################################### include $(EST)/config/systems/default.mak DEFAULT_JAVA_HOME=/usr/java1.1 ## Libraries needed for sockets based programs. OS_LIBS = -lsocket -lnsl ## the native audio module for this type of system NATIVE_AUDIO_MODULE = SUN16 ## echo -n doesn't work (well only sometimes ?) ECHO_N = /bin/printf "%s" SYSTEM_JAVA_INCLUDES = -I$(JAVA_HOME)/include/solaris festival/config/systems/rs6000_AIX4.1.mak0000644000175000017500000000566407304237411016337 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1999 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cs.cmu.edu) ## ## -------------------------------------------------------------------- ## ## For AIX 4.1 (from Stan Chen) ## ## ## ########################################################################### include $(EST)/config/systems/default.mak LINKFLAGS += -Wl,-bbigtoc festival/config/systems/sparc_SunOS4.1.mak0000644000175000017500000000575607304237411017105 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Thu Oct 2 1997 ## ## -------------------------------------------------------------------- ## ## Settings for SunOS 4.1 ## ## ## ########################################################################### include $(EST)/config/systems/sparc_SunOS4.mak festival/config/systems/sparc_SunOS4.mak0000644000175000017500000000607007304237411016734 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Thu Oct 2 1997 ## ## -------------------------------------------------------------------- ## ## Settings for SunOS 4. ## ## ## ########################################################################### include $(EST)/config/systems/default.mak ## the native audio module for this type of system NATIVE_AUDIO_MODULE = SUN16 festival/config/systems/sparc_SunOS5.5.mak0000644000175000017500000000602107306754504017105 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Thu Oct 2 1997 ## ## -------------------------------------------------------------------- ## ## Settings for SunOS 5.5 ## ## ## ########################################################################### CC_OTHER_FLAGS += -Dolder_solaris include $(EST)/config/systems/sparc_SunOS5.mak festival/config/systems/sparc_SunOS5.6.mak0000644000175000017500000000601707306754504017113 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## Date: Tue Nov 4 1997 ## ## -------------------------------------------------------------------- ## ## Settings for SunOS 5.6 ## ## ## ########################################################################### CC_OTHER_FLAGS += -Dolder_solaris include $(EST)/config/systems/sparc_SunOS5.mak festival/config/systems/sparc_SunOS5.7.mak0000644000175000017500000000563707304237411017112 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for SunOS 5.7 ## ## ## ########################################################################### include $(EST)/config/systems/sparc_SunOS5.mak festival/config/systems/sparc_SunOS5.8.mak0000644000175000017500000000563707304237411017113 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for SunOS 5.8 ## ## ## ########################################################################### include $(EST)/config/systems/sparc_SunOS5.mak festival/config/systems/sparc_SunOS5.mak0000644000175000017500000000676207304237411016745 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Thu Oct 2 1997 ## ## -------------------------------------------------------------------- ## ## Settings for SunOS 5.* (aka Solaris 2, aka 'Solaris') ## ## ## ########################################################################### include $(EST)/config/systems/default.mak ## Just guesses for what people are likely to have GCC=gcc27 SUNCC=suncc40 ## Libraries needed for sockets based programs. OS_LIBS = -lsocket -lnsl ## the native audio module for this type of system NATIVE_AUDIO_MODULE = SUN16 ## Official location for java DEFAULT_JAVA_HOME=/usr/java1.1 ## Tell gcc we are a v8 sparc or better. Any legacy machines lose. GCC_SYSTEM_OPTIONS = -mv8 ## echo -n doesn't work ECHO_N = /bin/printf "%s" JAVA_SYSTEM_INCLUDES = -I$(JAVA_HOME)/include/solaris ## Force use of nawk AWK=nawk festival/config/systems/unknown_DebianGNULinux.mak0000644000175000017500000000571007304237411021004 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: David Huggins-Daines ## ## -------------------------------------------------------------------- ## ## Settings for generic/unknown Debian GNU/Linux systems ## ## ## ########################################################################### include $(EST)/config/systems/Linux.mak include $(EST)/config/systems/DebianGNULinux.mak festival/config/systems/unknown_Linux.mak0000644000175000017500000000562707304237411017336 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Linux v2. ## ## ## ########################################################################### include $(EST)/config/systems/Linux.mak festival/config/systems/unknown_RedHatLinux.mak0000644000175000017500000000570607304237411020424 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Robert Clark (robert@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Settings for unknown Red Hat Linux versions ## ## ## ########################################################################### include $(EST)/config/systems/Linux.mak include $(EST)/config/systems/RedHatLinux.mak festival/config/systems/power_macintosh_Darwin.mak0000644000175000017500000000631710077563027021170 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## Date: Fri Oct 3 1997 ## ## -------------------------------------------------------------------- ## ## Settings for Apple Darwin. ## ## ## ########################################################################### include $(EST)/config/systems/default.mak ## the native audio module for this type of system NATIVE_AUDIO_MODULE = MACOSX ## echo -n doesn't work (well only sometimes ?) ECHO_N = /usr/bin/printf "%s" NAWK=awk GCC295=cc # GCC_MAKE_SHARED_LIB = ld -Bshareable -x -o XXX festival/config/systems/unknown_unknown.mak0000644000175000017500000000612307304237411017726 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Tue Oct 7 1997 ## ## -------------------------------------------------------------------- ## ## A default description used to bootstrap. ## ## ## ########################################################################### include $(EST)/config/systems/default.mak ## Must support -nt GNUTEST = gnutest ## echo -n doesn't work (well only sometimes ?) ECHO_N = echo festival/config/systems/x86_64_Darwin.mak0000644000175000017500000000575411363322454016725 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black (awb@cstr.ed.ac.uk) ## ## Date: Aug 3 2006 ## ## -------------------------------------------------------------------- ## ## Settings for Apple Darwin. ## ## Thanks to Brian West ## ########################################################################### include $(EST)/config/systems/x86_64_Darwin.mak festival/src/Makefile0000644000175000017500000000602407262605650013240 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Makefile for main src directory ## ## ## ########################################################################### TOP=.. DIRNAME=src LIB_BUILD_DIRS = arch modules BUILD_DIRS = $(LIB_BUILD_DIRS) main ALL_DIRS = include $(BUILD_DIRS) lib scripts FILES=Makefile ALL = .sub_directories include $(TOP)/config/common_make_rules tags: @ rm -f $(TOP)/FileList @ $(MAKE) --no-print-directory file-list @ etags `grep "src/.*[ch]c*$$" $(TOP)/FileList | sed 's/src\///'` festival/src/include/Makefile0000644000175000017500000000513307262605650014663 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../.. DIRNAME=src/include H = festival.h \ Phone.h intonation.h lexicon.h \ text.h fngram.h modules.h ModuleDescription.h \ module_support.h FILES=Makefile $(H) include $(TOP)/config/common_make_rules festival/src/include/festival.h0000644000175000017500000001606712431672620015214 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Top level .h file: main public functions */ /*=======================================================================*/ #ifndef __FESTIVAL_H__ #define __FESTIVAL_H__ #include #include using namespace std; #include "EST.h" #include "EST_cutils.h" #include "siod.h" #include "Phone.h" #ifndef streq #define streq(X,Y) (strcmp(X,Y)==0) #endif struct ModuleDescription; /* An iostream for outputing debug messages, switchable */ /* to /dev/null or cerr */ extern ostream *cdebug; #define cwarn cout extern "C" FILE* stddebug; extern int ft_server_socket; extern const char *festival_version; /* For server/client */ #define FESTIVAL_DEFAULT_PORT 1314 int festival_socket_client(const char *host,int port); int festival_start_server(int port); void festival_initialize(int load_init_files,int heap_size); void festival_init_lang(const EST_String &language); int festival_eval_command(const EST_String &expr); int festival_load_file(const EST_String &fname); int festival_say_file(const EST_String &fname); int festival_say_text(const EST_String &text); int festival_text_to_wave(const EST_String &text,EST_Wave &wave); void festival_repl(int interactive); void festival_server_mode(void); void festival_wait_for_spooler(void); void festival_tidy_up(); /* Never used and conflicts with some external system */ /* typedef void (*FT_Module)(EST_Utterance &utt); */ /* Feature functions */ void festival_def_nff(const EST_String &name,const EST_String &sname, EST_Item_featfunc func,const char *doc); typedef EST_Val (*FT_ff_pref_func)(EST_Item *s,const EST_String &name); void festival_def_ff_pref(const EST_String &pref,const EST_String &sname, FT_ff_pref_func func, const char *doc); EST_Val ffeature(EST_Item *s, const EST_String &name); /* proclaim a new module option Copyright to add to startup banner description is a computer readable description of the module */ void proclaim_module(const EST_String &name, const EST_String &banner_copyright, const ModuleDescription *description = NULL); void proclaim_module(const EST_String &name, const ModuleDescription *description = NULL); void init_module_subr(const char *name, LISP (*fcn)(LISP), const ModuleDescription *description); /* Some basic functions for accessing structures created by */ /* various modelling techniques */ EST_Val wagon_predict(EST_Item *s, LISP tree); LISP wagon_pd(EST_Item *s, LISP tree); EST_Val lr_predict(EST_Item *s, LISP lr_model); /* Grammar access functions */ EST_Ngrammar *get_ngram(const EST_String &name, const EST_String &filename = EST_String::Empty); EST_WFST *get_wfst(const EST_String &name, const EST_String &filename = EST_String::Empty); LISP lisp_wfst_transduce(LISP wfstname, LISP input); EST_String map_pos(LISP posmap, const EST_String &pos); LISP map_pos(LISP posmap, LISP pos); /* On error do a longjmp to appropriate place */ /* This is done as a macro so the compiler can tell its non-returnable */ #define festival_error() (errjmp_ok ? longjmp(*est_errjmp,1) : festival_tidy_up(),exit(-1)) /* Add new (utterance) module */ void festival_def_utt_module(const char *name, LISP (*fcn)(LISP), const char *docstring); void utt_cleanup(EST_Utterance &u); // delete all relations const EST_String utt_iform_string(EST_Utterance &utt); LISP utt_iform(EST_Utterance &utt); const EST_String utt_type(EST_Utterance &utt); void add_item_features(EST_Item *s,LISP features); extern const char *festival_libdir; // Module specific LISP/etc definitions void festival_init_modules(void); // Some general functions LISP ft_get_param(const EST_String &pname); // SIOD user defined types used by festival #define tc_festival_dummyobject tc_application_1 #define tc_festival_unit tc_application_2 #define tc_festival_unitdatabase tc_application_3 #define tc_festival_unitindex tc_application_4 #define tc_festival_join tc_application_5 #define tc_festival_schememoduledescription tc_application_6 #define tc_festival_unitcatalogue tc_application_7 // used to recognise our types #define tc_festival_first_type tc_festival_dummyobject #define tc_festival_last_type tc_festival_schememoduledescription #define is_festival_type(X) ((X) >= tc_festival_first_type && (X) <= tc_festival_last_type) class UnitDatabase *get_c_unitdatabase(LISP x); #define FESTIVAL_HEAP_SIZE 10000000 #endif festival/src/include/Phone.h0000644000175000017500000001425707262605650014454 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* Phone and PhoneSet class header file */ /* */ /*=======================================================================*/ #ifndef __PHONE_H__ #define __PHONE_H__ class Phone{ private: EST_String name; EST_StrStr_KVL features; public: Phone() {name = "";} EST_String &phone_name() {return name;} void set_phone_name(const EST_String &p) {name = p;} void add_feat(const EST_String &f, const EST_String &v) { features.add_item(f,v); } const EST_String &val(const EST_String &key) const { return features.val_def(key,"");} const EST_String &val(const EST_String &key,const EST_String &def) { return features.val_def(key,def); } int match_features(Phone *foreign); inline friend ostream& operator<<(ostream& s, Phone &p); Phone & operator =(const Phone &a); }; inline ostream& operator<<(ostream& s, Phone &p) { s << "[PHONE " << p.phone_name() << "]"; // s << p.features << endl; return s; } inline Phone &Phone::operator = (const Phone &a) { name = a.name; features = a.features; return *this; } class PhoneSet{ private: EST_String psetname; LISP silences; LISP map; LISP feature_defs; // List of features and values LISP phones; public: PhoneSet() {psetname = ""; phones=feature_defs=map=silences=NIL; gc_protect(&silences); gc_protect(&map); gc_protect(&feature_defs); gc_protect(&phones);} ~PhoneSet(); const EST_String &phone_set_name() const {return psetname;} void set_phone_set_name(const EST_String &p) {psetname = p;} int present(const EST_String &phone) const {return (siod_assoc_str(phone,phones) != NIL);} int is_silence(const EST_String &ph) const; void set_silences(LISP sils); void set_map(LISP m); LISP get_silences(void) {return silences;} LISP get_phones(void) {return phones;} LISP get_feature_defs(void) {return reverse(feature_defs);} int num_phones(void) const {return siod_llength(phones);} Phone *member(const EST_String &phone) const; int phnum(const char *phone) const; const char *phnum(const int n) const; int add_phone(Phone *phone); int feat_val(const EST_String &feat, const EST_String &val) { return (siod_member_str(val, car(cdr(siod_assoc_str(feat,feature_defs)))) != NIL); } void set_feature(const EST_String &name, LISP vals); inline friend ostream& operator<<(ostream& s, PhoneSet &p); Phone *find_matched_phone(Phone *phone); PhoneSet & operator =(const PhoneSet &a); }; inline ostream& operator<<(ostream& s, PhoneSet &p) { s << p.phone_set_name(); return s; } const EST_String &map_phone(const EST_String &fromphonename, const EST_String &fromsetname, const EST_String &tosetname); const EST_String &ph_feat(const EST_String &ph,const EST_String &feat); int ph_is_silence(const EST_String &ph); int ph_is_vowel(const EST_String &ph); int ph_is_consonant(const EST_String &ph); int ph_is_liquid(const EST_String &ph); int ph_is_approximant(const EST_String &ph); int ph_is_stop(const EST_String &ph); int ph_is_nasal(const EST_String &ph); int ph_is_fricative(const EST_String &ph); int ph_is_sonorant(const EST_String &ph); int ph_is_obstruent(const EST_String &ph); int ph_is_voiced(const EST_String &ph); int ph_is_sonorant(const EST_String &ph); int ph_is_syllabic(const EST_String &ph); int ph_sonority(const EST_String &ph); EST_String ph_silence(void); PhoneSet *phoneset_name_to_set(const EST_String &name); #endif festival/src/include/intonation.h0000644000175000017500000000622307262605650015557 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Shared intonation utilies */ /* */ /*=======================================================================*/ #ifndef __INTONATION_H__ #define __INTONATION_H__ EST_Item *add_target(EST_Utterance *u, EST_Item *seg, float pos,float val); EST_Item *add_IntEvent(EST_Utterance *u, EST_Item *syl, const EST_String &label); #endif /* __INTONATION_H__ */ festival/src/include/lexicon.h0000644000175000017500000001361407262605650015040 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Shared lexicon utilities */ /* Top level form (simply an s-expression of the input form */ /* */ /*=======================================================================*/ #ifndef __LEXICON_H__ #define __LEXICON_H__ #include "EST_Pathname.h" enum lex_type_t {lex_external, lex_internal}; class Lexicon{ private: lex_type_t type; EST_String name; EST_String ps_name; LISP addenda; // for personal local changes LISP posmap; int comp_num_entries; EST_Pathname bl_filename; FILE *binlexfp; EST_String lts_method; EST_String lts_ruleset; int blstart; LISP index_cache; void binlex_init(void); LISP lookup_addenda(const EST_String &word, LISP features); LISP lookup_complex(const EST_String &word, LISP features); LISP lookup_lts(const EST_String &word, LISP features); LISP bl_bsearch(const EST_String &word,LISP features, int start,int end,int depth); LISP bl_find_next_entry(int pos); LISP bl_find_actual_entry(int pos,const EST_String &word,LISP features); int lex_entry_match; LISP matched_lexical_entries; public: LISP pre_hooks; LISP post_hooks; Lexicon(); ~Lexicon(); const EST_String &lex_name() const {return name;} const EST_String &phoneset_name() const {return ps_name;} void set_lex_name(const EST_String &p) {name = p;} const EST_String &get_lex_name(void) const { return name; } void set_phoneset_name(const EST_String &p) {ps_name = p;} void set_lts_method(const EST_String &p) {lts_method = p;} void set_lts_ruleset(const EST_String &p) {lts_ruleset = p;} void set_pos_map(LISP p) {posmap = p;} const EST_String &get_lts_ruleset(void) const { return lts_ruleset; } void set_bl_filename(const EST_String &p) {bl_filename = p; if (binlexfp != NULL) fclose(binlexfp); binlexfp=NULL;} void add_addenda(LISP entry) {addenda = cons(entry,addenda);} LISP lookup(const EST_String &word,const LISP features); LISP lookup_all(const EST_String &word); EST_String str_lookup(const EST_String &word,const LISP features); int in_lexicon(const EST_String &word,LISP features); int num_matches() { return lex_entry_match; } void bl_lookup_cache(LISP cache, const EST_String &word, int &start, int &end, int &depth); void add_to_cache(LISP index_cache, const EST_String &word, int start,int mid, int end); inline friend ostream& operator<<(ostream& s, Lexicon &p); Lexicon & operator =(const Lexicon &a); }; inline ostream& operator<<(ostream& s, Lexicon &p) { s << "[LEXICON " << p.lex_name() << "]" ; return s; } inline Lexicon &Lexicon::operator = (const Lexicon &a) { name = a.name; addenda = a.addenda; bl_filename = a.bl_filename; binlexfp = NULL; lts_method = a.lts_method; return *this; } LISP lex_lookup_word(const EST_String &word,LISP features); EST_String lex_current_phoneset(void); LISP lex_select_lex(LISP lexname); LISP lex_syllabify(LISP phones); LISP lex_syllabify_phstress(LISP phones); int in_current_lexicon(const EST_String &word,LISP features); #endif /* __LEXICON_H__ */ festival/src/include/text.h0000644000175000017500000000721207262605650014360 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Shared text utilities */ /* */ /*=======================================================================*/ #ifndef __TEXT_H__ #define __TEXT_H__ EST_Item *add_token(EST_Utterance *u,EST_Token &t); void festival_token_init(void); LISP extract_tokens(LISP file, LISP tokens,LISP ofile); LISP new_token_utt(void); void tts_file_xxml(LISP filename); void tts_file_raw(LISP filename); LISP xxml_call_element_function(const EST_String &element, LISP atts, LISP elements, LISP utt); LISP xxml_get_tokens(const EST_String &line,LISP feats,LISP utt); typedef void (*TTS_app_tok)(EST_Item *token); typedef void (*TTS_app_utt)(LISP utt); LISP tts_chunk_stream(EST_TokenStream &ts, TTS_app_tok app_tok, TTS_app_utt app_utt, LISP eou_tree, LISP utt); void tts_file_user_mode(LISP filename, LISP params); #endif /* __TEXT_H__ */ festival/src/include/fngram.h0000644000175000017500000000576507262605650014661 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Authors: Alan W Black */ /* Date : October 1996 */ /*-----------------------------------------------------------------------*/ /* LISP level interface to ngrams */ /* */ /*=======================================================================*/ #ifndef __FNGRAM_H__ #define __FNGRAM_H__ const EST_Ngrammar &get_ngram(const EST_String &name); const EST_WFST &get_wfst(const EST_String &name); #endif // __FNGRAM_H__ festival/src/include/modules.h0000644000175000017500000000646007262605650015050 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : August 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* General functions shared between various simple modules */ /* e.e. POS, Word, Phrasify etc */ /* */ /*=======================================================================*/ #ifndef __MODULES_H__ #define __MODULES_H__ EST_Item *add_word(EST_Utterance *u, const EST_String &name); EST_Item *add_word(EST_Utterance *u, LISP word); EST_Item *add_segment(EST_Utterance *u, const EST_String &s); EST_Item *add_phrase(EST_Utterance *u); void create_phraseinput(EST_Utterance *u); #endif /* __MODULES_H__ */ festival/src/include/ModuleDescription.h0000644000175000017500000001321212442333465017020 0ustar awbawb /************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ #ifndef __MODULEDESCRIPTION_H__ #define __MODULEDESCRIPTION_H__ #include #include using namespace std; #include "EST_String.h" /** Machine readable descriptions of modules. Useful for help messages * and for verifying that a set of modules should work together. * * This is a struct rather than a class so that it can be initialised * in the source of the module. * @author Richard Caley * @version $Id: ModuleDescription.h,v 1.4 2014/11/25 14:32:04 robert Exp $ */ struct ModuleDescription { /**@name limits */ //@{ /// Number of lines of descriptive text. #define MD_MAX_DESCRIPTION_LINES (10) /// Space for input streams. #define MD_MAX_INPUT_STREAMS (5) /// Space for optional streams. #define MD_MAX_OPTIONAL_STREAMS (5) /// Space for output streams. #define MD_MAX_OUTPUT_STREAMS (5) /// Space for parameters. #define MD_MAX_PARAMETERS (10) //@} /**@name Parameter types * Use these for types to avoid typoes and to allow for a cleverer system * at a later date. */ //@{ /// t or nil #define mpt_bool "BOOL" /// Positive integer #define mpt_natnum "NATNUM" /// Integer #define mpt_int "INT" /// Floating point number #define mpt_float "FLOAT" /// Any string #define mpt_string "STRING" /// A UnitDatabase #define mpt_unitdatabase "UNITDATABASE" /// Anything #define mpt_other "OTHER" //@} /// name of module const char * name; /// version number of module float version; /// where it comes from const char * organisation; /// person(s) responsible const char * author; /// general description const char * description[MD_MAX_DESCRIPTION_LINES]; /// streams affected. struct stream_parameter { /// default stream name const char * name; /// what itis used for const char * description; }; /// Streams which must have values when the module is called. struct stream_parameter input_streams[MD_MAX_INPUT_STREAMS]; /// Streams which may or not be defined. struct stream_parameter optional_streams[MD_MAX_OPTIONAL_STREAMS]; /// Streams which will be defined after the module has run. struct stream_parameter output_streams[MD_MAX_OUTPUT_STREAMS]; /// Record for a parameter. struct parameter { /// Name of parameter const char * name; /// Type of value. const char * type; /// Default value assumed. const char * default_val; /// Human readable description of effect. const char * description; }; /// Parameters which effect the module. struct parameter parameters[MD_MAX_PARAMETERS]; /// Create human readable string from description. static EST_String to_string(const ModuleDescription &desc); /// Create a module description, initialising it properly. static struct ModuleDescription *create(); /// Print the description to the strream. static ostream &print(ostream &s, const ModuleDescription &desc); static int print(FILE *s, const ModuleDescription &desc); }; /// Output operator for descriptions. ostream &operator << (ostream &stream, const ModuleDescription &desc); //VAL_REGISTER_CLASS_DCLS(moddesc,ModuleDescription) // clang/llvm complains about this (Rob) //SIOD_REGISTER_CLASS_DCLS(moddesc,ModuleDescription) VAL_REGISTER_TYPE_DCLS(moddesc,ModuleDescription) SIOD_REGISTER_TYPE_DCLS(moddesc,ModuleDescription) #endif festival/src/include/module_support.h0000644000175000017500000001307407262605650016460 0ustar awbawb /************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Richard Caley (rjc@cstr.ed.ac.uk) */ /* Date: Tue Jul 29 1997 */ /* -------------------------------------------------------------------- */ /* Some things which are useful in modules. */ /* */ /*************************************************************************/ #ifndef __MODULE_SUPPORT_H__ #define __MODULE_SUPPORT_H__ #include "EST.h" #include "festival.h" #include "ModuleDescription.h" // To extract arguments passed as a list void unpack_multiple_args(LISP args, LISP &v1, LISP &v2, LISP &v3, LISP &v4); void unpack_multiple_args(LISP args, LISP &v1, LISP &v2, LISP &v3, LISP &v4, LISP &v5); // To extract arguments for a module, modules are called // (module-function Utterance StreamName1 StreamName2 StreamName3 ...) // To tell the unpacking functions what is expected of the stream. enum RelArgType { sat_existing, // must exist sat_new, // must be new sat_replace, // erase if there, then create sat_as_is // take what we find }; void unpack_relation_arg(EST_Utterance *utt, LISP lrelation_name, EST_String &relation_name, EST_Relation *&relation, RelArgType type); void unpack_module_args(LISP args, EST_Utterance *&utt); void unpack_module_args(LISP args, EST_Utterance *&utt, EST_String &relation1_name, EST_Relation *&relation1, RelArgType type1); void unpack_module_args(LISP args, EST_Utterance *&utt, EST_String &relation1_name, EST_Relation *&relation1, RelArgType type1, EST_String &relation2_name, EST_Relation *&relation2, RelArgType type2 ); void unpack_module_args(LISP args, EST_Utterance *&utt, EST_String &relation1_name, EST_Relation *&relation1, RelArgType type1, EST_String &relation2_name, EST_Relation *&relation2, RelArgType type2, EST_String &relation3_name, EST_Relation *&relation3, RelArgType type3 ); void unpack_module_args(LISP args, EST_Utterance *&utt, EST_String &relation1_name, EST_Relation *&relation1, RelArgType type1, EST_String &relation2_name, EST_Relation *&relation2, RelArgType type2, EST_String &relation3_name, EST_Relation *&relation3, RelArgType type3, EST_String &relation4_name, EST_Relation *&relation4, RelArgType type4 ); void unpack_module_args(LISP args, EST_Utterance *&utt, EST_String &relation1_name, EST_Relation *&relation1, RelArgType type1, EST_String &relation2_name, EST_Relation *&relation2, RelArgType type2, EST_String &relation3_name, EST_Relation *&relation3, RelArgType type3, EST_String &relation4_name, EST_Relation *&relation4, RelArgType type4, EST_String &relation5_name, EST_Relation *&relation5, RelArgType type5 ); LISP lisp_parameter_get(const EST_String parameter); int int_parameter_get(const EST_String parameter, int def=0); float float_parameter_get(const EST_String parameter, float def=0.0); bool bool_parameter_get(const EST_String parameter); EST_String string_parameter_get(const EST_String parameter, EST_String def=""); #endif festival/src/arch/Makefile0000644000175000017500000000555007262605650014160 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Makefile for architecture directory ## ## ## ########################################################################### TOP=../.. DIRNAME=src/arch #BUILD_DIRS = festival unitdb BUILD_DIRS = festival ALL_DIRS = $(BUILD_DIRS) FILES=Makefile ALL = $(BUILD_DIRS) include $(TOP)/config/common_make_rules festival/src/arch/festival/Makefile0000644000175000017500000000723210006206360015755 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### # # # Makefile for festival directory # # # # Contains basic interface functions, initialization, and functions for # # using various models # # # ########################################################################### TOP=../../.. DIRNAME=src/arch/festival H = festivalP.h SRCS = festival.cc Phone.cc utterance.cc features.cc \ wave.cc wagon_interp.cc linreg.cc \ audspio.cc server.cc client.cc web.cc tcl.cc \ wfst.cc ngram.cc viterbi.cc \ ModuleDescription.cc $(TSRCS) OBJS = $(SRCS:.cc=.o) FILES = Makefile $(SRCS) $(H) #LOCAL_TEMPLATES = -pti. INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules LOCAL_DEFINES += $(FESTIVAL_DEFINES) LOCAL_INCLUDES += $(FESTIVAL_INCLUDES) festival.o: festival.cc $(CXX_COMMAND_TEMPLATES) -DFTNAME='$(PROJECT_NAME)' -DFTLIBDIRC='$(FTLIBDIR)' -DFTVERSION='$(PROJECT_VERSION)' -DFTSTATE='$(PROJECT_STATE)' -DFTDATE='$(PROJECT_DATE)' -DFTOSTYPE=\"$(SYSTEM_TYPE)\" festival.cc festival/src/arch/festival/festival.cc0000644000175000017500000004632412445313754016465 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black and Paul Taylor */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Top level file for synthesizer */ /* */ /*=======================================================================*/ #include #include "EST_unix.h" #include "EST_Pathname.h" #include #include "festival.h" #include "festivalP.h" #include "siod.h" #include "ModuleDescription.h" void festival_lisp_funcs(void); void festival_lisp_vars(void); void festival_banner(void); void festival_load_default_files(void); // Visual studio 2012 needs this #ifdef SYSTEM_IS_WIN32 #define snprintf _snprintf #endif #define _S_S_S(S) #S #define STRINGIZE(S) _S_S_S(S) const char *festival_version = STRINGIZE(FTVERSION) ":" STRINGIZE(FTSTATE) " " STRINGIZE(FTDATE); // Allow the path to be passed in without quotes because Windoze command line // is stupid // Extra level of indirection needed to get an extra macro expansion. Yeuch. #ifdef FTLIBDIRC # define FTLIBDIR STRINGIZE(FTLIBDIRC) #endif #ifdef FTOSTYPEC # define FTOSTYPE STRINGIZE(FTOSTYPEC) #endif #ifndef FTLIBDIR #define FTLIBDIR "/projects/festival/lib/" #endif #ifndef FTOSTYPE #define FTOSTYPE "" #endif const char *festival_libdir = FTLIBDIR; ostream *cdebug; static int festival_server_port = 1314; static EST_StrList sub_copyrights; #if 0 LISP describe_module(LISP lname, LISP lstream); LISP describe_all_modules(void); #endif static int festival_initialized = 0; void festival_initialize(int load_init_files,int heap_size) { // all initialisation if (! festival_initialized ) { siod_init(heap_size); siod_est_init(); // add support for EST objects /* siod_fringe_init(); */ // and for talking to fringe. siod_prog_name = "festival"; cdebug = new ofstream("/dev/null"); // This wont work on Win/NT stddebug = fopen("/dev/null","w"); festival_lisp_vars(); festival_lisp_funcs(); if (load_init_files) festival_load_default_files(); festival_initialized = TRUE; } else { cerr << "festival_initialize() called more than once" << endl; } return; } void festival_repl(int interactive) { // top level read eval print loop siod_primary_prompt = "festival> "; siod_secondary_prompt = "> "; if (interactive) festival_banner(); siod_repl(interactive); } void festival_server_mode(void) { // Go into server mode LISP lport; lport = siod_get_lval("server_port",NULL); if (lport != NULL) festival_server_port = get_c_int(lport); festival_start_server(festival_server_port); } void festival_init_lang(const EST_String &language) { // Call the lisp function to setup for names language leval(cons(rintern("select_language"), cons(quote(rintern(language)),NIL)), NIL); } int festival_load_file(const EST_String &fname) { // Load and evaluate named file EST_String b; b = EST_String("(load ")+quote_string(fname,"\"","\\",1)+")"; // I used to do the above without the b intermediate variable // but that caused a crash for some compilers on some machines return festival_eval_command(b); } int festival_eval_command(const EST_String &command) { // Eval command but catch any errors // catch any errors and return true or false depending on success jmp_buf *old_errjmp = est_errjmp; int old_errjmp_ok = errjmp_ok; int rvalue = TRUE; LISP l; gc_protect(&l); errjmp_ok = 1; est_errjmp = walloc(jmp_buf,1); if (setjmp(*est_errjmp)) { rvalue = FALSE; } else { EST_String ll = command; // copy it; l = read_from_string((char *)ll); leval(l,NIL); rvalue = TRUE; } gc_unprotect(&l); // Restore error handler wfree(est_errjmp); est_errjmp = old_errjmp; errjmp_ok = old_errjmp_ok; return rvalue; } void festival_tidy_up(void) { // tidy up before exit // specifically this closes any ope files, which can be useful // if finding errors siod_tidy_up(); } void festival_banner(void) { /* Print out a banner, copyright and version */ if (siod_get_lval("hush_startup",NULL) == NIL) { EST_Litem *t; cout << "\n" << STRINGIZE(FTNAME) << " " << festival_version << endl; cout << "Copyright (C) University of Edinburgh, 1996-2010. " << "All rights reserved." << endl; if (sub_copyrights.length() > 0) { cout << "\n"; for (t = sub_copyrights.head(); t != 0; t = t->next()) cout << sub_copyrights.item(t); } cout << "For details type `(festival_warranty)'" << endl; } } void festival_def_utt_module(const char *name,LISP (*fcn)(LISP), const char *docstring) { // Define an utterance module to the LISP system init_subr_1(name,fcn,docstring); } int festival_say_file(const EST_String &fname) { /* Say this file as text */ // should use pathname on this return festival_eval_command(EST_String("(tts ")+ quote_string(fname,"\"","\\",1)+ " nil)"); } int festival_say_text(const EST_String &text) { /* Say this text */ return festival_eval_command(EST_String("(SayText ")+ quote_string(text,"\"","\\",1)+ ")"); } int festival_text_to_wave(const EST_String &text,EST_Wave &wave) { /* Convert text to waveform */ LISP lutt; EST_Wave *w; if (!festival_eval_command(EST_String("(set! wave_utt (SynthText ")+ quote_string(text,"\"","\\",1)+ "))")) return FALSE; lutt = siod_get_lval("wave_utt",NULL); if (!utterance_p(lutt)) return FALSE; w = get_utt_wave(utterance(lutt)); if (w == 0) return FALSE; wave = *w; return TRUE; } void festival_wait_for_spooler(void) { leval(cons(rintern("audio_mode"), cons(quote(rintern("close")),NIL)),NIL); } static LISP lisp_debug_output(LISP arg) { // switch debug output stream if (cdebug != &cerr) delete cdebug; if (stddebug != stderr) fclose(stddebug); if (arg == NIL) { // this might be a problem on non-Unix machines cdebug = new ofstream("/dev/null"); stddebug = fopen("/dev/null","w"); } else { cdebug = &cerr; stddebug = stderr; } return NIL; } void festival_load_default_files(void) { // Load in default files, init.scm. Users ~/.festivalrc // (or whatever you wish to call it) is loaded by init.scm EST_String userinitfile, home_str, initfile; // Load library init first initfile = (EST_String)EST_Pathname(festival_libdir).as_directory() + "init.scm"; if (access((const char *)initfile,R_OK) == 0) vload(initfile,FALSE); else cerr << "Initialization file " << initfile << " not found" << endl; } void festival_lisp_vars(void) { // set up specific lisp variables EST_TokenStream ts; int major,minor,subminor; siod_set_lval("libdir",strintern(festival_libdir)); if (!streq(FTOSTYPE,"")) siod_set_lval("*ostype*",cintern(FTOSTYPE)); siod_set_lval("festival_version", strcons(strlen(festival_version),festival_version)); ts.open_string(festival_version); ts.set_WhiteSpaceChars(". "); major = atoi(ts.get().string()); minor = atoi(ts.get().string()); subminor = atoi(ts.get().string()); ts.close(); siod_set_lval("festival_version_number", cons(flocons(major), cons(flocons(minor), cons(flocons(subminor),NIL)))); siod_set_lval("*modules*",NIL); siod_set_lval("*module-descriptions*",NIL); if (nas_supported) proclaim_module("nas"); if (esd_supported) proclaim_module("esd"); if (sun16_supported) proclaim_module("sun16audio"); if (freebsd16_supported) proclaim_module("freebsd16audio"); if (linux16_supported) proclaim_module("linux16audio"); if (macosx_supported) proclaim_module("macosxaudio"); if (win32audio_supported) proclaim_module("win32audio"); if (mplayer_supported) proclaim_module("mplayeraudio"); // Add etc-dir path and machine specific directory etc/$OSTYPE char *etcdir = walloc(char,strlen(festival_libdir)+strlen("etc/")+ strlen(FTOSTYPE)+3); sprintf(etcdir,"%s/etc/%s/",festival_libdir,FTOSTYPE); char *etcdircommon = walloc(char,strlen(festival_libdir)+strlen("etc/")+3); sprintf(etcdircommon,"%s/etc/",festival_libdir); // Modify my PATH to include these directories siod_set_lval("etc-path",cons(rintern(etcdir), cons(rintern(etcdircommon),NIL))); const char *path = getenv("PATH"); if (path == 0) path = ""; char *newpath = walloc(char,1024+strlen(path)+strlen(etcdir)+ strlen(etcdircommon)); sprintf(newpath,"PATH=%s:%s:%s",path,etcdir,etcdircommon); putenv(newpath); wfree(etcdir); wfree(etcdircommon); return; } static LISP lmake_tmp_filename(void) { EST_String tfile = make_tmp_filename(); return strintern(tfile); } LISP l_wagon(LISP si, LISP tree); LISP l_lr_predict(LISP si, LISP lr_model); void festival_unitdb_init(void); LISP Gen_Viterbi(LISP utt); inline int utf8_sequence_length(char c0) { // Get the expected length of UTF8 sequence given its most // significant byte return (( 0xE5000000 >> (( c0 >> 3 ) & 0x1E )) & 3 ) + 1; } LISP utf8_explode(LISP name) { /* return a list of utf-8 characters as strings */ LISP chars = NIL; const char *str = (const char*)get_c_string(name); char utf8char[5]; char c0; int charlength; while ((c0 = *str)) { charlength = utf8_sequence_length(c0); snprintf(utf8char, charlength + 1, "%s", str); chars = cons(strcons(charlength, utf8char), chars); str += charlength; } return reverse(chars); } int utf8_ord(const char *utf8_seq) { unsigned int len; int ord; unsigned char c0, c1, c2, c3; // Potential bytes in the UTF8 symbol c0 = utf8_seq[0]; len = utf8_sequence_length(c0); // Make sure the string sequence we received matches with the // expected length, and that the expected length is nonzero. if ( (len == 0) || (len != strlen(utf8_seq))) { return -1; } if (len == 1) { // ASCII sequence. return c0; } c1 = utf8_seq[1]; if (len == 2) { ord = ((c0 & 0x1F) << 6) | (c1 & 0x3F); if (ord < 0x80) return -1; return ord; } c2 = utf8_seq[2]; if (len == 3) { if ((c2 & 0xC0) != 0x80) return -1; ord = ((c0 & 0x0F) << 12) | ((c1 & 0x3F) << 6) | (c2 & 0x3F); if (ord < 0x800 || (ord >= 0xD800 && ord <= 0xDFFF)) return -1; return ord; } c3 = utf8_seq[3]; if (len == 4) { if ((c3 & 0xC0) != 0x80) return -1; ord = ((c0 & 0x7) << 18) | ((c1 & 0x3F) << 12) | ((c2 & 0x3F) << 6) | (c3 & 0x3F); if (ord < 0x10000 || ord > 0x10FFFF) return -1; return ord; } return -1; } LISP utf8_ord_wrapper(LISP utf8_char) { const char *ch = (const char *)get_c_string(utf8_char); return lisp_val(utf8_ord(ch)); } int utf8_chr(int ord, char* utf8char) { unsigned int utf8len; int i = 0; if (ord < 0x80) { utf8len = 1; } else if (ord < 0x800) { utf8len = 2; } else if (ord <= 0xFFFF) { utf8len = 3; } else if (ord <= 0x200000) { utf8len = 4; } else { // Replace invalid character with FFFD utf8len = 2; ord = 0xFFFD; } i = utf8len; // Index into utf8char utf8char[i--] = 0; switch (utf8len) { // These fallthrough deliberately case 6: utf8char[i--] = (ord | 0x80) & 0xBF; ord >>= 6; case 5: utf8char[i--] = (ord | 0x80) & 0xBF; ord >>= 6; case 4: utf8char[i--] = (ord | 0x80) & 0xBF; ord >>= 6; case 3: utf8char[i--] = (ord | 0x80) & 0xBF; ord >>= 6; case 2: utf8char[i--] = (ord | 0x80) & 0xBF; ord >>= 6; case 1: switch (utf8len) { case 0: case 1: utf8char[i--] = ord; break; case 2: utf8char[i--] = ord | 0xC0; break; case 3: utf8char[i--] = ord | 0xE0; break; case 4: utf8char[i--] = ord | 0xF0; } } return utf8len; } LISP utf8_chr_wrapper(LISP ord) { char ch[5]; int utf8len; utf8len = utf8_chr(get_c_int(ord), ch); if (utf8len == 0) { return NIL; } return strcons(utf8len, ch); } void festival_lisp_funcs(void) { // declare festival specific Lisp functions // Standard functions festival_utterance_init(); festival_features_init(); festival_wave_init(); festival_Phone_init(); festival_tcl_init(); // does nothing if TCL not selected festival_wfst_init(); festival_ngram_init(); // festival_unitdb_init(); // General ones festival_init_modules(); // Some other ones that aren't anywhere else init_subr_1("parse_url", lisp_parse_url, "(parse_url URL)\n\ Split URL into a list (protocol host port path) suitable\n\ for giving to fopen."); init_subr_0("make_tmp_filename",lmake_tmp_filename, "(make_tmp_filename)\n\ Return name of temporary file."); init_subr_1("debug_output",lisp_debug_output, "(debug_output ARG)\n\ If ARG is non-nil cause all future debug output to be sent to cerr,\n\ otherwise discard it (send it to /dev/null)."); init_subr_1("utf8explode", utf8_explode, "(utf8explode utf8string)\n\ Returns a list of utf-8 characters in given string."); init_subr_1("utf8ord", utf8_ord_wrapper, "(utf8ord utf8char)\n\ Returns the decimal value utf8 character. -1 if character is invalid."); init_subr_1("utf8chr", utf8_chr_wrapper, "(utf8chr ord)\n\ Returns the utf8 character corresponding to given decimal value. Invalid characters are replaced by 0xFFFD."); init_subr_2("wagon",l_wagon, "(wagon ITEM TREE)\n\ Apply the CART tree TREE to ITEM. This returns the full\n\ predicted form, you need to extract the value from the returned form\n\ itself. [see CART trees]"); init_subr_2("lr_predict",l_lr_predict, "(lr_predict ITEM LRMODEL)\n\ Apply the linear regression model LRMODEL to ITEM in. This\n\ returns float value by summing the product of the coeffients and values\n\ returned by the specified features in ITEM. [see Linear regression]"); init_subr_1("Gen_Viterbi",Gen_Viterbi, "(Gen_Viterbi UTT)\n\ Applies viterbi search algorithm based on the parameters in\n\ gen_vit_params. Basically allows user candidate selection function\n\ combined with ngrams."); #if 0 init_subr_2("describe_module", describe_module, "(describe_module NAME STREAM)\n\ Print a description of the named module."); init_subr_0("describe_all_modules", describe_all_modules, "(describe_all_modules)\n\ Print descriptions of all modules."); #endif } void proclaim_module(const EST_String &name, const EST_String &banner_copyright, const ModuleDescription *description) { // Add this name to the variable *modules*, so people can test for // it in the lisp world LISP mods = siod_get_lval("*modules*",NULL); LISP name_sym = rintern(name); siod_set_lval("*modules*",cons(name_sym,mods)); if (banner_copyright != "") sub_copyrights.append(name + ": " + banner_copyright); if (description != NULL) { LISP module_descriptions = siod_get_lval("*module-descriptions*",NULL); LISP scheme_desc = siod(description); siod_set_lval("*module-descriptions*", cons(cons(name_sym, cons(scheme_desc, NIL)), module_descriptions)); } } void proclaim_module(const EST_String &name, const ModuleDescription *description) { proclaim_module(name, "", description); } void init_module_subr(const char *name, LISP (*fcn)(LISP), const ModuleDescription *description) { char * desc_string = NULL; if (description) { EST_String desc(ModuleDescription::to_string(*description)); desc_string = wstrdup(desc); } init_lsubr((char *)name, fcn, desc_string); // delete desc_string; } LISP ft_get_param(const EST_String &pname) { EST_Features &p = Param(); if (p.present(pname)) return lisp_val(p.f(pname)); else return NIL; #if 0 LISP params,lpair; params = siod_get_lval("Parameter","no parameters"); lpair = siod_assoc_str(pname,params); if (lpair == NIL) return NIL; else return car(cdr(lpair)); #endif } void print_string(EST_String s) { cout << s << endl; } LISP map_pos(LISP posmap, LISP pos) { // Map specified features LISP l; if (consp(pos) || (pos == NIL)) return pos; for (l=posmap; l != NIL; l=cdr(l)) if (siod_member_str(get_c_string(pos),car(car(l))) != NIL) return car(cdr(car(l))); return pos; } EST_String map_pos(LISP posmap, const EST_String &pos) { LISP l; for (l=posmap; l != NIL; l=cdr(l)) if (siod_member_str(pos,car(car(l))) != NIL) return get_c_string(car(cdr(car(l)))); return pos; } festival/src/arch/festival/Phone.cc0000644000175000017500000004260311465010356015706 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Phone Name sybsystem, allowing definitions of PhoneSets */ /* by arbitrary features (and mapping) */ /* */ /*=======================================================================*/ #include #include "EST_unix.h" #include "festival.h" #include "festivalP.h" static void check_phoneset(void); static void ps_add_def(const EST_String &name, PhoneSet *ps); static LISP lisp_select_phoneset(LISP phoneset); static LISP phone_set_list = NULL; static PhoneSet *current_phoneset = NULL; static const EST_String f_cvox("cvox"); static const EST_String f_vc("vc"); static const EST_String f_ctype("ctype"); VAL_REGISTER_CLASS(phone,Phone) SIOD_REGISTER_CLASS(phone,Phone) VAL_REGISTER_CLASS(phoneset,PhoneSet) SIOD_REGISTER_CLASS(phoneset,PhoneSet) int Phone::match_features(Phone *foreign) { // Try to match the features of the foreign phone with this one EST_Litem *f; for (f=features.list.head(); f != 0; f=f->next()) { if ( features.list(f).v != foreign->val(features.list(f).k)) return FALSE; } return TRUE; } PhoneSet::~PhoneSet() { gc_unprotect(&silences); gc_unprotect(&map); gc_unprotect(&feature_defs); gc_unprotect(&phones); } Phone * PhoneSet::member(const EST_String &ph) const { LISP p = siod_assoc_str(ph,phones); if (p != 0) return phone(car(cdr(p))); else { cerr << "Phone \"" << ph << "\" not member of PhoneSet \"" << psetname << "\"" << endl; return 0; } } const char *PhoneSet::phnum(const int n) const { // return the nth phone name int i; LISP p; // Should use nth for this, but I want to controll the error case for (i=0,p=phones; p != NIL; p=cdr(p),i++) { if (i == n) return get_c_string(car(car(p))); } cerr << "Phone (phnum) " << n << " too large, not that many members in PhoneSet \"" << psetname << "\"" << endl; festival_error(); return NULL; } int PhoneSet::add_phone(Phone *phone) { // Add phone (deleting existing one with warning) LISP lpair; lpair = siod_assoc_str(phone->phone_name(),phones); if (lpair == NIL) { phones = cons(make_param_lisp(phone->phone_name(), siod(phone)), phones); return TRUE; } else return FALSE; } void PhoneSet::set_feature(const EST_String &name, LISP vals) { LISP lpair; lpair = siod_assoc_str(name, feature_defs); if (lpair == NIL) feature_defs = cons(make_param_lisp(name,vals),feature_defs); else { cerr << "PhoneSet: replacing feature definition of " << name << " PhoneSet " << psetname << endl; CAR(cdr(lpair)) = vals; } } int PhoneSet::phnum(const char *phone) const { // Return a unique number for this phone, i.e. position in list int i; LISP p; for (i=0,p=phones; p != NIL; p=cdr(p),i++) { if (streq(phone,get_c_string(car(car(p))))) return i; } cerr << "Phone \"" << phone << "\" not member of PhoneSet \"" << psetname << "\"" << endl; festival_error(); return -1; } Phone *PhoneSet::find_matched_phone(Phone *foreign) { // find a phone in the current set that matches the features // in foreign (from a different phoneset) LISP p; for (p=phones; p != NIL; p=cdr(p)) { if (phone(car(cdr(car(p))))->match_features(foreign)) return phone(car(cdr(car(p)))); } // could try harder cerr << "Cannot map phoneme " << *foreign << endl; festival_error(); return 0; } int PhoneSet::is_silence(const EST_String &ph) const { // TRUE is ph is silence return (siod_member_str(ph,silences) != NIL); } void PhoneSet::set_silences(LISP sils) { silences=sils; } void PhoneSet::set_map(LISP m) { map=m; } PhoneSet &PhoneSet::operator = (const PhoneSet &a) { psetname = a.psetname; silences = a.silences; map = a.map; feature_defs = a.feature_defs; phones = a.phones; return *this; } LISP make_phoneset(LISP args,LISP env) { // define a new phoneme set (void)env; PhoneSet *ps = new PhoneSet; Phone *phone; LISP f,p,pv; LISP name, features, phones; EST_String feat,val; int num_feats; name = car(args); features = car(cdr(args)); phones = car(cdr(cdr(args))); ps->set_phone_set_name(get_c_string(name)); // Define the phonetic features num_feats = siod_llength(features); for (f=features; f != NIL; f=cdr(f)) ps->set_feature(get_c_string(car(car(f))),cdr(car(f))); // Define the phones for (p=phones; p != NIL; p=cdr(p)) { if (siod_llength(cdr(car(p))) != num_feats) { cerr << "Wrong number of phone features for " << get_c_string(car(car(p))) << " in " << get_c_string(name) << endl; festival_error(); } phone = new Phone; phone->set_phone_name(get_c_string(car(car(p)))); for (pv=cdr(car(p)),f=features; f != NIL; pv=cdr(pv),f=cdr(f)) { feat = get_c_string(car(car(f))); val = get_c_string(car(pv)); if (ps->feat_val(feat,val)) phone->add_feat(feat,val); else { cerr << "Phone " << phone->phone_name() << " has invalid value " << get_c_string(car(pv)) << " for feature " << feat << endl; festival_error(); } } if (ps->add_phone(phone) == FALSE) { cerr << "Phone " << phone->phone_name() << " multiply defined " << endl; festival_error(); } } ps_add_def(ps->phone_set_name(),ps); current_phoneset = ps; // selects this one as current return NIL; } static LISP lisp_set_silence(LISP silences) { // Set list of names as silences for current phoneset check_phoneset(); current_phoneset->set_silences(silences); return silences; } PhoneSet *phoneset_name_to_set(const EST_String &name) { LISP lpair = siod_assoc_str(name,phone_set_list); if (lpair == NIL) { cerr << "Phoneset " << name << " not defined" << endl; festival_error(); } return phoneset(car(cdr(lpair))); } static LISP lisp_select_phoneset(LISP pset) { // Select named phoneset and make it current EST_String name = get_c_string(pset); LISP lpair; lpair = siod_assoc_str(name,phone_set_list); if (lpair == NIL) { cerr << "Phoneset " << name << " not defined" << endl; festival_error(); } else current_phoneset = phoneset(car(cdr(lpair))); return pset; } static void ps_add_def(const EST_String &name, PhoneSet *ps) { // Add phoneset to list of phonesets LISP lpair; if (phone_set_list == NIL) gc_protect(&phone_set_list); lpair = siod_assoc_str(name,phone_set_list); if (lpair == NIL) { phone_set_list = cons(cons(rintern(name), cons(siod(ps),NIL)), phone_set_list); } else { cwarn << "Phoneset \"" << name << "\" redefined" << endl; setcar(cdr(lpair),siod(ps)); } return; } static void check_phoneset(void) { // check if there is a phoneset defined if (current_phoneset == NULL) { cerr << "No phoneset currently selected"; festival_error(); } } static EST_Val ff_ph_feature(EST_Item *s,const EST_String &name) { // This function is called for all phone features. // It looks at the name to find out the // the actual name used to call this feature and removed the // ph_ prefix and uses the remainder as the phone feature name Phone *phone_def; if (!name.contains("ph_",0)) { cerr << "Not a phone feature function " << name << endl; festival_error(); } check_phoneset(); const EST_String &fname = name.after("ph_"); phone_def = current_phoneset->member(s->name()); if (phone_def == 0) { cerr << "Phone " << s->name() << " not in PhoneSet \"" << current_phoneset->phone_set_name() << "\"" << endl; festival_error(); } const EST_String &rrr = phone_def->val(fname,EST_String::Empty); if (rrr == EST_String::Empty) { cerr << "Phone " << s->name() << " does not have feature " << fname << endl; festival_error(); } return EST_Val(rrr); } static PhoneSet *find_phoneset(EST_String name) { // get the phone set from the phone set list LISP lpair; lpair = siod_assoc_str(name,phone_set_list); if (lpair == NIL) { cerr << "Phoneset \"" << name << "\" not defined" << endl; festival_error(); } return phoneset(car(cdr(lpair))); } const EST_String &map_phone(const EST_String &fromphonename, const EST_String &fromsetname, const EST_String &tosetname) { PhoneSet *fromset, *toset; Phone *fromphone, *tophone; fromset = find_phoneset(fromsetname); toset = find_phoneset(tosetname); // should check specific matches in fromset first fromphone = fromset->member(fromphonename); if (fromphone == 0) festival_error(); tophone = toset->find_matched_phone(fromphone); return tophone->phone_name(); } int ph_is_silence(const EST_String &ph) { // TRUE if this phone is silence check_phoneset(); return current_phoneset->is_silence(ph); } EST_String ph_silence(void) { // return the first silence in the current_phoneset EST_String s; check_phoneset(); if (current_phoneset->get_silences() == NIL) { cerr << "No silences set for PhoneSet\"" << current_phoneset->phone_set_name() << "\"" << endl; return "sil"; } else return get_c_string(car(current_phoneset->get_silences())); } int ph_is_vowel(const EST_String &ph) { // TRUE if this phone is a vowel, assumes the feature vc is used return (ph_feat(ph,f_vc) == "+"); } int ph_is_consonant(const EST_String &ph) { // TRUE if this phone is a consonant, assumes the feature vc is used return ((ph_feat(ph,f_vc) == "-") && !(ph_is_silence(ph))); } int ph_is_liquid(const EST_String &ph) { // TRUE if this phone is a liquid return (ph_feat(ph,f_ctype) == "l"); } int ph_is_approximant(const EST_String &ph) { // TRUE if this phone is an approximant return (ph_feat(ph,f_ctype) == "r"); } int ph_is_stop(const EST_String &ph) { // TRUE if this phone is a stop return (ph_feat(ph,f_ctype) == "s"); } int ph_is_fricative(const EST_String &ph) { // TRUE if this phone is a stop return (ph_feat(ph,f_ctype) == "f"); } int ph_is_nasal(const EST_String &ph) { // TRUE if this phone is a nasal return (ph_feat(ph,f_ctype) == "n"); } int ph_is_obstruent(const EST_String &ph) { // TRUE if this phone is a obstruent EST_String v = ph_feat(ph,f_ctype); return ((v == "s") || (v == "f") || (v == "a")); } int ph_is_sonorant(const EST_String &ph) { // TRUE if this phone is a sonorant return !ph_is_obstruent(ph); } int ph_is_voiced(const EST_String &ph) { // TRUE if this phone is a sonorant return (ph_feat(ph,f_cvox) == "+"); } int ph_is_syllabic(const EST_String &ph) { // TRUE if this phone is a syllabic consonant (or vowel) // Yes I know we just don't have this ... return (ph_feat(ph,f_vc) == "+"); } const EST_String &ph_feat(const EST_String &ph,const EST_String &feat) { // Values for this phone -- error is phone of feat doesn't exist Phone *phone_def; EST_String rrr; check_phoneset(); phone_def = current_phoneset->member(ph); if (phone_def == 0) { cerr << "Phone " << ph << " not in phone set " << current_phoneset->phone_set_name() << endl; festival_error(); } return phone_def->val(feat,EST_String::Empty); } int ph_sonority(const EST_String &ph) { // assumes standard phone features Phone *p; check_phoneset(); p = current_phoneset->member(ph); if (p == 0) return 1; if (p->val(f_vc) == "+") return 5; else if (p->val(f_ctype) == "l") // || glide return 4; else if (p->val(f_ctype) == "n") return 3; else if (p->val(f_cvox) == "+") // voiced obstruents (stop fric affric) return 2; else return 1; } LISP l_phoneset(LISP options) { // Return Lisp form of current phone set LISP description=NIL; check_phoneset(); if ((options == NIL) || (siod_member_str("silences",options))) { description = cons(make_param_lisp("silences", current_phoneset->get_silences()), description); } if ((options == NIL) || (siod_member_str("phones",options))) { LISP phones = current_phoneset->get_phones(); LISP features = current_phoneset->get_feature_defs(); LISP p,f,p_desc=NIL,f_desc=NIL; for (p=phones; p != NIL; p=cdr(p)) { f_desc = NIL; for (f=reverse(features); f != NIL; f=cdr(f)) { f_desc = cons(rintern(ph_feat(get_c_string(car(car(p))), get_c_string(car(car(f))))), f_desc); } p_desc = cons(cons(car(car(p)),f_desc),p_desc); } description = cons(make_param_lisp("phones",p_desc),description); } if ((options == NIL) || (siod_member_str("features",options))) { description = cons(make_param_lisp("features", current_phoneset->get_feature_defs()), description); } if ((options == NIL) || (siod_member_str("name",options))) { description = cons(make_param_str("name", current_phoneset->phone_set_name()), description); } return description; } LISP l_phoneset_list() { LISP phonesets = NIL; LISP l; for (l=phone_set_list; l != NIL; l=cdr(l)) phonesets = cons(car(car(l)),phonesets); return phonesets; } void festival_Phone_init(void) { // define Lisp accessor functions init_fsubr("defPhoneSet",make_phoneset, "(defPhoneSet PHONESETNAME FEATURES PHONEDEFS)\n\ Define a new phoneset named PHONESETNAME. Each phone is described with a\n\ set of features as described in FEATURES. Some of these FEATURES may\n\ be significant in various parts of the system. Copying an existing\n\ description is a good start. [see Phonesets]"); init_subr_1("PhoneSet.select",lisp_select_phoneset, "(PhoneSet.select PHONESETNAME)\n\ Select PHONESETNAME as current phoneset. [see Phonesets]"); init_subr_1("PhoneSet.silences",lisp_set_silence, "(PhoneSet.silences LIST)\n\ Declare LIST of phones as silences. The first in the list should be\n\ the \"most\" silent. [see Phonesets]"); init_subr_1("PhoneSet.description",l_phoneset, "(Phoneset.description OPTIONS)\n\ Returns a lisp for of the current phoneme set. Options is a list of\n\ parts of the definition you require. OPTIONS may include, silences,\n\ phones, features and/or name. If nil all are returned."); init_subr_0("PhoneSet.list",l_phoneset_list, "(Phoneset.list)\n\ List the names of all currently defined Phonesets."); // All feature functions starting with "ph_" festival_def_ff_pref("ph_","Segment",ff_ph_feature, "Segment.ph_*\n\ Access phoneset features for a segment. This definition covers multiple\n\ feature functions where ph_ may be extended with any features that\n\ are defined in the phoneset (e.g. vc, vlng, cplace etc.)."); } festival/src/arch/festival/utterance.cc0000644000175000017500000007322411465010356016632 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* EST_Utterance access functions (from Lisp) */ /* */ /*=======================================================================*/ #include #include "EST_unix.h" #include "festival.h" #include "festivalP.h" static LISP item_features(LISP sitem, LISP leval = NIL); static LISP item_features(EST_Item *s, bool evaluate_ff=false); static LISP stream_tree_to_lisp(EST_Item *s); LISP utt_iform(EST_Utterance &utt) { return read_from_lstring(strintern(utt_iform_string(utt))); } const EST_String utt_iform_string(EST_Utterance &utt) { return utt.f("iform").string(); } const EST_String utt_type(EST_Utterance &utt) { return utt.f("type").string(); } static LISP utt_flat_repr( LISP l_utt ) { EST_String flat_repr; EST_Utterance *utt = get_c_utt( l_utt ); utt_2_flat_repr( *utt, flat_repr ); return strcons(flat_repr.length(), flat_repr.str()); } static LISP utt_feat(LISP lutt, LISP feat) { EST_Utterance *u = utterance(lutt); EST_String f = get_c_string(feat); return lisp_val(u->f(f)); } static LISP item_utt(LISP i) { return siod(get_utt(item(i))); } static LISP item_sub_utt(LISP i) { EST_Utterance *u = new EST_Utterance; sub_utterance(*u,item(i)); return siod(u); } static LISP utt_set_feat(LISP u, LISP name, LISP value) { EST_String n = get_c_string(name); if (TYPEP(value,tc_flonum)) utterance(u)->f.set(n,get_c_float(value)); else if (val_p(value)) utterance(u)->f.set_val(n,val(value)); else utterance(u)->f.set(n,get_c_string(value)); return value; } static LISP utt_save(LISP utt, LISP fname, LISP ltype) { EST_Utterance *u = utterance(utt); EST_String filename = get_c_string(fname); if (fname == NIL) filename = "save.utt"; EST_String type = get_c_string(ltype); if (ltype == NIL) type = "est_ascii"; if (type == "est_ascii") { if (u->save(filename,type) != write_ok) { cerr << "utt.save: saving to \"" << filename << "\" failed" << endl; festival_error(); } } else { cerr << "utt.save: unknown save format" << endl; festival_error(); } return utt; } static LISP utt_save_relation(LISP utt, LISP rname, LISP fname, LISP evaluate_ff) { // Save relation in named file EST_Utterance *u = utterance(utt); EST_String relname = get_c_string(rname); EST_String filename = get_c_string(fname); bool a; if ((evaluate_ff == NIL) || (get_c_int(evaluate_ff) == 0)) a = false; else a = true; if (fname == NIL) filename = "save.utt"; EST_Relation *r = u->relation(relname); if (r->save(filename, a) != write_ok) { cerr << "utt.save.relation: saving to \"" << filename << "\" failed" << endl; festival_error(); } return utt; } static LISP utt_load(LISP utt, LISP fname) { EST_Utterance *u; if (utt == NIL) u = new EST_Utterance; else u = utterance(utt); EST_String filename = get_c_string(fname); if (u->load(filename) != 0) { cerr << "utt.load: loading from \"" << filename << "\" failed" << endl; festival_error(); } if (utt == NIL) return siod(u); else return utt; } static LISP utt_relation_load(LISP utt, LISP lrelname, LISP lfilename) { EST_Utterance *u; if (utt == NIL) u = new EST_Utterance; else u = utterance(utt); EST_String filename = get_c_string(lfilename); EST_String relname = get_c_string(lrelname); EST_Relation *rel = u->create_relation(relname); if (rel->load(filename,"esps") != 0) { cerr << "utt.load.relation: loading from \"" << filename << "\" failed" << endl; festival_error(); } if (utt == NIL) return siod(u); else return utt; } static LISP utt_evaluate_features(LISP utt) { EST_Utterance *u = utterance(utt); u->evaluate_all_features(); return NIL; } static LISP utt_evaluate_relation(LISP utt, LISP rname) { EST_Utterance *u = utterance(utt); EST_String relname = get_c_string(rname); EST_Relation *r = u->relation(relname); r->evaluate_item_features(); return NIL; } static LISP utt_copy_relation(LISP utt, LISP l_old_name, LISP l_new_name) { EST_Utterance *u = utterance(utt); EST_String old_name = get_c_string(l_old_name); EST_String new_name = get_c_string(l_new_name); u->create_relation(new_name); u->relation(new_name)->f = u->relation(old_name)->f; copy_relation(*u->relation(old_name), *u->relation(new_name)); return utt; } static LISP utt_copy_relation_and_items(LISP utt, LISP l_old_name, LISP l_new_name) { EST_Utterance *u = utterance(utt); EST_String old_name = get_c_string(l_old_name); EST_String new_name = get_c_string(l_new_name); u->create_relation(new_name); u->relation(new_name)->f = u->relation(old_name)->f; *u->relation(new_name) = *u->relation(old_name); return utt; } static LISP utt_relation_print(LISP utt, LISP l_name) { EST_Utterance *u = utterance(utt); EST_String name = get_c_string(l_name); cout << *u->relation(name); return NIL; } static LISP utt_relation_items(LISP utt, LISP rname) { EST_Utterance *u = utterance(utt); EST_String relationname = get_c_string(rname); EST_Item *i; LISP l = NIL; for (i=u->relation(relationname)->head(); i != 0; i=next_item(i)) l = cons(siod(i),l); return reverse(l); } // could be merged with above LISP utt_relation_tree(LISP utt, LISP sname) { EST_Utterance *u = utterance(utt); EST_String relname = get_c_string(sname); return stream_tree_to_lisp(u->relation(relname)->head()); } static LISP stream_tree_to_lisp(EST_Item *s) { if (s == 0) return NIL; else { LISP desc = cons(strintern(s->name()), cons(item_features(s, false),NIL)); return cons(cons(desc,stream_tree_to_lisp(s->down())), stream_tree_to_lisp(s->next())); } } static LISP set_item_name(LISP litem, LISP newname) { // Set a stream's name to newname EST_Item *s = item(litem); if (s != 0) s->set_name(get_c_string(newname)); return litem; } static LISP utt_relation(LISP utt, LISP relname) { EST_Utterance *u = utterance(utt); EST_String rn = get_c_string(relname); EST_Item *r; r = u->relation(rn)->head(); return siod(r); } static LISP utt_relation_create(LISP utt, LISP relname) { EST_Utterance *u = utterance(utt); EST_String rn = get_c_string(relname); u->create_relation(rn); return utt; } static LISP utt_relation_delete(LISP utt, LISP relname) { EST_Utterance *u = utterance(utt); EST_String rn = get_c_string(relname); u->remove_relation(rn); return utt; } static LISP utt_relationnames(LISP utt) { // Return list of relation names EST_Utterance *u = utterance(utt); LISP relnames = NIL; EST_Features::Entries p; for (p.begin(u->relations); p; p++) relnames = cons(rintern(p->k),relnames); return reverse(relnames); } static LISP item_relations(LISP si) { // Return list of relation names EST_Item *s = item(si); LISP relnames = NIL; EST_Litem *p; for (p = s->relations().list.head(); p; p=p->next()) relnames = cons(rintern(s->relations().list(p).k),relnames); return reverse(relnames); } static LISP item_relation_name(LISP si) { // Return list of relation names EST_Item *s = item(si); return rintern(s->relation_name()); } static LISP item_relation_remove(LISP li, LISP relname) { EST_String rn = get_c_string(relname); EST_Item *si = item(li); remove_item(si,rn); // Just in case someone tries to access this again // we set its contents to be 0 which will picked up by item delete (EST_Val *)USERVAL(li); EST_Val vv = est_val((EST_Item *)0); USERVAL(li) = new EST_Val(vv); return NIL; } static LISP utt_relation_append(LISP utt, LISP relname, LISP li) { EST_Utterance *u = utterance(utt); EST_String rn = get_c_string(relname); EST_Relation *r = u->relation(rn); EST_Item *s=0; if (!r) return NIL; if (item_p(li)) s = item(li); s = r->append(s); if (consp(li)) { s->set_name(get_c_string(car(li))); add_item_features(s,car(cdr(li))); } return siod(s); } static LISP item_next(LISP li) { return (li == NIL) ? NIL : siod(item(li)->next()); } static LISP item_prev(LISP li) { return (li == NIL) ? NIL : siod(item(li)->prev()); } static LISP item_up(LISP li) { return (li == NIL) ? NIL : siod(item(li)->up()); } static LISP item_down(LISP li) { return (li == NIL) ? NIL : siod(item(li)->down()); } static LISP item_parent(LISP li) { if (li == NIL) return NIL; else return siod(parent(item(li))); } static LISP item_daughter1(LISP li) { if (li == NIL) return NIL; else return siod(daughter1(item(li))); } static LISP item_daughter2(LISP li) { if (li == NIL) return NIL; else return siod(daughter2(item(li))); } static LISP item_daughtern(LISP li) { if (li == NIL) return NIL; else return siod(daughtern(item(li))); } static LISP item_link1(LISP li) { if (li == NIL) return NIL; else return siod(link1(item(li))); } static LISP item_link2(LISP li) { if (li == NIL) return NIL; else return siod(link2(item(li))); } static LISP item_linkn(LISP li) { if (li == NIL) return NIL; else return siod(linkn(item(li))); } static LISP item_next_link(LISP li) { if (li == NIL) return NIL; else return siod(next_link(item(li))); } static LISP item_linkedfrom(LISP li) { if (li == NIL) return NIL; else return siod(linkedfrom(item(li))); } static LISP item_next_leaf(LISP li) { return (li == NIL) ? NIL : siod(next_leaf(item(li))); } static LISP item_first_leaf(LISP li) { return (li == NIL) ? NIL : siod(first_leaf(item(li))); } static LISP item_last_leaf(LISP li) { return (li == NIL) ? NIL : siod(last_leaf(item(li))); } static LISP item_add_link(LISP lfrom, LISP lto) { add_link(item(lfrom),item(lto)); return NIL; } static LISP utt_id(LISP lutt, LISP l_id) { EST_Utterance *u = utterance(lutt); return siod(u->id(get_c_string(l_id))); } static LISP item_next_item(LISP li) { return (li == NIL) ? NIL : siod(next_item(item(li))); } static LISP item_append_daughter(LISP li,LISP nli) { EST_Item *l = item(li); EST_Item *s = 0; if (item_p(nli)) s = item(nli); s = l->append_daughter(s); if (consp(nli)) { s->set_name(get_c_string(car(nli))); add_item_features(s,car(cdr(nli))); } return siod(s); } static LISP item_prepend_daughter(LISP li,LISP nli) { EST_Item *l = item(li); EST_Item *s = 0; if (item_p(nli)) s = item(nli); s = l->prepend_daughter(s); if (consp(nli)) { s->set_name(get_c_string(car(nli))); add_item_features(s,car(cdr(nli))); } return siod(s); } static LISP item_insert_parent(LISP li,LISP nparent) { EST_Item *l = item(li); EST_Item *s = 0; if (item_p(nparent)) s = item(nparent); s = l->insert_parent(s); if (consp(nparent)) { s->set_name(get_c_string(car(nparent))); add_item_features(s,car(cdr(nparent))); } return siod(s); } static LISP item_insert(LISP li,LISP nli,LISP direction) { EST_Item *n = item(li); EST_String dir; EST_Item *s; if (item_p(nli)) s = item(nli); else s = 0; if (direction) dir = get_c_string(direction); else dir = "after"; if (dir == "after") s = n->insert_after(s); else if (dir == "before") s = n->insert_before(s); else if (dir == "above") s = n->insert_above(s); else if (dir == "below") s = n->insert_below(s); else { cerr << "item.insert: unknown direction \"" << dir << "\"" << endl; festival_error(); } if (consp(nli)) // specified information { s->set_name(get_c_string(car(nli))); add_item_features(s,car(cdr(nli))); } return siod(s); } static LISP item_move_tree(LISP from,LISP to) { EST_Item *f = item(from); EST_Item *t = item(to); if (move_sub_tree(f,t) == TRUE) return truth; else return NIL; } static LISP item_merge_item(LISP from,LISP to) { EST_Item *f = item(from); EST_Item *t = item(to); merge_item(f,t); return truth; } static LISP item_exchange_tree(LISP from,LISP to) { EST_Item *f = item(from); EST_Item *t = item(to); if (exchange_sub_trees(f,t) == TRUE) return truth; else return NIL; } static LISP item_relation(LISP lingitem,LISP relname) { EST_Item *li = item(lingitem); EST_String rn = get_c_string(relname); return siod(li->as_relation(rn)); } void utt_cleanup(EST_Utterance &u) { // Remove all relations // This is called in the Initialization to ensure we can // continue with a nice clean utterance u.relations.clear(); } LISP make_utterance(LISP args,LISP env) { /* Make an utterance structure from given input */ (void)env; EST_Utterance *u = new EST_Utterance; EST_String t; LISP lform; u->f.set("type",get_c_string(car(args))); lform = car(cdr(args)); u->f.set("iform",siod_sprint(lform)); return siod(u); } static LISP item_delete(LISP litem) { EST_Item *s = item(litem); s->unref_all(); delete (EST_Val *)USERVAL(litem); EST_Val vv = est_val((EST_Item *)0); USERVAL(litem) = new EST_Val(vv); return NIL; } static LISP item_remove_feature(LISP litem,LISP fname) { EST_Item *s = item(litem); EST_String f = get_c_string(fname); s->f_remove(f); return rintern("t"); } static LISP item_features(LISP litem, LISP leval) { // Return assoc list of features on this stream return item_features(item(litem), ((leval != NIL) ? true : false)); } static LISP item_features(EST_Item *s, bool evaluate_ff) { // Can't simply use features_to_lisp as evaluation requires access // to item. LISP features = NIL; EST_Val tv; EST_Features::Entries p; for (p.begin(s->features()); p != 0; ++p) { const EST_Val &v = p->v; LISP fpair; if (v.type() == val_int) fpair = make_param_int(p->k, v.Int()); else if (v.type() == val_float) fpair = make_param_float(p->k, v.Float()); else if (v.type() == val_type_feats) fpair = make_param_lisp(p->k, features_to_lisp(*feats(v))); else if ((v.type() == val_type_featfunc) && evaluate_ff) { tv = (featfunc(v))(s); if (tv.type() == val_int) fpair = make_param_int(p->k, tv.Int()); else if (tv.type() == val_float) { fpair = make_param_float(p->k, tv.Float()); } else fpair = make_param_lisp(p->k, strintern(tv.string())); } else fpair = make_param_lisp(p->k, strintern(v.string())); features = cons(fpair,features); } return reverse(features); } void add_item_features(EST_Item *s,LISP features) { // Add LISP specified features to s; LISP f; for (f=features; f != NIL; f=cdr(f)) s->set_val(get_c_string(car(car(f))), val_lisp(car(cdr(car(f))))); } void festival_utterance_init(void) { // declare utterance specific Lisp functions // Standard functions init_fsubr("Utterance",make_utterance, "(Utterance TYPE DATA)\n\ Build an utterance of type TYPE from DATA. Different TYPEs require\n\ different types of data. New types may be defined by defUttType.\n\ [see Utterance types]"); init_subr_2("utt.load",utt_load, "(utt.load UTT FILENAME)\n\ Loads UTT with the streams and stream items described in FILENAME.\n\ The format is Xlabel-like as saved by utt.save. If UTT is nil a new\n\ utterance is created, loaded and returned. If FILENAME is \"-\"\n\ the data is read from stdin."); init_subr_2("utt.feat",utt_feat, "(utt.feat UTT FEATNAME)\n\ Return value of feature name in UTT."); init_subr_3("utt.set_feat",utt_set_feat, "(utt.set_feat UTT FEATNAME VALUE)\n\ Set feature FEATNAME with VALUE in UTT."); init_subr_1("utt.flat_repr", utt_flat_repr, "(utt.flat_repr UTT)\n\ Returns a flat, string representation of the linguistic information\n\ contained in fully formed utterance structure UTT." ); init_subr_3("utt.relation.load",utt_relation_load, "(utt.relation.load UTT RELATIONNAME FILENAME)\n\ Loads (and creates) RELATIONNAME from FILENAME into UTT. FILENAME\n\ should contain simple Xlabel format information. The label part\n\ may contain the label proper followed by semi-colon separated\n\ pairs of feature and value."); init_subr_3("utt.save",utt_save, "(utt.save UTT FILENAME TYPE)\n\ Save UTT in FILENAME in an Xlabel-like format. If FILENAME is \"-\"\n\ then print output to stdout. TYPE may be nil or est_ascii"); init_subr_4("utt.save.relation",utt_save_relation, "(utt.save UTT RELATIONNAME FILENAME EVALUATE_FEATURES)\n\ Save relation RELATIONNAME in FILENAME in an Xlabel-like format. \n\ If FILENAME is \"-\" then print output to stdout."); init_subr_3("utt.copy_relation", utt_copy_relation, "(utt.copy_relation UTT FROM TO)\n\ copy relation \"from\" to a new relation \"to\". Note that items\n\ are NOT copied, simply linked into the new relation"); init_subr_3("utt.copy_relation_and_items", utt_copy_relation_and_items, "(utt.copy_relation_and_items UTT FROM TO)\n\ copy relation and contents of items \"from\" to a new relation \"to\""); init_subr_2("utt.relation.print", utt_relation_print, "(utt.relation.print UTT NAME)\n\ print contents of relation NAME"); init_subr_1("utt.evaluate", utt_evaluate_features, "(utt.evaluate UTT)\n\ evaluate all the features in UTT, replacing feature functions\n\ with their evaluation."); init_subr_2("utt.evaluate.relation", utt_evaluate_relation, "(utt.evaluate.relation UTT)\n\ evaluate all the features in RELATION in UTT, replacing feature functions\n\ with their evaluation."); init_subr_2("utt.relation.items",utt_relation_items, "(utt.relation.items UTT RELATIONNAME)\n\ Return a list of stream items in RELATIONNAME in UTT. \n\ If this relation is a tree, the parent streamitem is listed before its \n\ daughters."); init_subr_2("utt.relation_tree",utt_relation_tree, "(utt.relation_tree UTT RELATIONNAME)\n\ Return a tree of stream items in RELATIONNAME in UTT. This will give a\n\ simple list if the relation has no ups and downs. \n\ [see Accessing an utterance]"); init_subr_1("item.delete",item_delete, "(item.delete ITEM)\n\ Remove this item from all relations it is in and delete it."); init_subr_2("item.set_name",set_item_name, "(item.set_name ITEM NAME)\n\ Sets ITEM's name to NAME. [see Accessing an utterance]"); init_subr_2("item.features",item_features, "(item.features ITEM EVALUATE_FEATURES))\n\ Returns all features in ITEM as an assoc list."); init_subr_2("item.remove_feature",item_remove_feature, "(item.remove_feature ITEM FNAME)\n\ Remove feature named FNAME from ITEM. Returns t is successfully\n\ remove, nil if not found."); // New Utterance architecture (Relations and items) init_subr_2("utt.relation",utt_relation, "(utt.relation UTT RELATIONNAME)\n\ Return root item of relation RELATIONNAME in UTT."); init_subr_2("utt.relation.create",utt_relation_create, "(utt.relation.create UTT RELATIONNAME)\n\ Create new relation called RELATIONNAME in UTT."); init_subr_2("utt.relation.delete",utt_relation_delete, "(utt.relation.delete UTT RELATIONNAME)\n\ Delete relation from utt, it the stream items are not linked elsewhere\n\ in the utterance they will be deleted too."); init_subr_2("item.relation.remove",item_relation_remove, "(item.relation.remove ITEM RELATIONNAME)\n\ Remove this item from Relation, if it apears in no other relation it\n\ will be deleted too, in contrast item.delete will remove an item\n\ from all other relations, while this just removes it from this relation.\n\ Note this will also remove all daughters of this item in this \n\ relation from this relation."); init_subr_1("utt.relationnames",utt_relationnames, "(utt.relationnames UTT)\n\ List of all relations in this utterance."); init_subr_3("utt.relation.append",utt_relation_append, "(utt.relation.append UTT RELATIONNAME ITEM)\n\ Append ITEM to top of RELATIONNAM in UTT. ITEM may be\n\ a LISP description of an item or an item itself."); init_subr_1("item.next",item_next, "(item.next ITEM)\n\ Return the next ITEM in the current relation, or nil if there is\n\ no next."); init_subr_1("item.prev",item_prev, "(item.prev ITEM)\n\ Return the previous ITEM in the current relation, or nil if there\n\ is no previous."); init_subr_1("item.up",item_up, "(item.up ITEM)\n\ Return the item above ITEM, or nil if there is none."); init_subr_1("item.down",item_down, "(item.down ITEM)\n\ Return the item below ITEM, or nil if there is none."); init_subr_3("item.insert",item_insert, "(item.insert ITEM1 ITEM2 DIRECTION)\n\ Insert ITEM2 in ITEM1's relation with respect to DIRECTION. If DIRECTION\n\ is unspecified, after, is assumed. Valid DIRECTIONS as before, after,\n\ above and below. Use the functions item.insert_parent and\n\ item.append_daughter for specific tree adjoining. If ITEM2 is of\n\ type item then it is added directly, otherwise it is treated as a\n\ description of an item and new one is created."); // Relation tree access/creation functions init_subr_1("item.parent",item_parent, "(item.parent ITEM)\n\ Return the item of ITEM, or nil if there is none."); init_subr_1("item.daughter1",item_daughter1, "(item.daughter1 ITEM)\n\ Return the first daughter of ITEM, or nil if there is none."); init_subr_1("item.daughter2",item_daughter2, "(item.daughter2 ITEM)\n\ Return the second daughter of ITEM, or nil if there is none."); init_subr_1("item.daughtern",item_daughtern, "(item.daughtern ITEM)\n\ Return the last daughter of ITEM, or nil if there is none."); init_subr_1("item.next_leaf",item_next_leaf, "(item.next_leaf ITEM)\n\ Return the next leaf item (i.e. one with no daughters) in this \n\ relation. Note this may traverse up and down the relation tree \n\ significantly to find it."); init_subr_1("item.first_leaf",item_first_leaf, "(item.first_leaf ITEM)\n\ Returns he left most leaf in the tree dominated by ITEM. This \n\ is like calling item.daughter1 recursively until an item with no \n\ daughters is found."); init_subr_1("item.last_leaf",item_last_leaf, "(item.last_leaf ITEM)\n\ Returns he right most leaf in the tree dominated by ITEM. This \n\ is like calling item.daughtern recursively until an item with no \n\ daughters is found."); init_subr_2("item.append_daughter",item_append_daughter, "(item.append_daughter ITEM1 ITEM2)\n\ Add a ITEM2 a new daughter (right-most) to ITEM1 in the relation of \n\ ITEM1. If ITEM2 is of type item then it is added directly otherwise\n\ ITEM2 is treated as a description of an item and a one is created\n\ with that description (name features)."); init_subr_2("item.prepend_daughter",item_prepend_daughter, "(item.prepend_daughter ITEM1 ITEM2)\n\ Add a ITEM2 a new daughter (left-most) to ITEM1 in the relation of ITEM1.\n\ If ITEM2 is of type item then it is added directly otherwise\n\ ITEM2 is treated as a description of an item and a one is created\n\ with that description (name features)."); init_subr_2("item.insert_parent",item_insert_parent, "(item.insert_parent ITEM1 ITEM2)\n\ Insert a new parent between this ITEM1 and its parentm in ITEM1's \n\ relation. If ITEM2 is of type item then it is added directly, \n\ otherwise it is treated as a description of an item and one is created\n\ with that description (name features)."); // MLS access/creation functions init_subr_1("item.link1",item_link1, "(item.link1 ITEM)\n\ Return first item linked to ITEM in current relation."); init_subr_1("item.link2",item_link2, "(item.link2 ITEM)\n\ Return second item linked to ITEM in current relation."); init_subr_1("item.linkn",item_linkn, "(item.linkn ITEM)\n\ Return last item linked to ITEM in current relation."); init_subr_1("item.next_link",item_next_link, "(item.next_link ITEM)\n\ Return next item licked to the same item ITEM is linked to."); init_subr_1("item.linkedfrom",item_linkedfrom, "(item.linkedfrom ITEM)\n\ Return the item tht is linked to ITEM."); init_subr_2("item.add_link",item_add_link, "(item.add_link ITEMFROM ITEMTO)\n\ Add a link from ITEMFROM to ITEMTO is the relation ITEMFROM is in."); init_subr_1("item.next_item",item_next_item, "(item.next_item ITEM)\n\ Will give next item in this relation visiting every item in the \n\ relation until the end. Traverses in pre-order, root followed by \n\ daughters (then siblings)."); init_subr_2("utt.id", utt_id, "(utt.id UTT id_number)\n\ Return the item in UTT whose id matches id_number."); init_subr_2("item.relation",item_relation, "(item.relation ITEM RELATIONNAME)\n\ Return the item such whose relation is RELATIONNAME. If ITEM\n\ is not in RELATIONNAME then nil is return."); init_subr_1("item.relations",item_relations, "(item.relations ITEM)\n\ Return a list of names of the relations this item is in."); init_subr_1("item.relation.name",item_relation_name, "(item.relation.name ITEM)\n\ Return the name of the relation this ITEM is currently being viewed\n\ through."); init_subr_2("item.move_tree",item_move_tree, "(item.move_tree FROM TO)\n\ Move contents, and descendants of FROM to TO. Old daughters of TO are\n\ deleted. FROM will be deleted too if it is being viewed as the same\n\ same relation as TO. FROM will be deleted from its current place in\n\ TO's relation. Returns t if successful, returns nil if TO is within FROM."); init_subr_2("item.exchange_trees",item_exchange_tree, "(item.exchange_tree FROM TO)\n\ Exchanged contents of FROM and TO, and descendants of FROM and TO.\n\ Returns t if successful, or nil if FROM or TO contain each other."); init_subr_2("item.merge",item_merge_item, "(item.merge FROM TO)\n\ Merge FROM into TO making them the same items. All features in FROM\n\ are merged into TO and all references to FROM are made to point to TO."); init_subr_1("item.get_utt",item_utt, "(item.get_utt ITEM)\n\ Get utterance from given ITEM (if possible)."); init_subr_1("sub_utt",item_sub_utt, "(sub_utt ITEM)\n\ Return a new utterance that contains a copy of this item and all its\n\ descendants and related descendants."); init_subr_1("audio_mode",l_audio_mode, "(audio_mode MODE)\n\ Control audio specific modes. Five subcommands are supported. If\n\ MODE is async, start the audio spooler so that Festival need not wait\n\ for a waveform to complete playing before continuing. If MODE is\n\ sync wait for the audio spooler to empty, if running, and they cause\n\ future plays to wait for the playing to complete before continuing.\n\ Other MODEs are, close which waits for the audio spooler to finish\n\ any waveforms in the queue and then closes the spooler (it will restart\n\ on the next play), shutup, stops the current waveform playing and empties\n\ the queue, and query which lists the files in the queue. The queue may\n\ be up to five waveforms long. [see Audio output]"); } festival/src/arch/festival/features.cc0000644000175000017500000004257012051474700016455 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Functions for accessing parts of the utterance through "features" */ /* features are a cononical addressing method for relatively accessing */ /* information in an utterance from a given item in it. */ /* */ /* A feature name is is a dotted separated string of names ended by a */ /* a final feature. The names before the final feature are navigational */ /* consisting of a few builtin forms (i.e. next and prev) or names of */ /* relations that are to be followed, */ /* */ /* For example: */ /* "name" name of the current item */ /* "n.name" name of the next item (in the current relation */ /* "n.n.name" is the name of the next next item */ /* "R:SylStructure.parent.name" name of the parent in the */ /* SylStructure relation */ /* for example from an item in the Segment relation */ /* "p.R:SylStructure.parent.syl_break" */ /* is the syllable break level of the syllable related */ /* the previous segment. */ /* "R:SylStucture.parent.R:Syllable.p.syl_break" */ /* is the syllable break level of the syllable related */ /* this item's previous Syllable item. */ /* */ /* The following features are defined for all stream items */ /* name */ /* Other feature are defined through the C++ function festival_def_ff */ /* Note duration, is no longer predefined */ /* */ /* To extre features are defined here ph_* (i.. any feature prefixed by */ /* ph_) and lisp_*. ph_ function check the phoneme set and use the */ /* remainder of the name as a phone feature return its value for the */ /* current item (which will normally be a Segment stream item). e.g. */ /* ph_vc or ph_vheight */ /* The lisp_* feature will call a lisp function (the name following */ /* the lisp_) with two arguments the utterance and the stream item. */ /* this allows arbitrary new features without recompilation */ /* */ /* Another feature defined here is utt_* (any feature prefixed by utt_) */ /* It looks up the remainder of the name as a feature defined at the */ /* Utterance level. */ /* */ /*=======================================================================*/ #include #include #include "EST_unix.h" #include #include "festival.h" #include "festivalP.h" static LISP ff_pref_assoc(const char *name, LISP alist); static LISP ff_docstrings = NULL; static LISP ff_pref_list = NULL; void festival_def_nff(const EST_String &name,const EST_String &sname, EST_Item_featfunc func,const char *doc) { // define the given feature function with documentation register_featfunc(name,func); if (ff_docstrings == NIL) gc_protect(&ff_docstrings); EST_String id = sname + "." + name; ff_docstrings = cons(cons(rintern(id),cstrcons(doc)),ff_docstrings); siod_set_lval("ff_docstrings",ff_docstrings); return; } VAL_REGISTER_FUNCPTR(pref_ffunc,FT_ff_pref_func) SIOD_REGISTER_FUNCPTR(pref_ffunc,FT_ff_pref_func) void festival_def_ff_pref(const EST_String &pref,const EST_String &sname, FT_ff_pref_func func, const char *doc) { // define the given class of feature functions // All feature functions names with this prefix will go to this func LISP lpair; lpair = siod_assoc_str(pref,ff_pref_list); if (lpair == NIL) { if (ff_pref_list == NIL) gc_protect(&ff_pref_list); ff_pref_list = cons(cons(rintern(pref), cons(siod(func),NIL)), ff_pref_list); EST_String id = sname + "." + pref; ff_docstrings = cons(cons(rintern(id),cstrcons(doc)),ff_docstrings); siod_set_lval("ff_docstrings",ff_docstrings); } else { fprintf(stderr,"ffeature (prefix) %s duplicate definition\n", (const char *)pref); festival_error(); } return; } static LISP ff_pref_assoc(const char *name, LISP alist) { // Search list of ff_pref_funcs to see if name has an appropriate // prefix LISP l; const char *prefix; for (l=alist; CONSP(l); l=CDR(l)) { prefix = get_c_string(CAR(CAR(l))); if (strstr(name,prefix) == name) return CAR(l); } // not found return NIL; } static EST_Item *parent_to(EST_Item *s,const EST_String &relname) { // Follow parent relation of s until Item in relname // *includes* testing this s if (s == 0) return 0; if (s->relations().present(relname)) return s->as_relation(relname); else return parent_to(parent(s),relname); } static EST_Item *daughter1_to(EST_Item *s,const EST_String &relname) { // Follow daughter1 relation of s until Item in relname // *includes* testing this s if (s == 0) return 0; if (s->relations().present(relname)) return s->as_relation(relname); else return daughter1_to(daughter1(s),relname); } static EST_Item *daughtern_to(EST_Item *s,const EST_String &relname) { // Follow parent relation of s until Item in relname // *includes* testing this s if (s == 0) return 0; if (s->relations().present(relname)) return s->as_relation(relname); else return daughtern_to(daughtern(s),relname); } static EST_String Feature_Separator = "."; static EST_String Feature_PunctuationSymbols = EST_String::Empty; static EST_String Feature_PrePunctuationSymbols = EST_String::Empty; static EST_Val default_feature_value(0); // Moving this to be static gives me an extra second but of course // makes it thread unsafe, and makes reentrancy dubious static EST_TokenStream ts; EST_Val ffeature(EST_Item *item,const EST_String &fname) { // Select and apply feature function name to s and return result FT_ff_pref_func pfunc; EST_Item_featfunc func = 0; LISP lpair; EST_Item *s = item; if (item == 0) return default_feature_value; if (strchr(fname,'.') == 0) { // if its a simple name do it quickly, without tokenizing if ((func = get_featfunc(fname)) != 0) return (func)(item); else if ((lpair = ff_pref_assoc(fname,ff_pref_list)) != NIL) { pfunc = pref_ffunc(CAR(CDR(lpair))); return (pfunc)(item,fname); } else // it must be a feature name for this item return item->f(fname,default_feature_value); } ts.open_string(fname); ts.set_WhiteSpaceChars(Feature_Separator); ts.set_PunctuationSymbols(Feature_PunctuationSymbols); ts.set_PrePunctuationSymbols(Feature_PrePunctuationSymbols); while (!ts.eof()) { const EST_String &Sname = ts.get().string(); const char *name = Sname; if (streq(name,"n")) s=s->next(); else if (streq(name,"p")) s=s->prev(); else if (streq(name,"nn")) s=s->next()->next(); else if (streq(name,"pp")) s=s->prev()->prev(); else if (streq(name,"up")) // up down should really be private s=s->up(); else if (streq(name,"down")) s=s->down(); else if (streq(name,"parent")) s=parent(s); else if (streq(name,"parent_to")) s=parent_to(s,ts.get().string()); else if (streq(name,"daughter1_to")) s=daughter1_to(s,ts.get().string()); else if (streq(name,"daughtern_to")) s=daughtern_to(s,ts.get().string()); else if (streq(name,"root")) s=root(s); else if (streq(name,"daughter1")) s=daughter1(s); else if (streq(name,"daughter2")) s=daughter2(s); else if (streq(name,"daughtern")) s=daughtern(s); else if (streq(name,"last")) s=s->last(); else if (streq(name,"first")) s=s->first(); else if (strncmp(name,"R:",2) == 0) // new relation structure s = s->as_relation(&name[2]); else if (s->f_present(Sname)) { EST_String p = Sname; while (!ts.eof()) p = EST_String::cat(p,Feature_Separator,ts.get().string()); return s->f(p, default_feature_value); } else if ((func = get_featfunc(Sname)) != 0) return (func)(s); else if ((lpair = ff_pref_assoc(name,ff_pref_list)) != NIL) { pfunc = pref_ffunc(CAR(CDR(lpair))); return (pfunc)(s,Sname); } else // unrecognized s = 0; if (s==0) return default_feature_value; } cerr << "Invalid ffeature name: \"" << fname << "\"" << endl; festival_error(); return default_feature_value; } static LISP lisp_item_feature(LISP litem, LISP name) { // return the ffeature name for this stream EST_Item *s = item(litem); EST_String fname = get_c_string(name); return lisp_val(ffeature(s,fname)); } static LISP lisp_item_raw_feature(LISP litem, LISP name) { // return the ffeature name for this stream EST_Item *s = item(litem); EST_String fname = get_c_string(name); EST_Val v = ffeature(s,fname); if (v.type() == val_type_feats) return siod(feats(v)); else return lisp_val(ffeature(s,fname)); } static LISP lisp_item_set_feat(LISP litem, LISP name, LISP value) { // set the feature (locally) on this sitem EST_Item *s = item(litem); EST_String fname = get_c_string(name); if (fname.contains("R:")) { cerr << "item.set_feat: cannot set feat name containing " << "\"R:\"" << endl; festival_error(); } s->set_val(fname,val_lisp(value)); return value; } static LISP lisp_item_set_function(LISP litem, LISP name, LISP funcname) { EST_Item *s = item(litem); s->set_function(get_c_string(name),get_c_string(funcname)); return funcname; } static LISP lisp_relation_feature(LISP utt, LISP relname, LISP name) { // return the ffeature name for this stream EST_Utterance *u = utterance(utt); EST_String fname = get_c_string(name); return lisp_val(u->relation(get_c_string(relname))->f.val(fname)); } static LISP lisp_relation_set_feat(LISP utt, LISP relname, LISP name, LISP val) { // return the ffeature name for this stream EST_Utterance *u = utterance(utt); EST_String fname = get_c_string(name); u->relation(get_c_string(relname))->f.set_path(fname,val_lisp(val)); return val; } static LISP lisp_relation_remove_item_feat(LISP utt, LISP relname, LISP name) { // return the ffeature name for this stream EST_Utterance *u = utterance(utt); EST_String fname = get_c_string(name); u->relation(get_c_string(relname))->remove_item_feature(fname); return NIL; } static LISP lisp_relation_remove_feat(LISP utt, LISP relname, LISP name) { // return the ffeature name for this stream EST_Utterance *u = utterance(utt); EST_String fname = get_c_string(name); u->relation(get_c_string(relname))->f.remove(fname); return NIL; } void value_sort(EST_Features &f, const EST_String &field); static LISP lisp_feature_value_sort(LISP f, LISP name) { value_sort(*(feats(f)), get_c_string(name)); return NIL; } static EST_Val ff_lisp_func(EST_Item *i,const EST_String &name) { // This function is called for features functions starting lisp_ // It calls the lisp function following that with u and i // as arguments, the return value (which must be atomic) is // then passed back as a Val. I'm not sure if this will be // particularly efficient, but it will make development of // new features quicker as they can be done in Lisp without // changing the C++ code. EST_String lfunc_name = name.after("lisp_"); LISP r,l; l = cons(rintern(lfunc_name), cons(siod(i),NIL)); r = leval(l,NIL); if ((consp(r)) || (r == NIL)) { cerr << "FFeature Lisp function: " << lfunc_name << " returned non-atomic value" << endl; festival_error(); } else if (numberp(r)) return EST_Val(get_c_float(r)); return EST_Val(get_c_string(r)); } static EST_Val ff_utt_func(EST_Item *i, const EST_String &name) { // This function is called to retrieve utterance level features for // a given item. It uses the name after utt_ to be the feature // name. The value of this feature is assumed to be atomic. EST_String lfeat_name = name.after("utt_"); EST_Utterance *u = get_utt(i); return EST_Val(get_c_string(lisp_val(u->f(lfeat_name, 0)))); } void festival_features_init(void) { // declare feature specific Lisp functions festival_def_ff_pref("lisp_","any",ff_lisp_func, "ANY.lisp_*\n\ Apply Lisp function named after lisp_. The function is called with\n\ an stream item. It must return an atomic value.\n\ This method may be inefficient and is primarily desgined to allow\n\ quick prototyping of new feature functions."); festival_def_ff_pref( "utt_", "any", ff_utt_func, "ANY.utt_*\n" "Retrieve utterance level feature, given an item.\n" "It must be an atomic value."); init_subr_2("item.feat",lisp_item_feature, "(item.feat ITEM FEATNAME)\n\ Return value of FEATNAME (which may be a simple feature name or a\n\ pathname) of ITEM."); init_subr_2("item.raw_feat",lisp_item_raw_feature, "(item.raw_feat ITEM FEATNAME)\n\ Return value of FEATNAME as native features structure \n\ (which may be a simple feature name or a\n\ pathname) of ITEM."); init_subr_2("feats.value_sort",lisp_feature_value_sort, "(feats.value_sort FEATURES NAME)\n"); init_subr_3("item.set_feat",lisp_item_set_feat, "(item.set_feat ITEM FEATNAME VALUE)\n\ Set FEATNAME to VALUE in ITEM."); init_subr_3("item.set_function",lisp_item_set_function, "(item.set_function ITEM FEATNAME FEATFUNCNAME)\n\ Set FEATNAME to feature function name FEATFUNCNAME in ITEM."); init_subr_3("utt.relation.feat",lisp_relation_feature, "(utt.relation.feat UTT RELNAME FEATNAME)\n\ Return value of FEATNAME on relation RELNAME in UTT."); init_subr_3("utt.relation.remove_feat",lisp_relation_remove_feat, "(utt.relation.remove_feat UTT RELNAME FEATNAME)\n\ Remove FEATNAME on relation RELNAME in UTT."); init_subr_3("utt.relation.remove_item_feat",lisp_relation_remove_item_feat, "(utt.relation.remove_item_feat UTT RELNAME FEATNAME)\n\ Remove FEATNAME on every item in relation RELNAME in UTT."); init_subr_4("utt.relation.set_feat",lisp_relation_set_feat, "(utt.relation.set_feat UTT RELNAME FEATNAME VALUE)\n\ Set FEATNAME to VALUE on relation RELNAME in UTT."); } festival/src/arch/festival/wave.cc0000644000175000017500000004653711465010356015611 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : October 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Interface to various low level waveform functions from Lisp */ /* */ /*=======================================================================*/ #include #include "EST_unix.h" #include #include "festival.h" #include "festivalP.h" #ifdef WIN32 #include "winsock2.h" #endif static void utt_save_f0_from_targets(EST_Utterance *u,EST_String &filename); static float f0_interpolate(EST_Item *ptval, EST_Item *tval, float time); EST_Wave *get_utt_wave(EST_Utterance *u) { EST_Relation *r; if (((r = u->relation("Wave")) == 0) || (r->head() == 0)) { cerr << "no waveform in utterance" << endl; festival_error(); } return wave(r->head()->f("wave")); } static LISP wave_save(LISP lwave,LISP fname,LISP ftype,LISP stype) { EST_Wave *w = wave(lwave); EST_String filename,filetype,sampletype; if (fname == NIL) filename = "save.wav"; else filename = get_c_string(fname); if (ftype == NIL) { if (ft_get_param("Wavefiletype")) filetype = get_c_string(ft_get_param("Wavefiletype")); else filetype = "nist"; } else filetype = get_c_string(ftype); if (stype == NIL) { if (ft_get_param("Wavesampletype")) sampletype = get_c_string(ft_get_param("Wavesampletype")); else sampletype = "short"; } else sampletype = get_c_string(stype); if (w->save_file(filename,filetype,sampletype,EST_NATIVE_BO) != write_ok) { cerr << "utt.save.wave: failed to write wave to \"" << filename << "\"" << endl; festival_error(); } return truth; } static LISP wave_load(LISP fname,LISP ftype,LISP stype,LISP srate) { EST_Wave *w = new EST_Wave; EST_read_status r; if (ftype == NIL) r = w->load(get_c_string(fname)); else if (streq("raw",get_c_string(ftype))) r = w->load_file(get_c_string(fname), get_c_string(ftype), get_c_int(srate), get_c_string(stype), EST_NATIVE_BO, 1); else r = w->load(get_c_string(fname),get_c_string(ftype)); if (r != format_ok) cerr << "Cannot load wavefile: " << get_c_string(fname) << endl; return siod(w); } static LISP wave_copy(LISP w) { return siod(new EST_Wave(*wave(w))); } static LISP wave_append(LISP w1,LISP w2) { EST_Wave *wave1 = wave(w1); EST_Wave *wave2 = wave(w2); *wave1 += *wave2; return w1; } static LISP wave_info(LISP w1) { EST_Wave *w = wave(w1); return cons(make_param_float("num_samples", w->num_samples()), cons(make_param_float("sample_rate", w->sample_rate()), cons(make_param_float("num_channels", w->num_channels()), cons(make_param_str("file_type", w->file_type()), NIL)))); } static LISP wave_set(LISP lwave,LISP lx, LISP ly, LISP lv) { EST_Wave *t = wave(lwave); t->a(get_c_int(lx),get_c_int(ly)) = (short)get_c_float(lv); return lv; } static LISP wave_set_sample_rate(LISP lwave,LISP lsr) { EST_Wave *t = wave(lwave); t->set_sample_rate(get_c_int(lsr)); return lsr; } static LISP wave_get(LISP lwave,LISP lx, LISP ly) { EST_Wave *t = wave(lwave); return flocons(t->a(get_c_int(lx),get_c_int(ly))); } static LISP wave_resize(LISP lwave,LISP lsamples, LISP lchannels) { EST_Wave *t; if (lwave) t = wave(lwave); else t = new EST_Wave; t->resize(get_c_int(lsamples),get_c_int(lchannels)); return siod(t); } static LISP wave_resample(LISP w1,LISP newrate) { EST_Wave *w = wave(w1); w->resample(get_c_int(newrate)); return w1; } static LISP wave_rescale(LISP lw,LISP lgain,LISP normalize) { EST_Wave *w = wave(lw); float gain = get_c_float(lgain); if (normalize) w->rescale(gain,TRUE); else w->rescale(gain); return lw; } void play_wave(EST_Wave *w) { EST_Option al; LISP audio; if (audsp_mode) // asynchronous mode audsp_play_wave(w); else { if ((audio = ft_get_param("Audio_Method")) != NIL) al.add_item("-p",get_c_string(audio)); if ((audio = ft_get_param("Audio_Device")) != NIL) al.add_item("-audiodevice",get_c_string(audio)); if ((audio = ft_get_param("Audio_Command")) != NIL) al.add_item("-command",quote_string(get_c_string(audio))); if ((audio = ft_get_param("Audio_Required_Rate")) != NIL) al.add_item("-rate",get_c_string(audio)); if ((audio = ft_get_param("Audio_Required_Format")) != NIL) al.add_item("-otype",get_c_string(audio)); al.add_item("-quality","HIGH"); play_wave(*w,al); } } static LISP wave_play(LISP lw) { play_wave(wave(lw)); return truth; } static LISP track_save(LISP ltrack,LISP fname,LISP ftype) { EST_Track *t = track(ltrack); EST_String filename,filetype; filename = (fname == NIL) ? "save.track" : get_c_string(fname); filetype = (ftype == NIL) ? "est" : get_c_string(ftype); if (t->save(filename, filetype) != write_ok) { cerr << "track.save: failed to write track to \"" << filename << "\"" << endl; festival_error(); } return truth; } static LISP track_load(LISP fname,LISP ftype,LISP ishift) { EST_Track *t = new EST_Track; EST_read_status r; float is = 0.0; if (ishift) is = get_c_float(ishift); if (ftype == NIL) r = t->load(get_c_string(fname),is); else r = t->load(get_c_string(fname), get_c_string(ftype), is); if (r != format_ok) cerr << "Cannot load track: " << get_c_string(fname) << endl; return siod(t); } static LISP track_index_below(LISP ltrack, LISP ltime) { EST_Track *t = track(ltrack); int index = -1; if(ltime) { index = t->index_below(get_c_float(ltime)); return flocons(index); } return NIL; } static LISP track_resize(LISP ltrack,LISP lframes, LISP lchannels) { EST_Track *t; if (ltrack) t = track(ltrack); else t = new EST_Track; t->resize(get_c_int(lframes),get_c_int(lchannels)); return siod(t); } static LISP track_set(LISP ltrack,LISP lx, LISP ly, LISP lv) { EST_Track *t = track(ltrack); t->a(get_c_int(lx),get_c_int(ly)) = get_c_float(lv); return lv; } static LISP track_set_time(LISP ltrack,LISP lx, LISP lt) { EST_Track *t = track(ltrack); t->t(get_c_int(lx)) = get_c_float(lt); return lt; } static LISP track_get(LISP ltrack,LISP lx, LISP ly) { EST_Track *t = track(ltrack); return flocons(t->a(get_c_int(lx),get_c_int(ly))); } static LISP track_get_time(LISP ltrack,LISP lx) { EST_Track *t = track(ltrack); return flocons(t->t(get_c_int(lx))); } static LISP track_frames(LISP ltrack) { return flocons((float)track(ltrack)->num_frames()); } static LISP track_channels(LISP ltrack) { return flocons((float)track(ltrack)->num_channels()); } static LISP track_copy(LISP t) { return siod(new EST_Track(*track(t))); } static LISP track_insert(LISP argv, LISP env) { int i,j; /* TRACK1 X1 TRACK2 X2 COUNT */ EST_Track *t1 = track(leval(siod_nth(0,argv),env)); int x1 = get_c_int(leval(siod_nth(1,argv),env)); EST_Track *t2 = track(leval(siod_nth(2,argv),env)); int x2 = get_c_int(leval(siod_nth(3,argv),env)); int count = get_c_int(leval(siod_nth(4,argv),env)); if (t1->num_channels() != t2->num_channels()) { cerr << "track.insert: different number of channels" << t1->num_channels() << " != " << t2->num_channels() << endl; festival_error(); } if (x1 + count >= t1->num_frames()) t1->resize(x1+count,t1->num_channels()); for (i=0; inum_channels(); j++) t1->a(x1+i,j) = t2->a(x2+i,j); /* not sure this is right */ t1->t(x1+i) = (x1+i > 0 ? t1->t(x1+i-1) : 0) + t2->t(x2+i) - (x2+i > 0 ? t2->t(x2+i-1) : 0); } return siod_nth(1,argv); } static LISP utt_save_f0(LISP utt, LISP fname) { // Save utt's F0 in fname as an ESPS file EST_Utterance *u = utterance(utt); EST_String filename = get_c_string(fname); if ((u->relation_present("F0")) && (u->relation("F0")->head() != 0)) { EST_Track *f0 = track(u->relation("F0")->head()->f("f0")); if (f0->save(filename,"esps") != write_ok) { cerr << "utt.save.f0: failed to write f0 to \"" << filename << "\"" << endl; festival_error(); } } else if (u->relation("Target") != 0) utt_save_f0_from_targets(u,filename); else { cerr << "utt.save.f0: utterance doesn't contain F0 or Target stream" << endl; festival_error(); } return utt; } static void utt_save_f0_from_targets(EST_Utterance *u,EST_String &filename) { // Modifications by Gregor Moehler to do proper target tracing (GM) EST_Item *s; EST_Track f0; float p = 0.0; float length = u->relation("Segment")->last()->f("end"); int i,frames = (int)(length / 0.010); f0.resize(frames,4); EST_Item *ptval, *tval; ptval = tval = u->relation("Target")->first_leaf(); for (i=0,s=u->relation("Segment")->first(); s != 0; s=s->next()) { if (i >= frames) break; // may hit here one before end for ( ; p < s->F("end",0); p+=0.010,i++) { if (tval != 0 && p > (float)ffeature(tval,"pos")) { ptval = tval; tval = next_leaf(tval); } if (i >= frames) break; // may hit here one before end if ((ffeature(s,"ph_vc") == "+") || (ffeature(s,"ph_cvox") == "+")) { f0(i,0) = f0_interpolate(ptval,tval,p); f0(i,1) = 1; } else { f0(i,0) = 0; f0(i,1) = 0.0; // unvoiced; } } } f0.set_channel_name("F0",0); f0.set_channel_name("prob_voice",1); f0.fill_time(0.01); if (f0.save(filename,"esps") != write_ok) { cerr << "utt.save.f0: failed to write F0 to \"" << filename << "\"" << endl; festival_error(); } return; } static float f0_interpolate(EST_Item *ptval, EST_Item *tval, float time) { // GM: changed, to use proper targets // Return interpolated F0 at time t float p1,p0,d1,d0; d0=0; d1=0; if (tval == 0) // after last target return ffeature(ptval,"f0"); else if (time < (float) ffeature(ptval,"pos")) // before 1st target return ffeature(tval,"f0"); else { p0 = ffeature(ptval,"f0"); p1 = ffeature(tval,"f0"); d0 = ffeature(ptval,"pos"); d1 = ffeature(tval,"pos"); } if (p0 == 0.0 || d1 == d0) return p1; else if (p1 == 0.0) return p0; else return p0 + (p1-p0)*(time-d0)/(d1-d0); } static LISP utt_send_wave_client(LISP utt) { // Send the waveform to a client (must be acting as server) EST_Utterance *u = utterance(utt); EST_Wave *w; EST_String tmpfile = make_tmp_filename(); LISP ltype; EST_String type; w = get_utt_wave(u); if (ft_server_socket == -1) { cerr << "utt_send_wave_client: not in server mode" << endl; festival_error(); } ltype = ft_get_param("Wavefiletype"); if (ltype == NIL) type = "nist"; else type = get_c_string(ltype); w->save(tmpfile,type); #ifdef WIN32 send(ft_server_socket,"WV\n",3,0); #else write(ft_server_socket,"WV\n",3); #endif socket_send_file(ft_server_socket,tmpfile); unlink(tmpfile); return utt; } /* Asterisk support, see http://www.asterisk.org */ static LISP utt_send_wave_asterisk(LISP utt) { // Send the waveform to a client (must be acting as server) EST_Utterance *u = utterance(utt); EST_Wave *w; EST_String tmpfile = make_tmp_filename(); LISP ltype; EST_String type; w = get_utt_wave(u); if (ft_server_socket == -1) { cerr << "utt_send_wave_asterisk: not in server mode" << endl; festival_error(); } ltype = ft_get_param("Wavefiletype"); if (ltype == NIL) type = "nist"; else type = get_c_string(ltype); w->resample(8000); w->rescale(5); w->save(tmpfile,type); #ifdef WIN32 send(ft_server_socket,"WV\n",3,0); #else write(ft_server_socket,"WV\n",3); #endif socket_send_file(ft_server_socket,tmpfile); unlink(tmpfile); return utt; } static LISP send_sexpr_to_client(LISP l) { EST_String tmpfile = make_tmp_filename(); FILE *fd; fd = fopen(tmpfile,"w"); lprin1f(l,fd); fprintf(fd,"\n"); fclose(fd); #ifdef WIN32 send(ft_server_socket,"LP\n",3,0); #else write(ft_server_socket,"LP\n",3); #endif socket_send_file(ft_server_socket,tmpfile); unlink(tmpfile); return l; } void festival_wave_init(void) { // declare utterance (wave) specific Lisp functions init_subr_4("wave.save",wave_save, "(wave.save WAVE FILENAME FILETYPE SAMPLETYPE)\n\ Save WAVE in FILENAME, respecting FILETYPE and SAMPLETYPE if specifed\n\ if these last two arguments are unspecified the global parameters\n\ Wavefiletype and Wavesampletype are used. Returns t is successful\n\ and throws an error if not."); init_subr_4("wave.load",wave_load, "(wave.load FILENAME FILETYPE SAMPLETYPE SAMPLERATE)\n\ Load and return a wave from FILENAME. Respect FILETYPE is specified\n\ if not specified respect whatever header is on the file. SAMPLETYPE\n\ and SAMPLERATE are only used if FILETYPE is raw."); init_subr_1("wave.copy",wave_copy, "(wave.copy WAVE)\n\ Return a copy of WAVE."); init_subr_2("wave.append",wave_append, "(wave.copy WAVE1 WAVE2)\n\ Destuctively append WAVE2 to WAVE1 and return WAVE1."); init_subr_1("wave.info",wave_info, "(wave.info WAVE)\n\ Returns assoc list of info about this wave."); init_subr_2("wave.resample",wave_resample, "(wave.resample WAVE NEWRATE)\n\ Resamples WAVE to NEWRATE."); init_subr_3("wave.rescale",wave_rescale, "(wave.rescale WAVE GAIN NORMALIZE)\n\ If NORMALIZE is specified and non-nil, maximizes the waveform first\n\ before applying the gain."); init_subr_1("wave.play",wave_play, "(wave.play WAVE)\n\ Play wave of selected audio"); init_subr_3("wave.resize",wave_resize, "(wave.resize WAVE NEWSAMPLES NEWCHANNELS)\n\ Resize WAVE to have NEWSAMPLES number of frames and NEWCHANNELS\n\ number of channels. If WAVE is nil a new wave is made of the\n\ requested size."); init_subr_4("wave.set",wave_set, "(wave.set WAVE X Y V)\n\ Set position X Y to V in WAVE.") ; init_subr_3("wave.get",wave_get, "(wave.get WAVE X Y)\n\ Get value of X Y in WAVE."); init_subr_2("wave.set_sample_rate",wave_set_sample_rate, "(wave.set_sample_rate WAVE SR)\n\ set sample rate to SR."); init_subr_3("track.save",track_save, "(track.save TRACK FILENAME FILETYPE)\n\ Save TRACK in FILENAME, in format FILETYPE, est is used if FILETYPE\n\ is unspecified or nil."); init_subr_3("track.load",track_load, "(track.load FILENAME FILETYPE ISHIFT)\n\ Load and return a track from FILENAME. Respect FILETYPE is specified\n\ and ISHIFT if specified."); init_subr_1("track.copy",track_copy, "(track.copy TRACK)\n\ Return a copy of TRACK."); init_subr_2("track.index_below",track_index_below, "(track.index_below TRACK TIME)\n\ Returns the first frame index before this time."); init_subr_3("track.resize",track_resize, "(track.resize TRACK NEWFRAMES NEWCHANNELS)\n\ Resize TRACK to have NEWFRAMES number of frames and NEWCHANNELS\n\ number of channels. If TRACK is nil a new track is made of the\n\ requested size."); init_subr_1("track.num_frames",track_frames, "(track.num_frames TRACK)\n\ Returns number of frames in TRACK."); init_subr_1("track.num_channels",track_channels, "(track.num_channels TRACK)\n\ Returns number of channels in TRACK."); init_subr_4("track.set",track_set, "(track.set TRACK X Y V)\n\ Set position X Y to V in TRACK."); init_subr_3("track.get",track_get, "(track.get TRACK X Y)\n\ Get value of X Y in TRACK."); init_subr_3("track.set_time",track_set_time, "(track.set_time TRACK X TIME)\n\ Set time at X to TIME in TRACK."); init_subr_2("track.get_time",track_get_time, "(track.get_time TRACK X)\n\ Get time of X in TRACK."); init_fsubr("track.insert",track_insert, "(track.insert TRACK1 X1 TRACK2 X2 COUNT)\n\ Insert TRACK2 from X2 to X2+COUNT into TRACK1 at X1. TRACK1 is resized\n\ as required."); init_subr_1("utt.send.wave.client",utt_send_wave_client, "(utt.send.wave.client UTT)\n\ Sends wave in UTT to client. If not in server mode gives an error\n\ Note the client must be expecting to receive the waveform."); init_subr_1("utt.send.wave.asterisk",utt_send_wave_asterisk, "(utt.send.wave.asterisk UTT)\n\ Sends wave in UTT to client. If not in server mode gives an error\n\ Note the client must be expecting to receive the waveform. The waveform\n\ is rescaled and resampled according to what asterisk needs"); init_subr_1("send_sexpr_to_client", send_sexpr_to_client, "(send_sexpr_to_client SEXPR)\n\ Sends given sexpression to currently connected client."); init_subr_2("utt.save.f0",utt_save_f0, "(utt.save.f0 UTT FILENAME)\n\ Save F0 of UTT as esps track file in FILENAME."); } festival/src/arch/festival/wagon_interp.cc0000644000175000017500000001543511332061214017324 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : May 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* A simple interpreter for CART trees as produced by Wagon */ /* */ /*=======================================================================*/ #include #include "EST_unix.h" #include #include "festival.h" #define ques_oper_str(X) (get_c_string(car(cdr(X)))) #define ques_operand(X) (car(cdr(cdr(X)))) static int wagon_ask(EST_Item *s, LISP tree, EST_TKVL *fcache); static LISP l_wagon_predict(EST_Item *s, LISP tree, EST_TKVL *fcache); /* It seems to be worth building a feature cache */ EST_Val wagon_predict(EST_Item *s, LISP tree) { LISP answer,val; EST_TKVL *fcache; fcache = new EST_TKVL; // feature cache saves some calls answer = l_wagon_predict(s,tree,fcache); delete fcache; // Decide if this is a number of a string val = car(siod_last(answer)); if (!FLONUMP(val)) // just in case return EST_Val(get_c_string(val)); else if (!CONSP(car(answer))) return EST_Val(get_c_float(val)); else return EST_Val(get_c_string(val)); } LISP wagon_pd(EST_Item *s, LISP tree) { // return probability distribution LISP answer; EST_TKVL *fcache; fcache = new EST_TKVL; // feature cache saves some calls answer = l_wagon_predict(s,tree,fcache); delete fcache; return answer; } static LISP l_wagon_predict(EST_Item *s, LISP tree, EST_TKVL *fcache) { // Use the tree to predict if (cdr(tree) == NIL) return car(tree); else if (wagon_ask(s,car(tree),fcache) == TRUE) return l_wagon_predict(s,car(cdr(tree)),fcache); else return l_wagon_predict(s,car(cdr(cdr(tree))),fcache); } static int wagon_ask(EST_Item *s, LISP question, EST_TKVL *fcache) { // Ask a question of this stream item EST_Val answer; const char *str_oper; const EST_String fname = get_c_string(car(question)); LISP operand; /* printf("wagon_ask %s\n",(const char *)siod_sprint(question)); */ if (!fcache->present(fname)) { answer = ffeature(s,fname); fcache->add_item(fname,answer,0); } else answer = fcache->val(fname); str_oper = ques_oper_str(question); operand = ques_operand(question); // So that you can use LISP variables in the operand (or any LISP) // check if we've got a , if so eval it. if ((consp(operand)) && (!consp(car(operand))) && (streq("+internal-comma",get_c_string(car(operand))))) operand = leval(cdr(operand),NIL); if (streq("is",str_oper)) if (answer.string() == get_c_string(operand)) { /* printf("wagon_ask %s is %s\n",(const char *)answer.string(),(const char *)get_c_string(operand)); */ return TRUE; } else { /* printf("wagon_ask %s isnot %s\n",(const char *)answer.string(),(const char *)get_c_string(operand)); */ return FALSE; } else if (streq("=",str_oper)) if (answer == get_c_float(operand)) return TRUE; else return FALSE; else if (streq("<",str_oper)) if ((float)answer < get_c_float(operand)) return TRUE; else return FALSE; else if (streq(">",str_oper)) if ((float)answer > get_c_float(operand)) return TRUE; else return FALSE; else if (streq("matches",str_oper)) if (answer.string().matches(make_regex(get_c_string(operand)))) return TRUE; else return FALSE; else if (streq("in",str_oper)) if (siod_member_str(answer.string(),operand) != NIL) return TRUE; else return FALSE; else { cerr << "Decision tree: unknown question operator: \"" << str_oper << "\"\n"; festival_error(); } return 0; } LISP l_wagon(LISP si, LISP tree) { // Lisp level binding for tree prediction EST_Item *s = item(si); LISP answer; EST_TKVL *fcache; fcache = new EST_TKVL; // feature cache saves some calls answer = l_wagon_predict(s,tree,fcache); delete fcache; return answer; } festival/src/arch/festival/linreg.cc0000644000175000017500000000770510126474330016120 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : May 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* A simple interpreter for Linear Regression models */ /* */ /*=======================================================================*/ #include #include "EST_unix.h" #include "festival.h" #define FFEATURE_NAME(X) (get_c_string(car(X))) #define FFEATURE_WEIGHT(X) (get_c_float(car(cdr(X)))) #define FFEATURE_MAPCLASS(X) (car(cdr(cdr(X)))) EST_Val lr_predict(EST_Item *s, LISP lr_model) { EST_Val v = 0.0; float answer; LISP f; const char *ffeature_name, *last_name=""; answer = FFEATURE_WEIGHT(car(lr_model)); // Intercept; for (f=cdr(lr_model); CONSP(f); f=CDR(f)) { ffeature_name = FFEATURE_NAME(CAR(f)); if (!streq(ffeature_name,last_name)) v = ffeature(s,ffeature_name); if (siod_llength(CAR(f)) == 3) { // A map class is specified if (siod_member_str(v.string(),FFEATURE_MAPCLASS(CAR(f))) != NIL) answer += FFEATURE_WEIGHT(CAR(f)); } else answer += FFEATURE_WEIGHT(CAR(f)) * (float)v; last_name = ffeature_name; } return EST_Val(answer); } LISP l_lr_predict(LISP si, LISP lr_model) { EST_Item *s = item(si); EST_Val answer; answer = lr_predict(s,lr_model); return flocons(answer.Float()); } festival/src/arch/festival/audspio.cc0000644000175000017500000002016110126474330016273 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : September 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Interface with the audio spooler */ /* */ /*=======================================================================*/ #include #include "EST_unix.h" #include #include "festival.h" #include "festivalP.h" #ifdef NO_SPOOLER void audsp_play_wave(EST_Wave *w) { cerr << "no spooler available\n"; } LISP l_audio_mode(LISP mode) { return NIL; } #else #include #include static int start_sub_process(int *fds,int argc,char **argv); static char **enargen(const char *command,int *argc); static void audsp_send(const char *c); static int *pipe_open(const char *command); static void pipe_close(int *fds); static int *audfds; static int audsp_num=0; static int audsp_pid = 0; void audsp_play_wave(EST_Wave *w) { EST_String tpref = make_tmp_filename(); char *tmpfilename = walloc(char,tpref.length()+20); sprintf(tmpfilename,"%s_aud_%05d",(const char *)tpref,audsp_num++); w->save(tmpfilename,"nist"); audsp_send(EST_String("play ")+tmpfilename+EST_String(" ")+ itoString(w->sample_rate())); wfree(tmpfilename); } static void audsp_send(const char *c) { char reply[4]; int pid; int statusp; pid = waitpid((pid_t)audsp_pid,&statusp,WNOHANG); if (pid != 0) { cerr << "Audio spooler has died unexpectedly" << endl; audsp_mode = FALSE; festival_error(); } write(audfds[0],c,strlen(c)); write(audfds[0],"\n",1); read(audfds[1],reply,3); /* confirmation */ } LISP l_audio_mode(LISP mode) { // Switch audio mode LISP audio=NIL; LISP command=NIL; if (mode == NIL) { cerr << "audio_mode: nil is not a valid mode\n"; festival_error(); } else if (streq("async",get_c_string(mode))) { // Asynchronous mode using the audio spooler. if (audsp_mode == FALSE) { audio = ft_get_param("Audio_Method"); command = ft_get_param("Audio_Command"); audfds = pipe_open("audsp"); if (audio != NIL) audsp_send(EST_String("method ")+get_c_string(audio)); if (command != NIL) { // command needs to be a single line so delete an newlines EST_String flattened = get_c_string(command); flattened.gsub("\\\n"," "); flattened.gsub("\n"," "); audsp_send(EST_String("command ")+flattened); } if ((audio = ft_get_param("Audio_Required_Rate")) != NIL) audsp_send(EST_String("rate ")+get_c_string(audio)); if ((audio = ft_get_param("Audio_Required_Format")) != NIL) audsp_send(EST_String("otype ")+get_c_string(audio)); if ((audio = ft_get_param("Audio_Device")) != NIL) audsp_send(EST_String("device ")+get_c_string(audio)); audsp_mode = TRUE; } } else if (streq("sync",get_c_string(mode))) { // Synchronous mode if (audsp_mode) pipe_close(audfds); audsp_mode = FALSE; } else if (streq("shutup",get_c_string(mode))) { if (audsp_mode) audsp_send("shutup"); else { cerr << "audio_mode: not in async mode, can't shutup\n"; festival_error(); } } else if (streq("close",get_c_string(mode))) { // return only when queue is empty if (audsp_mode) audsp_send("close"); } else if (streq("query",get_c_string(mode))) { if (audsp_mode) audsp_send("query"); else { cerr << "audio_mode: not in async mode, can't query\n"; festival_error(); } } else { cerr << "audio_mode: unknown mode \"" << get_c_string(mode) << "\"\n"; festival_error(); } return mode; } static void pipe_close(int *fds) { // Close down the pipes close(fds[0]); close(fds[1]); } static int *pipe_open(const char *command) { // Starts a subprocess with its stdin and stdout bounad to pipes // the ends of which are returned in an array int argc; char **argv; int *fds; argv = enargen(command,&argc); fds = walloc(int,2); if (start_sub_process(fds,argc,argv) != 0) { cerr << "pipe_open: failed to start subprocess: \n" << endl; cerr << "pipe_open: \"" << command << "\"\n"; festival_error(); } return fds; } static int start_sub_process(int *fds, int argc, char **argv) { // start sub_process with stdin and stdout bound to pipes whose ends // are in fds[0] and fds[1] int pid; int in[2]; int out[2]; (void)argc; if ((pipe(in) != 0) || (pipe(out) != 0)) { cerr << "pipe_open: failed to open pipes\n"; festival_error(); } switch (pid=fork()) { case 0: /* child */ close(in[1]); /* close the end child isn't using */ dup2(in[0],0); /* reassign stdin to the pipe */ close(out[0]); dup2(out[1],1); /* reassign stdout to the pipe */ execvp(argv[0],argv); cerr << "pipe_open: failed to start " << argv[0] << endl; exit(-1); /* should only get here on failure */ case -1: cerr << "pipe_open: fork failed\n"; festival_error(); default: /* parent */ close(in[0]); /* Close unused sides of the pipes */ close(out[1]); fds[0] = in[1]; fds[1] = out[0]; } audsp_pid = pid; return 0; } static char **enargen(const char *command,int *argc) { EST_TokenStream ts; char **argv; int i; ts.open_string(command); for (i=0; ts.get() != ""; i++); ts.close(); *argc = i; argv = walloc(char *,i+1); ts.open_string(command); for (i=0; i < *argc; i++) argv[i] = wstrdup(ts.get().string()); argv[i] = 0; return argv; } #endif festival/src/arch/festival/server.cc0000644000175000017500000002274110327756206016153 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan Black and Richard Tobin */ /* Date : December 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Socket support to run Festival as a server process */ /* */ /* The original socket code is derived Richard Tobin's scokpipe/pipesock */ /* examples, but have be substantially changed to be more general */ /* for Festival */ /* */ /* The logging, access control, file transfer stuff are all new */ /* */ /*=======================================================================*/ #include #include #include #include #include #include #include "EST_unix.h" #include "EST_socket.h" #include "festival.h" #include "festivalP.h" #define DEFAULT_MAX_CLIENTS 10 /* The folloing gives a server that never forks */ /* and only accepts one client at a time. This is good for */ /* OSs with an expensive implementation of fork and or waitpid (e.g. NT) */ #ifdef WIN32 #define SINGLE_CLIENT 1 #endif static int client_access_check(int fd,int client); static EST_String log_time_stamp(int client); static void log_message(int client,const char *message); int ft_server_socket = -1; ostream *cslog = NULL; int festival_start_server(int port) { // Never exits except by signals struct sockaddr_in serv_addr; int fd, fd1; int statusp; int client_name=0; int max_clients, num_clients, pid; LISP lmax_clients, llog_file; lmax_clients = siod_get_lval("server_max_client",NULL); if (lmax_clients != NULL) max_clients = get_c_int(lmax_clients); else max_clients = DEFAULT_MAX_CLIENTS; num_clients = 0; llog_file = siod_get_lval("server_log_file",NULL); if (llog_file == NIL) cslog = cdebug; else if (llog_file == siod_get_lval("t",NULL)) cslog = &cout; else cslog = new ofstream(get_c_string(llog_file),ios::app); if (!socket_initialise()) { festival_error(); } fd = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP); if (NOT_A_SOCKET(fd)) { int n = socket_error(); cerr << "socket: socket failed (" << n << ")\n"; festival_error(); } int one = 1; if (setsockopt(fd, SOL_SOCKET,SO_REUSEADDR,(char *)&one,sizeof(int)) < 0) { cerr << "socket: SO_REUSEADDR failed" << endl; festival_error(); } memset(&serv_addr, 0, sizeof(serv_addr)); serv_addr.sin_family = AF_INET; serv_addr.sin_port = htons(port); serv_addr.sin_addr.s_addr = htonl(INADDR_ANY); if (bind(fd, (struct sockaddr *)&serv_addr, sizeof(serv_addr)) != 0) { cerr << "socket: bind failed" << endl; festival_error(); } if (listen(fd, 5) != 0) { cerr << "socket: listen failed" << endl; festival_error(); } #if SINGLE_CLIENT log_message(0,EST_String("Festival server (non-forking) started on port ")+ itoString(port)); #else log_message(0,EST_String("Festival server started on port ")+ itoString(port)); #endif fflush(stdout); fflush(stderr); fflush(stdin); while(1) // never exits except by signals { if((fd1 = accept(fd, 0, 0)) < 0) { cerr << "socket: accept failed"; festival_error(); } client_name++; if (client_access_check(fd1,client_name) == FALSE) { close(fd1); continue; } #ifdef SINGLE_CLIENT ft_server_socket = fd1; repl_from_socket(fd1); log_message(client_name,"disconnected"); #else num_clients++; // Fork new image of festival and call interpreter if (num_clients > max_clients) { log_message(client_name,"failed: too many clients"); num_clients--; } else if ((pid=fork()) == 0) { ft_server_socket = fd1; repl_from_socket(fd1); log_message(client_name,"disconnected"); exit(0); } else if (pid < 0) { log_message(client_name,"failed to fork new client"); num_clients--; } while ((num_clients > 0) && (waitpid(0,&statusp,WNOHANG) != 0)) num_clients--; #endif close(fd1); } return 0; } static int client_access_check(int fd,int client) { // Check client against various possible checks to see if they // are allowed access to the server LISP passwd, access_list, deny_list; int client_access = TRUE; struct sockaddr_in peer; socklen_t addrlen=sizeof(peer); struct hostent *clienthost; const char *client_hostname; const char *client_hostnum; const char *reason = ""; getpeername(fd,(struct sockaddr *)&peer,&addrlen); clienthost = gethostbyaddr((char *)&peer.sin_addr, sizeof(peer.sin_addr),AF_INET); client_hostnum = inet_ntoa(peer.sin_addr); if (streq(client_hostnum,"0.0.0.0") || streq(client_hostnum,"127.0.0.1")) // its me ! client_hostname = "localhost"; else if (clienthost == 0) // failed to get a name client_hostname = client_hostnum; else client_hostname = clienthost->h_name; if (((deny_list = siod_get_lval("server_deny_list",NULL)) != NIL) && (siod_regex_member_str(client_hostname,deny_list) != NIL)) { client_access = FALSE; reason = "in deny list"; } else if ((access_list = siod_get_lval("server_access_list",NULL)) != NIL) { client_access = FALSE; // by default now reason = "not in access list"; if (siod_regex_member_str(client_hostname,access_list) != NIL) { reason = ""; client_access = TRUE; } } passwd = siod_get_lval("server_passwd",NULL); if ((client_access == TRUE) && (passwd != NULL)) { char *client_passwd = walloc(char,strlen(get_c_string(passwd))+1); read(fd,client_passwd,strlen(get_c_string(passwd))); client_passwd[strlen(get_c_string(passwd))] = '\0'; if (streq(get_c_string(passwd),client_passwd)) client_access = TRUE; else { client_access = FALSE; reason = "bad passwd"; } wfree(client_passwd); } char *message = walloc(char,20+strlen(client_hostname)+strlen(reason)); if (client_access == TRUE) { sprintf(message,"accepted from %s",client_hostname); log_message(client,message); } else { sprintf(message,"rejected from %s %s",client_hostname,reason); log_message(client,message); } wfree(message); return client_access; } static void log_message(int client, const char *message) { // log the message in log file *cslog << log_time_stamp(client) << message << endl; } static EST_String log_time_stamp(int client) { // returns a string with client id, time and date char lst[1024]; time_t thetime = time(0); char *cthetime = ctime(&thetime); cthetime[24] = '\0'; // get rid of \n if (client == 0) sprintf(lst,"server %s : ",cthetime); else sprintf(lst,"client(%d) %s : ",client,cthetime); return lst; } festival/src/arch/festival/client.cc0000644000175000017500000001045110126474330016106 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan Black (with Richard Tobin) */ /* Date : December 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Low level client functions, separated from the other scoket functions */ /* so things that link with these don't need the whole system */ /* */ /*=======================================================================*/ #include #include #include #include #include #include #include "EST_unix.h" #include "EST_socket.h" #include "festival.h" #include "festivalP.h" static EST_Regex ipnum("[0-9]+\\.[0-9]+\\.[0-9]+\\.[0-9]+"); int festival_socket_client(const char *host,int port) { // Return an FD to a remote server struct sockaddr_in serv_addr; struct hostent *serverhost; EST_String shost; int fd; if (!socket_initialise()) { festival_error(); } fd = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP); if (NOT_A_SOCKET(fd)) { int n = socket_error(); cerr << "socket: socket failed (" << n << ")\n"; festival_error(); } memset(&serv_addr, 0, sizeof(serv_addr)); shost = host; if (shost.matches(ipnum)) serv_addr.sin_addr.s_addr = inet_addr(host); else { serverhost = gethostbyname(host); if (serverhost == (struct hostent *)0) { cerr << "socket: gethostbyname failed" << endl; festival_error(); } memmove(&serv_addr.sin_addr,serverhost->h_addr, serverhost->h_length); } serv_addr.sin_family = AF_INET; serv_addr.sin_port = htons(port); if (connect(fd, (struct sockaddr *)&serv_addr, sizeof(serv_addr)) != 0) { cerr << "socket: connect failed" << endl; festival_error(); } return fd; } festival/src/arch/festival/web.cc0000644000175000017500000001525510126474330015414 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Authors: Alan W Black */ /* Date : November 1996 */ /*-----------------------------------------------------------------------*/ /* Some basic functions for dealing with the web and urls */ /* */ /*=======================================================================*/ #include #include "EST_unix.h" #include #include "festival.h" #include "festivalP.h" #include "EST_String.h" #if 0 static int getc_unbuffered(int fd,int *c); #endif LISP parse_url(EST_String url) { EST_String protocol, host, port, path; if (!parse_url(url, protocol, host, port, path)) err("can't parse URL", url); return cons(strintern(protocol), cons(strintern(host), cons(strintern(port), cons(strintern(path), NIL)))); } LISP lisp_parse_url(LISP l_url) { EST_String url(get_c_string(l_url)); return parse_url(url); } #if 0 LISP lisp_get_url(LISP url,LISP filename) { // Copy file identified by URL and copy it to filename // Current only file:/.../... and http:/.../... are supported EST_TokenStream us; EST_String host,file,port; char *comm = walloc(char,32+strlen(get_c_string(url))); char *getstr = walloc(char,32+strlen(get_c_string(filename))); int wwwserver=-1,c; FILE *fd, *fin, *fout; // first parse the url us.open_string(get_c_string(url)); us.set_WhiteSpaceChars(""); us.set_SingleCharSymbols(":/"); if (us.peek() == "http") { us.get(); if ((us.get() != ":") || (us.get() != "/") || (us.get() != "/")) { cerr << "url_get: malformed url" << endl; festival_error(); } host = us.get().string(); // upto next / if (us.peek() == ":") // a port is specified { us.get(); port = us.get().string(); } else port = "80"; // standard port for http servers file = us.get_upto_eoln(); sprintf(comm,"telnet %s %s",(const char *)host,(const char *)port); wwwserver = festival_socket_client(host,atoi(port)); if (wwwserver < 0) { cerr << "get_url: can't access server\n"; festival_error(); } fout = fdopen(wwwserver,"wb"); fprintf(fout,"GET %s\n",(const char *)file); fflush(fout); if ((fd=fopen(get_c_string(filename),"wb")) == NULL) { cerr << "get_url: can't open outputfile \"" << get_c_string(filename) << "\"\n"; festival_error(); } else { while (getc_unbuffered(wwwserver,&c) != EOF) putc(c,fd); fclose(fd); } close(wwwserver); wfree(comm); wfree(getstr); } else if (us.peek() == "file") { us.get(); if (us.get() != ":") { cerr << "url_get: malformed url" << endl; festival_error(); } file = us.get_upto_eoln(); if ((fin = fopen(file,"rb")) == NULL) { cerr << "get_url: unable to access file url \"" << get_c_string(url) << "\"\n"; festival_error(); } if ((fd=fopen(get_c_string(filename),"wb")) == NULL) { cerr << "get_url: can't open outputfile \"" << get_c_string(filename) << "\"\n"; fclose(fin); festival_error(); } else { while ((c=getc(fin)) != EOF) putc(c,fd); fclose(fd); fclose(fin); } } else { cerr << "get_url: unrecognizable url \"" << get_c_string(url) << "\"\n"; festival_error(); } return NIL; } LISP l_open_socket(LISP host, LISP port, LISP how) { // Open socket to remote server int fd; char *how_c; fd = festival_socket_client(get_c_string(host),get_c_int(port)); if (streq(get_c_string(how),"rw")) { // return list of r and w FILEDESCRIPTORS return cons(siod_fdopen_c(fd, EST_String(get_c_string(host))+ ":"+get_c_string(port), "rb"), cons(siod_fdopen_c(fd, EST_String(get_c_string(host))+ ":"+get_c_string(port), "wb"),NIL)); } else { if (how == NIL) how_c = "wb"; else how_c = get_c_string(how); return siod_fdopen_c(fd, EST_String(get_c_string(host))+":"+get_c_string(port), how_c); } } static int getc_unbuffered(int fd,int *rc) { // An attempted to get rid of the buffering char c; int n; n = read(fd,&c,1); *rc = c; if (n == 0) return EOF; else return 0; } #endif festival/src/arch/festival/tcl.cc0000644000175000017500000001120710126474330015412 0ustar awbawb/* * Copyright (C)1997 Jacques H. de Villiers * Copyright (C)1997 Center for Spoken Language Understanding, * Oregon Graduate Institute of Science & Technology * * The authors hereby grant permission to use, copy, modify, distribute, * and license this software and its documentation for any purpose, provided * that existing copyright notices are retained in all copies and that this * notice is included verbatim in any distributions. No written agreement, * license, or royalty fee is required for any of the authorized uses. * Modifications to this software may be copyrighted by their authors * and need not follow the licensing terms described here, provided that * the new terms are clearly indicated on the first page of each file where * they apply. * * IN NO EVENT SHALL THE AUTHORS OR DISTRIBUTORS BE LIABLE TO ANY PARTY * FOR DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES * ARISING OUT OF THE USE OF THIS SOFTWARE, ITS DOCUMENTATION, OR ANY * DERIVATIVES THEREOF, EVEN IF THE AUTHORS HAVE BEEN ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. * * THE AUTHORS AND DISTRIBUTORS SPECIFICALLY DISCLAIM ANY WARRANTIES, * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE, AND NON-INFRINGEMENT. THIS SOFTWARE * IS PROVIDED ON AN "AS IS" BASIS, AND THE AUTHORS AND DISTRIBUTORS HAVE * NO OBLIGATION TO PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR * MODIFICATIONS. * * GOVERNMENT USE: If you are acquiring this software on behalf of the * U.S. government, the Government shall have only "Restricted Rights" * in the software and related documentation as defined in the Federal * Acquisition Regulations (FARs) in Clause 52.227.19 (c) (2). If you * are acquiring the software on behalf of the Department of Defense, the * software shall be classified as "Commercial Computer Software" and the * Government shall have only "Restricted Rights" as defined in Clause * 252.227-7013 (c) (1) of DFARs. Notwithstanding the foregoing, the * authors grant the U.S. Government and others acting in its behalf * permission to use and distribute the software in accordance with the * terms specified in this license. * * * * Sat May 17 13:43:47 BST 1997 * * minor modifications to make it work in 1.1.4 -- awb@cstr.ed.ac.uk * *--------------------------------------------------------------------------- * This module adds a Tcl interpreter to Scheme, Scheme commands to * retrieve wave and phoneme info. A Scheme command to invoke Tcl and a * Tcl command to invoke Scheme. We'll load CSLUsh packages into the Tcl * interpreter to implement a client/server architecture. *--------------------------------------------------------------------------- */ #include #include #include "festival.h" #include "festivalP.h" #if SUPPORT_TCL #include static int festCmd(ClientData d, Tcl_Interp *interp, int argc, char *argv []); /* Scheme: (cslush TCL_COMMAND_STRING) * * Invoke Tcl interpreter from within Festival's scheme runtime. * Returns a LISP string, throws an error if the Tcl command fails. * For the time being a single, global Tcl interp will suffice. */ static Tcl_Interp *tcl_interpreter=NULL; static LISP tcl_eval(LISP tcl_command) { char *cmd=get_c_string(tcl_command); if (!tcl_interpreter) { tcl_interpreter=Tcl_CreateInterp(); Tcl_Init(tcl_interpreter); Tcl_CreateCommand(tcl_interpreter,"festival",festCmd,NULL,NULL); } if (Tcl_Eval(tcl_interpreter,cmd)!=TCL_OK) { cerr << tcl_interpreter->result << endl; festival_error(); } return strintern(tcl_interpreter->result); } /* TCL: festival scheme_command * * Invoke the scheme interpreter that wraps our Tcl interp. * Result is a Tcl string * (The Scheme interpreter aborts the cslush command if a scheme error * occurs. So much for it being reentrant.) */ static int festCmd(ClientData d, Tcl_Interp *interp, int argc, char *argv []) { LISP cmd; if(argc!=2) { Tcl_AppendResult(interp,"wrong # args: should be \"",argv[0], " scheme_command\"",NULL); return TCL_ERROR; } d=d; /* stop compiler warning */ cmd=strintern(argv[1]); cmd=read_from_string(cmd); cmd=leval(cmd,NIL); Tcl_SetResult(interp,get_c_string(cmd),TCL_VOLATILE); return TCL_OK; } /* register our new Scheme functions with Festival */ void festival_tcl_init(void) { init_subr_1("tcl_eval",tcl_eval, "(tcl_eval STRING)\n\ Evaluate STRING as a Tcl command, return the result as a LISP string"); proclaim_module("tcl"); } #else /* no TCL */ void festival_tcl_init(void) { // nothing to do } #endif festival/src/arch/festival/wfst.cc0000644000175000017500000001233210126474330015613 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Authors: Alan W Black */ /* Date : December 1997 */ /*-----------------------------------------------------------------------*/ /* Access to WFST classes */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "festivalP.h" static LISP wfst_loaded_list = NIL; static EST_WFST *load_wfst(const EST_String &filename); static LISP add_wfst(const EST_String &name,EST_WFST *n); SIOD_REGISTER_CLASS(wfst,EST_WFST) static LISP lisp_load_wfst(LISP name, LISP filename) { EST_WFST *n; n = load_wfst(get_c_string(filename)); add_wfst(get_c_string(name),n); return name; } static EST_WFST *load_wfst(const EST_String &filename) { EST_WFST *n = new EST_WFST(); if (n->load(filename) != 0) { fprintf(stderr,"WFST: failed to read wfst from \"%s\"\n", (const char *)filename); festival_error(); } return n; } static LISP add_wfst(const EST_String &name,EST_WFST *n) { LISP lpair; lpair = siod_assoc_str(name,wfst_loaded_list); if (wfst_loaded_list == NIL) gc_protect(&wfst_loaded_list); LISP lwfst = siod(n); if (lpair == NIL) wfst_loaded_list = cons(cons(strintern(name),cons(lwfst,NIL)),wfst_loaded_list); else { cwarn << "WFST: " << name << " recreated" << endl; setcar(cdr(lpair),lwfst); } return lwfst; } EST_WFST *get_wfst(const EST_String &name,const EST_String &filename) { // Find ngram named name, returns NULL if none; LISP lpair; lpair = siod_assoc_str(name,wfst_loaded_list); if (lpair == NIL) { if (filename != EST_String::Empty) { EST_WFST *n = load_wfst(filename); add_wfst(name,n); return n; } else { cwarn << "WFST: no wfst named \"" << name << "\" loaded" << endl; return 0; } } else return wfst(car(cdr(lpair))); } LISP lisp_wfst_transduce(LISP wfstname, LISP input) { EST_WFST *wfst = get_wfst(get_c_string(wfstname)); EST_StrList in,out; int r; if (consp(input)) siod_list_to_strlist(input,in); else siod_list_to_strlist(stringexplode(get_c_string(input)),in); r = transduce(*wfst,in,out); if (r == FALSE) return rintern("FAILED"); else return siod_strlist_to_list(out); } void festival_wfst_init() { init_subr_2("wfst.load",lisp_load_wfst, "(wfst.load NAME FILENAME)\n\ Load a WFST from FILENAME and store it named NAME for later access."); init_subr_2("wfst.transduce",lisp_wfst_transduce, "(wfst.trasduce WFSTNAME INPUT)\n\ Transduce list INPUT (or exploded INPUT if its an atom) to a list of \n\ outputs. The atom FAILED is return if the transduction fails."); } festival/src/arch/festival/ngram.cc0000644000175000017500000001134010126474330015732 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Authors: Alan W Black */ /* Date : December 1997 */ /*-----------------------------------------------------------------------*/ /* Access to the Ngrammar */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "festivalP.h" static LISP ngram_loaded_list = NIL; static EST_Ngrammar *load_ngram(const EST_String &filename); static LISP add_ngram(const EST_String &name,EST_Ngrammar *n); SIOD_REGISTER_CLASS(ngrammar,EST_Ngrammar) static LISP lisp_load_ngram(LISP name, LISP filename) { EST_Ngrammar *n; n = load_ngram(get_c_string(filename)); add_ngram(get_c_string(name),n); return name; } static EST_Ngrammar *load_ngram(const EST_String &filename) { EST_Ngrammar *n = new EST_Ngrammar(); if (n->load(filename) != 0) { fprintf(stderr,"Ngrammar: failed to read ngrammar from \"%s\"", (const char *)filename); festival_error(); } return n; } static LISP add_ngram(const EST_String &name,EST_Ngrammar *n) { LISP lpair; lpair = siod_assoc_str(name,ngram_loaded_list); if (ngram_loaded_list == NIL) { // First time round so do a little initialization gc_protect(&ngram_loaded_list); } LISP ng = siod(n); if (lpair == NIL) ngram_loaded_list = cons(cons(strintern(name),cons(ng,NIL)),ngram_loaded_list); else { cwarn << "Ngrammar: " << name << " recreated" << endl; setcar(cdr(lpair),ng); } return ng; } EST_Ngrammar *get_ngram(const EST_String &name,const EST_String &filename) { // Find ngram named name, returns NULL if none; LISP lpair; lpair = siod_assoc_str(name,ngram_loaded_list); if (lpair == NIL) { if (filename != EST_String::Empty) { EST_Ngrammar *n = load_ngram(filename); add_ngram(name,n); return n; } else { cwarn << "Ngrammar: no ngram named \"" << name << "\"" << endl; return 0; } } else return ngrammar(car(cdr(lpair))); } void festival_ngram_init() { init_subr_2("ngram.load",lisp_load_ngram, "(ngram.load NAME FILENAME)\n\ Load an ngram from FILENAME and store it named NAME for later access."); } festival/src/arch/festival/viterbi.cc0000644000175000017500000001660011465010356016277 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1999 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Authors: Alan W Black */ /* Date : February 1999 */ /*-----------------------------------------------------------------------*/ /* Generic Viterbi search specifications through scheme */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "lexicon.h" static EST_VTCandidate *gv_candlist(EST_Item *s,EST_Features &f); static EST_VTPath *gv_npath(EST_VTPath *p,EST_VTCandidate *c,EST_Features &f); static double gv_find_wfst_prob(EST_VTPath *p,EST_WFST *wfst, int n,int &state); static double gv_find_ngram_prob(EST_VTPath *p,EST_Ngrammar *ngram, int n,int &state,EST_Features &f); LISP Gen_Viterbi(LISP utt) { // For each syllable predict intonation events. EST_Utterance *u = utterance(utt); LISP params = siod_get_lval("gen_vit_params","no gen_vit_params"); EST_Features f; EST_WFST *wfst = 0; EST_Ngrammar *ngram = 0; int num_states; // Add some defaults f.set("gscale_s",1.0); f.set("gscale_p",0.0); f.set("Relation","Syllable"); f.set("return_feat","gen_vit_val"); lisp_to_features(params,f); if (f.present("ngramname")) { ngram = get_ngram(f.S("ngramname")); num_states = ngram->num_states(); } else { wfst = get_wfst(f.S("wfstname")); num_states = wfst->num_states(); } EST_Viterbi_Decoder v(gv_candlist,gv_npath,num_states); v.f = f; v.initialise(u->relation(f.S("Relation"))); v.search(); v.result("gv_id"); if (f.present("debug")) { v.copy_feature("nprob"); v.copy_feature("prob"); v.copy_feature("score"); v.copy_feature("total_score"); } // Map internal ids back to strings for (EST_Item *p=u->relation(f.S("Relation"))->head(); p != 0; p=p->next()) if (wfst == 0) p->set(f.S("return_feat"),ngram->get_vocab_word(p->I("gv_id"))); else p->set(f.S("return_feat"),wfst->in_symbol(p->I("gv_id"))); return utt; } static EST_VTCandidate *gv_candlist(EST_Item *s,EST_Features &f) { LISP p; LISP l; EST_VTCandidate *c; EST_VTCandidate *all_c = 0; EST_WFST *w = 0; EST_Ngrammar *n = 0; float prob; // Call user function to get candidate probabilities p = leval(cons(rintern(f.S("cand_function")), cons(siod(s),NIL)),NIL); if (f.present("ngramname")) n = get_ngram(f.S("ngramname")); else w = get_wfst(f.S("wfstname")); for (l=p; l != NIL; l=cdr(l)) { prob = get_c_float(car(cdr(car(l)))); if (f.present("debug")) s->set(EST_String("cand_")+get_c_string(car(car(l))),prob); if (prob != 0) { c = new EST_VTCandidate; if (w == 0) c->name = n->get_vocab_word(get_c_string(car(car(l)))); else c->name = w->in_symbol(get_c_string(car(car(l)))); c->score = log(prob); c->s = s; c->next = all_c; all_c = c; } } return all_c; } static EST_VTPath *gv_npath(EST_VTPath *p,EST_VTCandidate *c,EST_Features &f) { EST_VTPath *np = new EST_VTPath; double prob,lprob; EST_WFST *wfst = 0; EST_Ngrammar *ngram = 0; if (f.present("ngramname")) ngram = get_ngram(f.S("ngramname")); else wfst = get_wfst(f.S("wfstname")); np->c = c; np->from = p; int n = c->name.Int(); if (wfst == 0) prob = gv_find_ngram_prob(p,ngram,n,np->state,f); else prob = gv_find_wfst_prob(p,wfst,n,np->state); prob = f.F("gscale_p") + (prob * (1-f.F("gscale_p"))); if (prob == 0) lprob = log(0.00000001); else lprob = log(prob); if (p==0) np->score = (c->score+lprob); else np->score = (c->score+lprob) + p->score; if (f.present("debug")) { np->f.set("prob",prob); np->f.set("score",c->score); np->f.set("nprob",prob*(exp(c->score))); np->f.set("total_score",np->score); } return np; } static double gv_find_wfst_prob(EST_VTPath *p,EST_WFST *wfst, int n,int &state) { float prob; int oldstate; if (p == 0) oldstate = wfst->start_state(); else oldstate = p->state; state = wfst->transition(oldstate,n,n,prob); return prob; } static double gv_find_ngram_prob(EST_VTPath *p,EST_Ngrammar *ngram, int n,int &state,EST_Features &f) { int oldstate=0; double prob; if (p == 0) { // This could be done once before the search is called int order = ngram->order(); int i; EST_IVector window(order); if (order > 1) window.a_no_check(order-1) = n; if (order > 2) window.a_no_check(order-2) = ngram->get_vocab_word(f.S("p_word")); for (i = order-3; i>=0; i--) window.a_no_check(i) = ngram->get_vocab_word(f.S("pp_word")); oldstate = ngram->find_state_id(window); } else oldstate = p->state; state = ngram->find_next_state_id(oldstate,n); const EST_DiscreteProbDistribution &pd = ngram->prob_dist(oldstate); if (pd.samples() == 0) prob = 0; else prob = (double)pd.probability(n); return prob; } festival/src/arch/festival/ModuleDescription.cc0000644000175000017500000001464012442333465020273 0ustar awbawb /************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Richard Caley (rjc@cstr.ed.ac.uk) */ /* Date: Tue Aug 26 1997 */ /* -------------------------------------------------------------------- */ /* Machine readable description of a module. */ /* */ /*************************************************************************/ #include #include "siod.h" #include "ModuleDescription.h" // to make life easier static inline EST_String S(const char *s) { return EST_String(s); } struct ModuleDescription *ModuleDescription::create() { struct ModuleDescription *desc = new ModuleDescription; desc->name = ""; desc->version = 0.0; desc->organisation = ""; desc->author = ""; desc->description[0] = NULL; desc->input_streams[0].name = NULL; desc->optional_streams[0].name = NULL; desc->output_streams[0].name = NULL; desc->parameters[0].name = NULL; return desc; } EST_String ModuleDescription::to_string(const ModuleDescription &desc) { EST_String s; char buf[10]; int i; if (desc.input_streams[0].name || desc.optional_streams[0].name || desc.output_streams[0].name) { s += S("(") + desc.name + S(" UTT"); for(i=0; i init_modules.cc init_modules.cc: Makefile $(TOP)/config/modincludes.inc @ echo "Making init_modules.cc" @ rm -f init_modules.cc @ echo >init_modules.cc @ echo "/* This file is autogenerated by src/modules/Makefile */" >>init_modules.cc @ echo >>init_modules.cc @ echo "#include \"EST_unix.h\"" >>init_modules.cc @ echo "#include \"stdio.h\"" >>init_modules.cc @ echo "#include \"festival.h\"" >>init_modules.cc @ echo >>init_modules.cc @ for i in $(LIB_BUILD_DIRS) $(EXTRA_LIB_BUILD_DIRS) ; \ do \ if [ ! -f "$$i/NoInit" ] ; then \ echo "void festival_"$$i"_init(void);" ; \ fi \ done >> init_modules.cc @ echo >>init_modules.cc @ echo "void festival_init_modules(void)" >>init_modules.cc @ echo "{" >>init_modules.cc @ for i in $(LIB_BUILD_DIRS) $(EXTRA_LIB_BUILD_DIRS) ; \ do \ if [ ! -f "$$i/NoInit" ] ; then \ echo " festival_"$$i"_init();" ; \ fi \ done >> init_modules.cc @ echo "}" >>init_modules.cc @ echo >>init_modules.cc festival/src/modules/VCLocalRules0000755000175000017500000000026507330012566015467 0ustar awbawb# Special extra rules for this directory subtypes.cc : . utilities\find_subtypes.exe @echo "Making subtypes.cc" utilities\find_subtypes $(DIRS) - $(ABSTRACT_TYPES) >subtypes.cc festival/src/modules/Lexicon/Makefile0000644000175000017500000000575007262605650016316 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### # # # Contains implementation of a general lexicon system # # # ########################################################################### TOP=../../.. DIRNAME=src/modules/Lexicon H = lts.h lexiconP.h SRCS = lex_aux.cc lexicon.cc lts.cc lts_rules.cc complex.cc lex_ff.cc OBJS = $(SRCS:.cc=.o) FILES=Makefile $(SRCS) $(H) LOCAL_INCLUDES = -I../include INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/Lexicon/lex_aux.cc0000644000175000017500000001521312102246371016613 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Basic lexicon utilities */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "lexicon.h" #include "lexiconP.h" static void split_stress(LISP phones, LISP &phs, LISP &stresses); static char *v_stress(const char *ph,int &stress); static int syl_contains_vowel(LISP phones); static int syl_breakable(LISP syl, LISP rest); LISP lex_syllabify(LISP phones) { /* Given a simple list of phones, syllabify them and add stress */ LISP syl,syls,p; int stress = 1; for (syl=NIL,syls=NIL,p=phones; p != NIL; p=cdr(p)) { syl = cons(car(p),syl); if (syl_breakable(syl,cdr(p))) { syls = cons(cons(reverse(syl),cons(flocons(stress),NIL)),syls); stress = 0; syl = NIL; } } return reverse(syls); } LISP lex_syllabify_phstress(LISP phones) { /* Given a list of phones where vowels may have stress numeral, */ /* as found in BEEP and CMU syllabify them */ LISP syl,syls,p,phs,stresses,s; int stress = 0; const char *ph; split_stress(phones,phs,stresses); for (syl=NIL,syls=NIL,p=phs,s=stresses; p != NIL; p=cdr(p),s=cdr(s)) { ph = get_c_string(car(p)); if (!streq(ph,ph_silence())) syl = cons(car(p),syl); if (car(s) && (!streq(get_c_string(car(s)),"0"))) stress = 1; // should worry about 2 stress too if (streq(ph,ph_silence()) || syl_breakable(syl,cdr(p))) { syls = cons(cons(reverse(syl),cons(flocons(stress),NIL)),syls); stress = 0; syl = NIL; } } return reverse(syls); } static void split_stress(LISP phones, LISP &phs, LISP &stresses) { // unpack the list of phones. When they come from certain types // of lexical entries (CMU, BEEP) vowels may have a 1 or 2 at their // end to denote stress. // This returns two list of equal length, one with the phones and // one with nils (for each phone) except when there is an explicit // stress number LISP p,np,ns; char *nph; int stress; for (p=phones,np=ns=NIL; p != NIL; p=cdr(p)) { stress = 0; nph = v_stress(get_c_string(car(p)),stress); if (streq(nph,"-")) // a break of some sort np = cons(rintern(ph_silence()),np); else np = cons(rintern(nph),np); wfree(nph); if (stress != 0) ns = cons(flocons(stress),ns); else ns = cons(NIL,ns); } phs = reverse(np); stresses = reverse(ns); } static char *v_stress(const char *ph,int &stress) { // Checks to see if final character is a numeral, if so treats // is as stress value. char *nph; if ((strlen(ph) > 1) && ((ph[strlen(ph)-1] == '1') || (ph[strlen(ph)-1] == '2') || (ph[strlen(ph)-1] == '0'))) { stress = ph[strlen(ph)-1]-'0'; nph = wstrdup(ph); nph[strlen(ph)-1] = '\0'; return nph; } else return wstrdup(ph); } static int syl_breakable(LISP syl, LISP rest) { if (rest == NIL) return TRUE; else if (!syl_contains_vowel(rest)) return FALSE; // must be a vowel remaining in rest else if (syl_contains_vowel(syl)) { if (ph_is_vowel(get_c_string(car(rest)))) return TRUE; else if (cdr(rest) == NIL) return FALSE; int p = ph_sonority(get_c_string(car(syl))); int n = ph_sonority(get_c_string(car(rest))); int nn = ph_sonority(get_c_string(car(cdr(rest)))); if ((p <= n) && (n <= nn)) return TRUE; else return FALSE; } else return FALSE; } static int syl_contains_vowel(LISP phones) { // So we can support "vowels" like ah2, oy2 (i.e. vowels with // stress markings) we need to make this a hack. Vowels are // assumed to start with one of aiueo LISP p; for (p=phones; p !=NIL; p=cdr(p)) if (strchr("aiueoAIUEO",get_c_string(car(p))[0]) != NULL) return TRUE; else if (ph_is_vowel(get_c_string(car(p)))) return TRUE; else if (ph_is_silence(get_c_string(car(p)))) return FALSE; return FALSE; } festival/src/modules/Lexicon/lexicon.cc0000644000175000017500000005623610256005423016620 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* lexicons: lookup */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "lexicon.h" #include "lexiconP.h" #include "lts.h" static int bl_match_entry(LISP entry,const EST_String &word); static int match_features(LISP req_feats, LISP act_feats); // doesn't need to be exact -- used it jumping back chunks when // searching for entries with same head word #define DEFAULT_LEX_ENTRY_SIZE 40 // Depth to build the index cache to #define CACHE_DEPTH 8 static LISP lexicon_list = NIL; static Lexicon *current_lex = NULL; VAL_REGISTER_CLASS(lexicon,Lexicon) SIOD_REGISTER_CLASS(lexicon,Lexicon) Lexicon::Lexicon() { type = lex_external; name = ""; binlexfp = NULL; posmap = NIL; gc_protect(&posmap); addenda = NIL; gc_protect(&addenda); index_cache = NIL; gc_protect(&index_cache); matched_lexical_entries = NIL; gc_protect(&matched_lexical_entries); pre_hooks = NIL; gc_protect(&pre_hooks); post_hooks = NIL; gc_protect(&post_hooks); bl_filename = ""; lts_method=""; } Lexicon::~Lexicon() { if (binlexfp != NULL) fclose(binlexfp); gc_unprotect(&addenda); gc_unprotect(&index_cache); gc_unprotect(&posmap); gc_unprotect(&matched_lexical_entries); gc_unprotect(&pre_hooks); gc_unprotect(&post_hooks); } LISP Lexicon::lookup(const EST_String &word, const LISP features) { LISP entry,mapped,hooked, entry2; EST_String sword; if (pre_hooks != NIL) { // We could just call it and it wont do anything if NIL // but this check may save some extra consing hooked = apply_hooks_right(pre_hooks, cons(strintern(word),cons(features,NIL))); sword = get_c_string(car(hooked)); mapped = map_pos(posmap,car(cdr(hooked))); } else { sword = word; mapped = map_pos(posmap,features); } // Check addenda, them complex, then lts if ((entry = lookup_addenda(sword,mapped)) == NIL) { if ((entry = lookup_complex(sword,mapped)) == NIL) entry = lookup_lts(sword,mapped); } // After adding the city Nice to the addendum, FT said "nice" wrong, there- else if(mapped != NIL) // fore we try to find an entry with matching pos. if(car(cdr(entry)) != NIL && // addendum pos is not NIL mapped != car(cdr(entry))) // and does not match if((entry2 = lookup_complex(sword,mapped)) != NIL) if(mapped == car(cdr(entry2))) // comp. lex. pos does match entry = entry2; if (post_hooks != NIL) { return apply_hooks_right(post_hooks,cons(entry,NIL)); } else return entry; } LISP Lexicon::lookup_all(const EST_String &word) { // Find all entries that match word. LISP entries = NIL; LISP l; for (l=addenda; l != 0; l=cdr(l)) if (bl_match_entry(car(l),word) == 0) entries = cons(car(l),entries); lookup_complex(word,flocons(-1)); return reverse(append(matched_lexical_entries,entries)); } int Lexicon::in_lexicon(const EST_String &word,LISP features) { // Checks to see if this word is in the lexicons (addenda or // compiled lexicon. Ignores any letter to sound method. // This is used to determine if further analysis on this word might // be worthwhile if ((lookup_addenda(word,features) != NIL) || (lookup_complex(word,features) != NIL)) return TRUE; else return FALSE; } EST_String Lexicon::str_lookup(const EST_String &word, const LISP features) { LISP entry = lookup(word,features); return siod_sprint(entry); } LISP Lexicon::lookup_addenda(const EST_String &word,LISP features) { // Lookup addenda (for first match) LISP l,potential=NIL; for (l=addenda; l != 0; l=cdr(l)) if (bl_match_entry(car(l),word) == 0) { if (potential == NIL) // if nothing matches features this'll do potential = car(l); if (match_features(features,(car(cdr(car(l)))))) return car(l); } // If there isn't a complete match, a match of name is sufficient return potential; } LISP Lexicon::lookup_complex(const EST_String &word,LISP features) { // Lookup the word in a compile lexicon file int start,end; if (bl_filename == "") return NIL; // there isn't a compiled lexicon binlex_init(); int depth = 0; matched_lexical_entries = NIL; lex_entry_match = 0; bl_lookup_cache(index_cache,word,start,end,depth); return bl_bsearch(word,features,start,end,depth); } void Lexicon::bl_lookup_cache(LISP cache, const EST_String &word, int &start, int &end, int &depth) { // Look up word in the index cache to get a better idea where // to look in the real file int fc; if (cdr(cache) == NIL) // hit bottom { start = get_c_int(car(car(cache))); end = get_c_int(cdr(car(cache))); } else if ((fc=fcompare(word,get_c_string(car(cdr(cache))),NULL)) < 0) bl_lookup_cache(siod_nth(2,cache),word,start,end,++depth); else if (fc == 0) { start = get_c_int(car(car(cache))); end = get_c_int(cdr(car(cache))); } else bl_lookup_cache(siod_nth(3,cache),word,start,end,++depth); } void Lexicon::add_to_cache(LISP cache, const EST_String &word, int start,int mid, int end) { int fc; if (cdr(cache) == NIL) // hit bottom { LISP a,b; a = cons(cons(flocons(start),flocons(mid)),NIL); b = cons(cons(flocons(mid),flocons(end)),NIL); setcdr(cache,cons(strintern(word),cons(a,cons(b,NIL)))); } else if ((fc = fcompare(word,get_c_string(car(cdr(cache))),NULL)) < 0) add_to_cache(siod_nth(2,cache),word,start,mid,end); else if (fc == 0) return; /* already in cache */ else add_to_cache(siod_nth(3,cache),word,start,mid,end); } void Lexicon::binlex_init(void) { // Open the compiled lexicon file if not already open char magic_number[20]; int end; if (binlexfp == NULL) { if (bl_filename == "") { cerr << "Lexicon: no compile file given" << endl; festival_error(); } binlexfp = fopen(bl_filename,"rb"); if (binlexfp == NULL) { cerr << "Lexicon: compile file \"" << bl_filename << "\" not found or unreadble " << endl; festival_error(); } fread(magic_number,sizeof(char),4,binlexfp); magic_number[4] = '\0'; if ((EST_String)"MNCM" == (EST_String)magic_number) { // A compiled lexicon plus features // Also different entry format (pos distributions) LISP features = lreadf(binlexfp); comp_num_entries = get_param_int("num_entries",features,-1); } else if ((EST_String)"MNCL" == (EST_String)magic_number) { comp_num_entries = -1; } else { cerr << "Lexicon: compile file \"" << bl_filename << "\" not a compiled lexicon " << endl; festival_error(); } blstart = ftell(binlexfp); fseek(binlexfp,0L, SEEK_END); end = ftell(binlexfp); index_cache = cons(cons(flocons(blstart),flocons(end)),NIL); } } LISP Lexicon::lookup_lts(const EST_String &word,LISP features) { // Look up using letter to sound rules if ((lts_method == "") || (lts_method == "Error")) { cerr << "LEXICON: Word " << word << " (plus features) not found in lexicon " << endl; festival_error(); } else if (lts_method == "lts_rules") return lts(word,features,lts_ruleset); else if (lts_method == "none") return cons(strintern(word),cons(NIL,cons(NIL,NIL))); else if (lts_method == "function") return leval(cons(rintern("lex_user_unknown_word"), cons(quote(strintern(word)), cons(quote(features),NIL))), NIL); else return leval(cons(rintern(lts_method), cons(quote(strintern(word)), cons(quote(features),NIL))), NIL); return NIL; } LISP Lexicon::bl_bsearch(const EST_String &word,LISP features, int start,int end, int depth) { // Do a binary search for word in file LISP closest_entry; int mid, compare; if (start==end) // only happens if first item has been tested (and failed) return NIL; else if ((end-start) < 10) { if (start == blstart) { mid = start; end = start; } else return NIL; // failed } else mid = start + (end-start)/2; closest_entry = bl_find_next_entry(mid); if ((depth < CACHE_DEPTH) && (end-start > 256)) { add_to_cache(index_cache,get_c_string(car(closest_entry)), start,mid,end); } compare = bl_match_entry(closest_entry,word); /* printf("%s %s %d %d %d\n", (const char *)word, get_c_string(car(closest_entry)), start,mid,end); */ if (compare == 0) return bl_find_actual_entry(mid,word,features); else if (compare < 0) // too far return bl_bsearch(word,features,start,mid,++depth); else // not far enough return bl_bsearch(word,features,mid,end,++depth); } LISP Lexicon::bl_find_next_entry(int pos) { // Read the next full entry after this point int c; fseek(binlexfp,(long)pos,SEEK_SET); while ((c = getc(binlexfp)) != '\n') if (c == EOF) return NIL; return lreadf(binlexfp); } static int bl_match_entry(LISP entry,const EST_String &word) { return fcompare(word,get_c_string(car(entry)),NULL); } LISP Lexicon::bl_find_actual_entry(int pos,const EST_String &word,LISP features) { // Well there may be a number of entries with the same head word // we want to find the one that matches (with features) or the // the first one. To do this we must go back until we find one that // doesn't match and then search through each for a real match // If no features have a match the first headword match is returned // as that pronunciation is probably still better than letter to // sound rules LISP n_entry; LISP f_entry = NIL; do { pos -= DEFAULT_LEX_ENTRY_SIZE; if (pos < blstart) { pos = blstart; fseek(binlexfp,pos,SEEK_SET); break; } n_entry = bl_find_next_entry(pos); } while (bl_match_entry(n_entry,word) == 0); // now pos is definitely before the first entry whose head is word n_entry = lreadf(binlexfp); lex_entry_match = 0; matched_lexical_entries = NIL; while (bl_match_entry(n_entry,word) >= 0) { if (bl_match_entry(n_entry,word) == 0) { if (f_entry == NIL) f_entry = n_entry; matched_lexical_entries = cons(n_entry,matched_lexical_entries); lex_entry_match++; if (match_features(features,(car(cdr(n_entry))))) return n_entry; } n_entry = lreadf(binlexfp); if (siod_eof(n_entry)) return NIL; } return f_entry; } static int match_features(LISP req_feats, LISP act_feats) { // Match required features with actualy features // required should be less specific than actual LISP l,m; if ((req_feats == NIL) || (eql(req_feats,act_feats))) return TRUE; else if (consp(req_feats) && consp(act_feats)) { for (l=req_feats; l != 0; l=cdr(l)) { for (m=act_feats; m != 0; m=cdr(m)) if (eql(car(l),car(m))) break; if (m == NIL) return FALSE; } return TRUE; } else return FALSE; } static void check_current_lex(void) { // check there is a current lexicon if (current_lex == NULL) { cerr << "No lexicon" << endl; festival_error(); } } static void lex_add_lexicon(const EST_String &name, Lexicon *l) { // Add lexicon to list of lexicons LISP lpair; lpair = siod_assoc_str(name,lexicon_list); if (lexicon_list == NIL) gc_protect(&lexicon_list); if (lpair == NIL) { lexicon_list = cons(cons(strintern(name), cons(siod(l),NIL)), lexicon_list); } else { cwarn << "lexicon " << name << " recreated" << endl; setcar(cdr(lpair),siod(l)); } return; } // // This functions provide the LISP access to the lexicons including // access and selection // static LISP lex_set_compile_file(LISP fname) { EST_String filename = get_c_string(fname); check_current_lex(); current_lex->set_bl_filename(filename); return fname; } static LISP lex_add_entry(LISP entry) { check_current_lex(); current_lex->add_addenda(entry); return NIL; } static LISP lex_set_lts_method(LISP method) { EST_String smethod; check_current_lex(); if (method == NIL) smethod = "none"; else smethod = get_c_string(method); current_lex->set_lts_method(smethod); return method; } static LISP lex_set_lts_ruleset(LISP rulesetname) { EST_String sruleset; check_current_lex(); if (rulesetname == NIL) { cerr << "LEXICON: no ruleset name given\n"; festival_error(); } else sruleset = get_c_string(rulesetname); current_lex->set_lts_ruleset(sruleset); return rulesetname; } static LISP lex_set_pos_map(LISP posmap) { check_current_lex(); current_lex->set_pos_map(posmap); return posmap; } static LISP lex_set_phoneset(LISP psname) { EST_String phoneset = get_c_string(psname); check_current_lex(); current_lex->set_phoneset_name(phoneset); return psname; } static LISP lex_set_pre_hooks(LISP hooks) { LISP last_hooks; check_current_lex(); last_hooks = current_lex->pre_hooks; current_lex->pre_hooks = hooks; return last_hooks; } static LISP lex_set_post_hooks(LISP hooks) { LISP last_hooks; check_current_lex(); last_hooks = current_lex->post_hooks; current_lex->post_hooks = hooks; return last_hooks; } EST_String lex_current_phoneset(void) { check_current_lex(); return current_lex->phoneset_name(); } static LISP lex_create_lex(LISP lexname) { // Make a new lexicon current (and select it) Lexicon *l = new Lexicon; EST_String name = get_c_string(lexname); l->set_lex_name(name); lex_add_lexicon(name,l); current_lex = l; return lexname; } LISP lex_select_lex(LISP lexname) { // Select named lexicon and make it current, return name of previous EST_String name = get_c_string(lexname); LISP lpair, lastname; lpair = siod_assoc_str(name,lexicon_list); if (current_lex == NULL) { cerr << "lexicon: no current lexicon -- shouldn't happen\n"; festival_error(); } else lastname = rintern((const char *)current_lex->get_lex_name()); if (lpair == NIL) { cerr << "lexicon " << name << " not defined" << endl; festival_error(); } else current_lex = lexicon(car(cdr(lpair))); return lastname; } LISP lex_list(void) { // List names of all current defined lexicons LISP lexica = NIL; LISP l; for (l=lexicon_list; l != NIL; l=cdr(l)) lexica = cons(car(car(l)),lexica); return lexica; } static LISP lex_lookup_lisp(LISP lword, LISP features) { return lex_lookup_word(get_c_string(lword),features); } static LISP lex_lookup_all(LISP lword) { check_current_lex(); return current_lex->lookup_all(get_c_string(lword)); } static LISP lex_entrycount(LISP lword) { check_current_lex(); // assumes -1 is never in the pos field current_lex->lookup(get_c_string(lword),flocons(-1)); return flocons(current_lex->num_matches()); } LISP lex_lookup_word(const EST_String &word, LISP features) { check_current_lex(); return current_lex->lookup(word,features); } int in_current_lexicon(const EST_String &word, LISP features) { check_current_lex(); return current_lex->in_lexicon(word,features); } void festival_lex_ff_init(void); void festival_Lexicon_init(void) { // define lexicon related functions festival_lex_ff_init(); init_subr_1("lex.set.compile.file",lex_set_compile_file, "(lex.set.compile.file COMPFILENAME)\n\ Set the current lexicon's compile file to COMPFILENAME. COMPFILENAME\n\ is a compiled lexicon file created by lex.compile.\n\ [see Defining lexicons]"); init_subr_0("lex.list",lex_list, "(lex.list)\n\ List names of all currently defined lexicons."); init_subr_1("lex.set.lts.method",lex_set_lts_method, "(lex.set.lts.method METHOD)\n\ Set the current lexicon's letter-to-sound method to METHOD. METHOD\n\ can take any of the following values: Error (the default) signal a\n\ festival error if a word is not found in the lexicon; lts_rules use the\n\ letter to sound rule set named by lts_ruleset; none return\n\ simply nil in the pronunciation field; function use call the two argument\n\ function lex_user_unknown_word (as set by the user) with the word and\n\ features to provide an entry. [see Letter to sound rules]"); init_subr_1("lex.set.lts.ruleset",lex_set_lts_ruleset, "(lex.set.lts.ruleset RULESETNAME)\n\ Set the current lexicon's letter-to-sound ruleset to RULESETNAME.\n\ A ruleset of that name must already be defined. This is used if\n\ lts.method is set to lts_rules. [see Letter to sound rules]"); init_subr_1("lex.set.pos.map",lex_set_pos_map, "(lex.set.pos.map POSMAP)\n\ A reverse assoc-list mapping part of speech tags to the lexical\n\ part of speech tag set. [see Lexical entries]"); init_subr_1("lex.set.pre_hooks",lex_set_pre_hooks, "(lex.set.pre_hooks HOOKS)\n\ Set a function or list of functions that are to be applied to the entry\n\ before lookup. Returns previous value [see Lexical entries]"); init_subr_1("lex.set.post_hooks",lex_set_post_hooks, "(lex.set.post_hooks HOOKS)\n\ Set a function or list of functions that are to be applied to the entry\n\ after lookup. Returns previous value [see Lexical entries]"); init_subr_1("lex.set.phoneset",lex_set_phoneset, "(lex.set.phoneset PHONESETNAME)\n\ Set current lexicon's phone set to PHONESETNAME. PHONESETNAME must be\n\ a currently defined (and, of course, loaded) phone set.\n\ [see Defining lexicons]"); init_subr_1("lex.add.entry",lex_add_entry, "(lex.add.entry ENTRY)\n\ Add ENTRY to the addenda of the current lexicon. As the addenda is\n\ checked before the compiled lexicon or letter to sound rules, this will\n\ cause ENTRY to be found before all others. If a word already in the\n\ addenda is added again the most recent addition will be found (part of\n\ speech tags are respected in the look up). [see Lexical entries]"); init_subr_1("lex.select",lex_select_lex, "(lex.select LEXNAME)\n\ Select LEXNAME as current lexicon. The name of the previously selected\n\ lexicon is returned."); init_subr_1("lex.create",lex_create_lex, "(lex.create LEXNAME)\n\ Create a new lexicon of name LEXNAME. If it already exists, the old one\n\ is deleted first. [see Defining lexicons]"); init_subr_2("lex.lookup",lex_lookup_lisp, "(lex.lookup WORD FEATURES)\n\ Lookup word in current lexicon. The addenda is checked first, if WORD\n\ with matching FEATURES (so far this is only the part of speech tag) is\n\ not found the compiled lexicon is checked. Only if the word is still not\n\ found the letter to sound rules (or whatever method specified by the\n\ current lexicon's lts.method is used). [see Lookup process]"); init_subr_1("lex.lookup_all",lex_lookup_all, "(lex.lookup_all WORD)\n\ Return list of all entries in the addenda and compiled lexicon that\n\ match this word. The letter to sound rules and user defined unknown\n\ word function is ignored."); init_subr_1("lex.entrycount",lex_entrycount, "(lex.entrycount WORD)\n\ Return the number of entries in the compiled lexicon that match this\n\ word. This is used in detecting homographs."); init_subr_1("lex.syllabify.phstress",lex_syllabify_phstress, "(lex.syllabify.phstress PHONELIST)\n\ Syllabify the given phone list (if current phone set). Vowels may have\n\ the numerals 0, 1, or 2 as suffixes, if so these are taken to be stress\n\ for the syllable they are in. This format is similar to the entry format\n\ in the CMU and BEEP lexicons. [see Defining lexicons]"); init_subr_2("lex.compile",lexicon_compile, "(lex.compile ENTRYFILE COMPILEFILE)\n\ Compile the list of lexical entries in ENTRYFILE into a compiled file in\n\ COMPILEFILE. [see Defining lexicons]"); init_fsubr("lts.ruleset",lts_def_ruleset, "(lts.ruleset NAME RULES SETS)\n\ Define a new set of letter to sound rules. [see Letter to sound rules]"); init_subr_2("lts.apply",lts_apply_ruleset, "(lts.apply WORD RULESETNAME)\n\ Apply lts ruleset RULESETNAME to word returning result. \n\ [see Letter to sound rules]"); init_subr_2("lts.in.alphabet",lts_in_alphabet, "(lts.in.alphabet WORD RULESETNAME)\n\ Returns t is all characters in symbol word (or items in list WORD)\n\ are in the alphabet of letter to sound ruleset name RULESETNAME. nil\n\ otherwise. [see Letter to sound rules]"); init_subr_0("lts.list",lts_list, "(lts.list)\n\ Return list of all current defined LTS rulesets."); } festival/src/modules/Lexicon/lts.cc0000644000175000017500000001077410126474330015761 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* A front end to the letter to sound rule system(s) */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "lexicon.h" #include "lts.h" static LISP lts_create_entry(const EST_String &word, LISP features,LISP syllables); static LISP map_phones(LISP phones); LISP lts(const EST_String &word,LISP features,const EST_String &rulesetname) { /* Return lexical entry for given word in best possible way */ LISP phones; EST_String dword = downcase(word); LISP lword = strintern(dword); LISP lrulesetname = rintern(rulesetname); if (lts_in_alphabet(lword,lrulesetname) != NIL) { // this check doesn't guarantee success phones = lts_apply_ruleset(lword,lrulesetname); } else phones = NIL; // otherwise can't do anything return lts_create_entry(word, features, lex_syllabify(map_phones(phones))); } static LISP lts_create_entry(const EST_String &word,LISP features,LISP syllables) { /* build an entry from information */ return cons(strcons(strlen(word),word), cons(features,cons(syllables,NIL))); } static LISP map_phones(LISP phones) { // map list of phones to lexical list of phones // Users responsibility to get right phone set return phones; // If lts rulesets get their own phonesets then the following should // be used (with appropriate changes) #if 0 LISP mapped = NIL,p; EST_String lexset,mappedph; lexset = lex_current_phoneset(); if (lexset != "nrl") { for (p=phones; p != NIL; p=cdr(p)) { mappedph = map_phone(get_c_string(car(p)),"nrl",lexset); mapped = cons(rintern(mappedph),mapped); } return reverse(mapped); } else return phones; #endif } festival/src/modules/Lexicon/lts_rules.cc0000644000175000017500000003160310126474330017165 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : September 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* A letter to sound rule system that allows rules to be specified */ /* externally. This is specified desined to use the existing Welsh */ /* letter to rules developed by Briony Williams for the Welsh */ /* synthesizer. This form came from a program by Greg Lee (Univ of */ /* Hawaii), but this is not using his code. */ /* */ /* Multiple rule sets are supported alloing multiple applications of */ /* varying rule sets */ /* */ /* A set of rules consists of set definitions, and rules */ /* Each rule consists of a left hand side and a write hand side */ /* The LHS consist of a left context [ change chars ] right context */ /* The RHS consist of the new symbols */ /* */ /* The rules are interpreted in order, the first match rule gives */ /* the replacement, and the search start again with the pointer */ /* incremented */ /* */ /*=======================================================================*/ #include #include #include "festival.h" #include "lts.h" class LTS_Ruleset{ private: EST_String p_name; int num_rules; LISP p_rules; LISP p_alphabet; LISP p_sets; // short hand for sets LISP normalize(LISP rules); int item_match(LISP actual_item, LISP rule_item); int context_match(LISP actual_context, LISP rule_context); int match_rule(LISP lc, LISP remainder, LISP rule, LISP *rest); LISP rewrite(LISP lc, LISP remainder, LISP rules, LISP *rest); LISP this_match(LISP remainder, LISP rule_this); void update_alphabet(LISP newitems); public: LTS_Ruleset(LISP name, LISP rules, LISP sets); ~LTS_Ruleset(void); const EST_String &name(void) const {return p_name;} LISP apply(LISP word); LISP check_alpha(LISP word); }; static LISP fix_postfix_ops(LISP l); static LISP lts_rules_list = NIL; #define LTS_LC(R) (car(R)) #define LTS_THIS(R) (car(cdr(R))) #define LTS_RC(R) (car(cdr(cdr(R)))) #define LTS_RHS(R) (car(cdr(cdr(cdr(R))))) VAL_REGISTER_CLASS(ltsruleset,LTS_Ruleset) SIOD_REGISTER_CLASS(ltsruleset,LTS_Ruleset) LTS_Ruleset::LTS_Ruleset(LISP name, LISP rules, LISP sets) { p_alphabet = NIL; gc_protect(&p_alphabet); p_name = get_c_string(name); p_sets = sets; gc_protect(&p_sets); p_rules = normalize(rules); gc_protect(&p_rules); } LTS_Ruleset::~LTS_Ruleset(void) { gc_unprotect(&p_sets); gc_unprotect(&p_rules); gc_unprotect(&p_alphabet); } LISP LTS_Ruleset::normalize(LISP rules) { // Change the rule format to I can access it faster LISP r, rc, t, lc, rhs, c; LISP nrs = NIL; int state; for (r=rules; r != NIL; r=cdr(r)) { lc = t = rc = rhs = NIL; state = 0; for (c=car(r); c != NIL; c = cdr(c)) { if (state == 0) { if (streq("[",get_c_string(car(c)))) state = 1; else lc = cons(car(c),lc); } else if (state == 1) { if (streq("]",get_c_string(car(c)))) state = 2; else t = cons(car(c),t); } else if (state == 2) { if (streq("=",get_c_string(car(c)))) { state = 3; rhs = cdr(c); break; } else rc = cons(car(c),rc); } else { cerr << "LTS_Rules:: misparsed a rule\n"; cerr << "LTS_Rules:: "; pprint(car(r)); festival_error(); } } update_alphabet(t); if ((state != 3) || (t == NIL)) { cerr << "LTS_Rules:: misparsed a rule\n"; cerr << "LTS_Rules:: "; pprint(car(r)); festival_error(); } nrs = cons(cons(fix_postfix_ops(lc),cons(reverse(t),cons(reverse(rc), cons(rhs,NIL)))),nrs); } return reverse(nrs); } void LTS_Ruleset::update_alphabet(LISP newitems) { // Add new items to alphabet is not already there LISP p; for (p=newitems; p != NIL; p=cdr(p)) if (siod_member_str(get_c_string(car(p)),p_alphabet) == NIL) p_alphabet = cons(car(p),p_alphabet); } static LISP fix_postfix_ops(LISP l) { // This list have been built in reverse so the postfix operators * and + // are wrong. Destrictively fix them LISP p,q; for (p=l; p != NIL; p=cdr(p)) if ((streq("*",get_c_string(car(p)))) || (streq("+",get_c_string(car(p))))) { if (cdr(p) == NIL) { cerr << "LTS_Rules:: malformed left context\n"; pprint(reverse(l)); } q = car(p); CAR(p) = car(cdr(p)); CAR(cdr(p)) = q; p = cdr(p); } return l; } LISP LTS_Ruleset::apply(LISP word) { // Apply rules to word LISP lc,remainder,result,newremainder,r,l; int i; lc = cons(rintern("#"),NIL); remainder = append(word,lc); // add # at end of right context result = NIL; for (; !streq("#",get_c_string(car(remainder))); ) { r = rewrite(lc,remainder,p_rules,&newremainder); result = append(reverse(r),result); for (i=0,l=remainder; i < siod_llength(remainder)-siod_llength(newremainder); i++,l=cdr(l)) lc=cons(car(l),lc); remainder = newremainder; } return reverse(result); } LISP LTS_Ruleset::check_alpha(LISP word) { // Check all characters in word can (possibly) be mapped by this ruleset LISP word_chars,p; if (consp(word)) word_chars = word; else word_chars = symbolexplode(word); for (p=word_chars; p != NIL; p=cdr(p)) if (siod_member_str(get_c_string(car(p)),p_alphabet) == NIL) return NIL; return rintern("t"); } LISP LTS_Ruleset::rewrite(LISP lc, LISP remainder, LISP rules, LISP *rest) { // Find a rule to match this context LISP r,t; for (r=rules; r != NIL; r=cdr(r)) if (match_rule(lc,remainder,car(r),rest) == TRUE) return LTS_RHS(car(r)); cerr << "LTS_Ruleset " << p_name << ": no rule matches: \n"; cerr << "LTS_Ruleset: "; for (t=reverse(lc); t != NIL; t = cdr(t)) cerr << get_c_string(car(t)) << " "; cerr << "*here* "; for (t=remainder; t != NIL; t = cdr(t)) cerr << get_c_string(car(t)) << " "; cerr << endl; festival_error(); return NIL; } int LTS_Ruleset::match_rule(LISP lc, LISP remainder, LISP rule, LISP *rest) { // Match this rule *rest = this_match(remainder,LTS_THIS(rule)); return ((*rest != NIL) && (context_match(*rest,LTS_RC(rule))) && (context_match(lc,LTS_LC(rule)))); } int LTS_Ruleset::context_match(LISP acontext, LISP rcontext) { // TRUE if rule context is initial sub list of actual context if (rcontext == NIL) return TRUE; else if ((cdr(rcontext) != NIL) && (streq("*",get_c_string(car(cdr(rcontext)))))) return ((context_match(acontext,cdr(cdr(rcontext)))) || (context_match(acontext,cons(car(rcontext), cdr(cdr(rcontext))))) || ((item_match(car(acontext),car(rcontext))) && context_match(cdr(acontext),rcontext))); else if ((cdr(rcontext) != NIL) && (streq("+",get_c_string(car(cdr(rcontext)))))) return ((item_match(car(acontext),car(rcontext))) && (context_match(cdr(acontext), cons(car(rcontext), cons(rintern("*"), cdr(cdr(rcontext))))))); else if (item_match(car(acontext),car(rcontext))) return context_match(cdr(acontext),cdr(rcontext)); else return FALSE; #if 0 for (a=actual_context,r=rule_context; r != NIL; a=cdr(a),r=cdr(r)) { if (item_match(car(a),car(r))) return FALSE; } return TRUE; #endif } LISP LTS_Ruleset::this_match(LISP remainder, LISP rule_this) { // Match the centre of the rule to the remainder. Returning // the new remainder if the match is successful LISP a,r; for (a=remainder,r=rule_this; r != NIL; a=cdr(a),r=cdr(r)) if (!item_match(car(a),car(r))) return NIL; return a; } int LTS_Ruleset::item_match(LISP actual_item, LISP rule_item) { // Checks for a match, possible rule_item is a set name // returns remainder if match, NIL otherwise. if (streq(get_c_string(actual_item),get_c_string(rule_item))) return TRUE; else { LISP lpair = assq(rule_item,p_sets); if (lpair == NIL) return FALSE; else if (siod_member_str(get_c_string(actual_item),cdr(lpair)) != NIL) return TRUE; else return FALSE; } } LISP lts_def_ruleset(LISP args, LISP penv) { // Define a new rule set (void)penv; LTS_Ruleset *rs = new LTS_Ruleset(car(args), car(cdr(cdr(args))), car(cdr(args))); LISP name = car(args); LISP lpair; if (lts_rules_list == NIL) gc_protect(<s_rules_list); lpair = siod_assoc_str(get_c_string(name),lts_rules_list); if (lpair == NIL) { lts_rules_list = cons(cons(name, cons(siod(rs),NIL)), lts_rules_list); } else { cwarn << "LTS_Rules: " << get_c_string(name) << " recreated" << endl; setcar(cdr(lpair),siod(rs)); } return name; } LISP lts_in_alphabet(LISP word, LISP rulesetname) { // check if this word is in the input alphabet for ruleset LTS_Ruleset *rs; LISP lpair; lpair = siod_assoc_str(get_c_string(rulesetname),lts_rules_list); if (lpair == NIL) { cerr << "LTS_Rules: no rule set named \"" << get_c_string(rulesetname) << "\"\n"; festival_error(); } else { rs = ltsruleset(car(cdr(lpair))); return rs->check_alpha(word); } return NIL; } LISP lts_list() { // List all currently defined rule sets LISP r,rulesets = NIL; for (r=lts_rules_list; r != NIL; r=cdr(r)) rulesets = cons(car(car(r)),rulesets); return rulesets; } LISP lts_apply_ruleset(LISP word, LISP rulesetname) { // Apply the rule set to word (an atom or list of atoms) LTS_Ruleset *rs; LISP lpair; lpair = siod_assoc_str(get_c_string(rulesetname),lts_rules_list); if (lpair == NIL) { cerr << "LTS_Rule: no rule set named \"" << get_c_string(rulesetname) << "\"\n"; festival_error(); } else { rs = ltsruleset(car(cdr(lpair))); if (consp(word)) return rs->apply(word); else return rs->apply(symbolexplode(word)); } return NIL; } festival/src/modules/Lexicon/complex.cc0000644000175000017500000001703312370701677016633 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : June 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Compile a lexicon from set of exntries */ /* */ /*=======================================================================*/ #include #include #include "festival.h" #include "lexicon.h" #include "lexiconP.h" #include "lts.h" static LISP check_and_fix(LISP entry); static void check_sylphones(const char *name,LISP syls); struct LIST_ent_struct { EST_String word; char *pos; char *entry; LIST_ent_struct *next; }; typedef LIST_ent_struct *LIST_ent; static LISP lex_lts_set = NIL; static LISP lex_syllabification = NIL; int entry_compare(const void *e1, const void *e2) { LIST_ent le1 = *((LIST_ent *)e1); LIST_ent le2 = *((LIST_ent *)e2); int rcode; if ((rcode = fcompare(le1->word,le2->word,0)) != 0) return rcode; else if ((rcode = strcmp(le1->pos,le2->pos)) != 0) return rcode; else if ((rcode = strcmp(le1->word,le2->word)) != 0) return rcode; else { // its a homograph but to ensure its in the same order // not matter which machine we're on return strcmp(le1->entry,le2->entry); } } LISP lexicon_compile(LISP finname, LISP foutname) { // Take a file of entries and process them checking phones // syllabifying if necessary. Sorts them and writes them to // fout FILE *fin, *fout; LISP entry; LIST_ent entries = NULL,e; LIST_ent *ent_list; int num_entries=0,i; EST_String tmpname; if ((fin=fopen(get_c_string(finname),"rb")) == NULL) { cerr << "Lexicon compile: unable to open " << get_c_string(finname) << " for reading\n"; festival_error(); } lex_lts_set = siod_get_lval("lex_lts_set",NULL); lex_syllabification = siod_get_lval("lex_syllabification",NULL); while (!siod_eof((entry = lreadf(fin)))) { e = new LIST_ent_struct; *cdebug << "Processing entry " << get_c_string(car(entry)) << endl; entry = check_and_fix(entry); e->word = get_c_string(car(entry)); e->pos = wstrdup(siod_sprint(car(cdr(entry)))); e->entry = wstrdup(siod_sprint(entry)); e->next = entries; entries = e; num_entries++; } fclose(fin); // Make it into an array for sorting ent_list = new LIST_ent[num_entries]; for (i=0,e=entries; i < num_entries; i++,e=e->next) ent_list[i] = e; qsort(ent_list,num_entries,sizeof(LIST_ent),entry_compare); if ((fout=fopen(get_c_string(foutname),"wb")) == NULL) { cerr << "Lexicon compile: unable to open " << get_c_string(foutname) << " for writing\n"; fclose(fin); festival_error(); } fprintf(fout,"MNCL\n"); for (i=0; i < num_entries; i++) { fprintf(fout,"%s\n",ent_list[i]->entry); wfree(ent_list[i]->pos); wfree(ent_list[i]->entry); delete ent_list[i]; } delete ent_list; fclose(fout); cwarn << "Compiled lexicon \"" << get_c_string(finname) << "\" into \"" << get_c_string(foutname) << "\" " << num_entries << " entries\n"; return NIL; } static LISP check_and_fix(LISP entry) { // Check shape and that phones are in current phone set // Syllabify entry if required LISP syls; if (siod_llength(entry) < 2) { cerr << "Lexicon compile: entry: "; lprint(entry); cerr << "has too few fields\n"; festival_error(); } else if (CONSP(car(entry))) { cerr << "Lexicon compile: entry: "; lprint(entry); cerr << "has non-atomic head word\n"; festival_error(); } // else if (CONSP(car(cdr(entry)))) // The lookup code allows for this so why not allow it here. // { // cerr << "Lexicon compile: entry: "; // lprint(entry); // cerr << "has non-atomic pos field\n"; // festival_error(); // } if ((lex_syllabification == NIL) && (siod_atomic_list(car(cdr(cdr(entry)))))) { // syllabify them (in an old special way) LISP phones = car(cdr(cdr(entry))); if (lex_lts_set != NIL) phones = lts_apply_ruleset(phones,lex_lts_set); syls = lex_syllabify_phstress(phones); check_sylphones(get_c_string(car(entry)),syls); } else if ((lex_syllabification != NIL) && (atomp(lex_syllabification)) && (streq(get_c_string(lex_syllabification),"NONE"))) syls = car(cdr(cdr(entry))); else syls = apply_hooks(lex_syllabification,car(cdr(cdr(entry)))); return cons(car(entry),cons(car(cdr(entry)), cons(syls,cdr(cdr(cdr(entry)))))); } static void check_sylphones(const char *name,LISP syls) { // check shape of syllables, and that phones are valid LISP s,p; for (s=syls; s != NIL; s=cdr(s)) { if (siod_llength(car(s)) != 2) { cerr << "Malformed lexical entry: \"" << name << "\" syllable malformed\n"; festival_error(); } if (!siod_atomic_list(car(car(s)))) { cerr << "Malformed lexical entry: \"" << name << "\" syllable phone list malformed\n"; festival_error(); } for (p=car(car(s)); p != NIL; p=cdr(p)) ; } } festival/src/modules/Lexicon/lex_ff.cc0000644000175000017500000002431411465010356016416 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : May 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* Word based ffeature functions */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "lexiconP.h" static EST_String Phrase("Phrase"); static EST_Val f_content("content"); static EST_Val f_string0("0"); static EST_Val f_string1("1"); static EST_Val ff_word_gpos(EST_Item *s) { /* Part of speech by guessing, returns, prep, det, aux, content */ /* from simple lookup list */ EST_String word; LISP l; LISP guess_pos; word = downcase(s->name()); guess_pos = siod_get_lval("guess_pos","no guess_pos set"); for (l=guess_pos; l != NIL; l=cdr(l)) if (siod_member_str(word,cdr(car(l)))) return EST_Val(get_c_string(car(car(l)))); return f_content; } EST_Val ff_word_contentp(EST_Item *s) { /* 1 if this is a content word, 0 otherwise */ if (ff_word_gpos(s) == "content") return f_string1; else return f_string0; } static EST_Val ff_word_n_content(EST_Item *s) { // returns the next content word after s EST_Item *p; for (p=s->as_relation("Word")->next(); p != 0; p = p->next()) { if (ff_word_gpos(p) == "content") return EST_Val(p->name()); } return f_string0; } static EST_Val ff_word_nn_content(EST_Item *s) { // returns the next next content word after s int count = 0; EST_Item *p; for (p=s->as_relation("Word")->next(); p != 0; p = p->next()) { if (ff_word_gpos(p) == "content") { count ++; if (count == 2) return EST_Val(p->name()); } } return f_string0; } static EST_Val ff_word_p_content(EST_Item *s) { // returns the previous content word after s EST_Item *p; for (p=s->as_relation("Word")->prev(); p != 0; p = p->prev()) if (ff_word_gpos(p) == "content") return EST_Val(p->name()); return f_string0; } static EST_Val ff_word_pp_content(EST_Item *s) { // returns the previous previous content word after s int count = 0; EST_Item *p; for (p=s->as_relation("Word")->prev(); p != 0; p = p->prev()) { if (ff_word_gpos(p) == "content") { count ++; if (count == 2) return EST_Val(p->name()); } } return f_string0; } static EST_Val ff_content_words_out(EST_Item *s) { EST_Item *nn = s->as_relation(Phrase); EST_Item *p; int pos=0; for (p=nn->next(); p; p=p->next()) { if (ff_word_gpos(p) == "content") pos++; } // don't think you can get here return EST_Val(pos); } static EST_Val ff_content_words_in(EST_Item *s) { EST_Item *nn = s->as_relation(Phrase); EST_Item *p; int pos=0; for (p=nn->prev(); p; p=p->prev()) { if (ff_word_gpos(p) == "content") pos++; } // don't think you can get here return EST_Val(pos); } static EST_Val ff_word_cap(EST_Item *s) { // "1" is the word starts with a capital letter const char *word = s->name(); if ((word[0] >= 'A') && (word[0] <='Z')) return f_string1; else return f_string0; } static EST_Val ff_syl_onset_type(EST_Item *s) { // Return van Santen's classification of onset type in to one // of three forms: // -V contains only voiceless consonants // +V-S contains voiced obstruents but no sonorants // +S contains just sonorants EST_Item *nn = s->as_relation("SylStructure"); EST_Item *p; int vox=FALSE; int sonorant=FALSE; for (p=daughter1(nn); p->next() != 0; p=p->next()) { if (ph_is_vowel(p->name())) break; if (ph_is_voiced(p->name())) vox = TRUE; if (ph_is_sonorant(p->name())) sonorant = TRUE; } if (p==daughter1(nn)) // null-onset case return EST_Val("+V-S"); else if (sonorant) return EST_Val("+S"); else if (vox) return EST_Val("+V-S"); else return EST_Val("-V"); } static EST_Val ff_syl_coda_type(EST_Item *s) { // Return van Santen's classification of onset type in to one // of three forms: // -V contains only voiceless consonants // +V-S contains voiced obstruents but no sonorants // +S contains just sonorants EST_Item *nn = s->as_relation("SylStructure"); EST_Item *p; int vox=FALSE; int sonorant=FALSE; for (p=daughter1(nn); p->next() != 0; p=p->next()) { if (ph_is_vowel(p->name())) break; } if (p->next() == 0) // empty coda return EST_Val("+S"); for (p=p->next(); p != 0; p=p->next()) { if (ph_is_voiced(p->name())) vox = TRUE; if (ph_is_sonorant(p->name())) sonorant = TRUE; } if (sonorant) return EST_Val("+S"); else if (vox) return EST_Val("+V-S"); else return EST_Val("-V"); } void festival_lex_ff_init(void) { festival_def_nff("gpos","Word",ff_word_gpos, "Word.gpos\n\ Returns a guess at the part of speech of this word. The lisp a-list\n\ guess_pos is used to load up this word. If no part of speech is\n\ found in there \"content\" is returned. This allows a quick efficient\n\ method for part of speech tagging into closed class and content words."); festival_def_nff("contentp","Word",ff_word_contentp, "Word.contentp\n\ Returns 1 if this word is a content word as defined by gpos, 0 otherwise."); festival_def_nff("cap","Word",ff_word_cap, "Word.cap\n\ Returns 1 if this word starts with a capital letter, 0 otherwise."); festival_def_nff("n_content","Word",ff_word_n_content, "Word.n_content\n\ Next content word. Note this doesn't use the standard n. notation as\n\ it may have to search a number of words forward before finding a\n\ non-function word. Uses gpos to define content/function word distinction.\n\ This also works for Tokens."); festival_def_nff("nn_content","Word",ff_word_nn_content, "Word.nn_content\n\ Next next content word. Note this doesn't use the standard n.n. notation\n\ as it may have to search a number of words forward before finding the \n\ second non-function word. Uses gpos to define content/function word\n\ distinction. This also works for Tokens."); festival_def_nff("p_content","Word",ff_word_p_content, "Word.p_content\n\ Previous content word. Note this doesn't use the standard p. notation\n\ as it may have to search a number of words backward before finding the \n\ first non-function word. Uses gpos to define content/function word\n\ distinction. This also works for Tokens."); festival_def_nff("pp_content","Word",ff_word_pp_content, "Word.pp_content\n\ Previous previous content word. Note this doesn't use the standard p.p.\n\ notation as it may have to search a number of words backward before\n\ finding the first non-function word. Uses gpos to define \n\ content/function word distinction. This also works for Tokens."); festival_def_nff("content_words_out","Word",ff_content_words_out, "Word.content_words_out\n\ Number of content words to end of this phrase."); festival_def_nff("content_words_in","Word",ff_content_words_in, "Word.content_words_in\n\ Number of content words from start this phrase."); festival_def_nff("syl_onset_type","Syllable",ff_syl_onset_type, "Syllable.syl_onset_type\n\ Return the van Santen and Hirschberg classification. -V for unvoiced,\n\ +V-S for voiced but no sonorants, and +S for sonorants."); festival_def_nff("syl_coda_type","Syllable",ff_syl_coda_type, "Syllable.syl_coda_type\n\ Return the van Santen and Hirschberg classification. -V for unvoiced,\n\ +V-S for voiced but no sonorants, and +S for sonorants."); } festival/src/modules/Lexicon/lts.h0000644000175000017500000000630707262605650015630 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* access to letter to sound rule system(s) */ /* */ /*=======================================================================*/ #ifndef __LTS_H__ #define __LTS_H__ LISP lts(const EST_String &word,LISP features,const EST_String &rulesetname); LISP lts_apply_ruleset(LISP word, LISP rulesetname); LISP lts_in_alphabet(LISP word, LISP rulesetname); LISP lts_def_ruleset(LISP args, LISP penv); LISP lts_list(); #endif /* __LTS_H__ */ festival/src/modules/Lexicon/lexiconP.h0000644000175000017500000000602607262605650016605 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : February 1997 */ /*-----------------------------------------------------------------------*/ /* */ /* Private functions for lexicon directory */ /* */ /*=======================================================================*/ #ifndef __LEXICONP_H__ #define __LEXICONP_H__ LISP lexicon_compile(LISP finname, LISP foutname); #endif /* __LEXICONP_H__ */ festival/src/modules/base/Makefile0000644000175000017500000000534507262605650015627 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../../.. DIRNAME=src/modules/base H = TSRCS = SRCS = modules.cc module_support.cc parameters.cc ff.cc \ pos.cc phrasify.cc word.cc postlex.cc phrinfo.cc $(TSRCS) OBJS = $(SRCS:.cc=.o) FILES = Makefile $(SRCS) $(H) LOCAL_INCLUDES = -I../include INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/base/modules.cc0000644000175000017500000002004410126474330016127 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Some basic initialization functions for modules */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "lexicon.h" #include "modules.h" #include "intonation.h" static void create_words(EST_Utterance *u); static void create_segments(EST_Utterance *u); static void create_wave(EST_Utterance *u); static void create_phones(EST_Utterance *u); LISP FT_Initialize_Utt(LISP utt) { // Main utterance intialization routine // creates appropriate streams and loads them from the input EST_Utterance *u = get_c_utt(utt); EST_String type; *cdebug << "Initialize module\n"; type = utt_type(*u); utt_cleanup(*u); // delete all relations if (type == "Words") create_words(u); else if (type == "Text") ; else if (type == "Segments") create_segments(u); else if (type == "Phones") create_phones(u); else if (type == "Phrase") create_phraseinput(u); else if (type == "Wave") create_wave(u); else { // error cerr << "Unknown utterance type \"" << type << "\" for initialization " << endl; festival_error(); } return utt; } void create_words(EST_Utterance *u) { // Add words from IForm LISP lwords,w; EST_Item *word; u->create_relation("Word"); lwords = utt_iform(*u); for (w=lwords; w != NIL; w=cdr(w)) { if (consp(car(w))) // word has features too { word = add_word(u,get_c_string(car(car(w)))); add_item_features(word,car(cdr(car(w)))); } else add_word(u,get_c_string(car(w))); } } void create_wave(EST_Utterance *u) { // Get the fname for the wave and load it EST_Item *item = 0; LISP lwave; EST_Wave *wave = new EST_Wave; lwave = utt_iform(*u); if (wave->load(get_c_string(lwave)) != format_ok) { cerr << "Cannot load wavefile: " << get_c_string(lwave) << endl; festival_error(); } item = u->create_relation("Wave")->append(); item->set_val("wave",est_val(wave)); } void create_segments(EST_Utterance *u) { // Add segments from IForm LISP lsegs,s,targs,t; EST_String seg; EST_Item *Seg;; float start,end,dur,tpos,tval; u->create_relation("Segment"); u->create_relation("Target"); lsegs = utt_iform(*u); end = 0.0; for (s=lsegs; s != NIL; s=cdr(s)) { seg = get_c_string(car(car(s))); dur = get_c_float(car(cdr(car(s)))); targs = cdr(cdr(car(s))); Seg = add_segment(u,seg); start = end; end += dur; Seg->set("end",end); for (t=targs; t != NIL; t=cdr(t)) { tpos = start + (get_c_float(car(car(t)))); tval = get_c_float(car(cdr(car(t)))); add_target(u,Seg,tpos,tval); } } } static void create_phones(EST_Utterance *u) { // Add phones from IForm LISP lsegs,s; EST_String seg; u->create_relation("Segment"); lsegs = utt_iform(*u); for (s=lsegs; s != NIL; s=cdr(s)) { seg = get_c_string(car(s)); add_segment(u,seg); } } LISP FT_Initialize_Utt(LISP args); LISP FT_Classic_Phrasify_Utt(LISP args); LISP FT_Classic_Word_Utt(LISP args); LISP FT_Unilex_Word_Utt(LISP args); LISP FT_Classic_POS_Utt(LISP args); LISP FT_PostLex_Utt(LISP utt); void festival_ff_init(void); void festival_base_init(void) { // Thing I haven't put anywhere else yet festival_ff_init(); // basic feature functions // Basic EST_Utterance modules festival_def_utt_module("Initialize",FT_Initialize_Utt, "(Initialize UTT)\n\ This module should be called first on all utterances it does some\n\ necessary initialization of the utterance and loads the base\n\ streams with the information from the input form."); festival_def_utt_module("Classic_Phrasify",FT_Classic_Phrasify_Utt, "(Classic_Phrasify UTT)\n\ Creates phrases from words, if pos_supported is non-nil, a more elaborate\n\ system of prediction is used. Here probability models based on part of\n\ speech and B/NB distribution are used to predict breaks. This system\n\ uses standard Viterbi decoding techniques. If pos_supported is nil,\n\ a simple CART-based prediction model is used. [see Phrase breaks]"); festival_def_utt_module("Classic_Word",FT_Classic_Word_Utt, "(Classic_Word UTT)\n\ Build the syllable/segment/SylStructure from the given words using the\n\ Lexicon. Uses part of speech information in the lexicon look up if\n\ present."); festival_def_utt_module("Unilex_Word",FT_Unilex_Word_Utt, "(Unilex_Word UTT)\n\ Build the syllable/segment/SylStructure from the given words using the\n\ Lexicon. Uses part of speech information in the lexicon look up if\n\ present."); festival_def_utt_module("Classic_POS",FT_Classic_POS_Utt, "(Classic_POS UTT)\n\ Predict part of speech tags for the existing word stream. If the variable\n\ pos_lex_name is nil nothing happens, otherwise it is assumed to point to\n\ a lexicon file giving part of speech distribution for words. An ngram\n\ model file should be in pos_ngram_name. The system uses standard\n\ Viterbi decoding techniques. [see POS tagging]"); festival_def_utt_module("Builtin_PostLex",FT_PostLex_Utt, "(Builtin_PostLex UTT)\n\ Post-lexical rules. Currently only vowel reduction applied to each\n\ syllable using postlex_vowel_reduce_cart_tree, and the table of \n\ vowel reduction pairs in postlex_vowel_reduce_table."); } festival/src/modules/base/module_support.cc0000644000175000017500000001701207262605650017551 0ustar awbawb /************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Richard Caley (rjc@cstr.ed.ac.uk) */ /* Date: Tue Jul 29 1997 */ /* -------------------------------------------------------------------- */ /* Some things useful in modules. */ /* */ /*************************************************************************/ #include "module_support.h" #define CAR6(x) CAR(CDR5(x)) #define CDR6(x) CDR(CDR5(x)) #define CAR7(x) CAR(CDR6(x)) #define CDR7(x) CDR(CDR6(x)) #define CDR_to1(X) ((X!=NIL)&&CDR1(X)) #define CDR_to2(X) (CDR_to1(X)&&CDR2(X)) #define CDR_to3(X) (CDR_to2(X)&&CDR3(X)) #define CDR_to4(X) (CDR_to3(X)&&CDR4(X)) #define CDR_to5(X) (CDR_to4(X)&&CDR5(X)) #define CDR_to6(X) (CDR_to5(X)&&CDR6(X)) #define CDR_to7(X) (CDR_to6(X)&&CDR7(X)) void unpack_multiple_args(LISP args, LISP &v1, LISP &v2, LISP &v3, LISP &v4) { if (args) { v1 = CAR1(args); if (CDR1(args)) { v2 = CAR2(args); if (CDR2(args)) { v3 = CAR3(args); if (CDR3(args)) v4 = CAR4(args); } } } } void unpack_multiple_args(LISP args, LISP &v1, LISP &v2, LISP &v3, LISP &v4, LISP &v5) { unpack_multiple_args(args, v1, v2, v3, v4); if (CDR4(args)) v5 = CAR5(args); } void unpack_relation_arg(EST_Utterance *utt, LISP lrel_name, EST_String &relation_name, EST_Relation *&relation, RelArgType type) { if (lrel_name) relation_name = get_c_string(lrel_name); if(utt->relation(relation_name)) relation = utt->relation(relation_name); if (type==sat_existing) { if(!relation) err("no relation", relation_name); } else if (type==sat_new || type==sat_replace) { if (relation) if (type==sat_new) err("relation exists", relation_name); utt->create_relation(relation_name); relation = &(*(utt->relation(relation_name))); } } void unpack_module_args(LISP args, EST_Utterance *&utt) { if (args) { LISP lutt = CAR1(args); utt = get_c_utt(lutt); return; } err("no utterance given", NIL); } void unpack_module_args(LISP args, EST_Utterance *&utt, EST_String &relation1_name, EST_Relation *&relation1, RelArgType type1) { unpack_module_args(args, utt); unpack_relation_arg(utt, CDR_to1(args)?CAR2(args):NIL, relation1_name, relation1, type1); } void unpack_module_args(LISP args, EST_Utterance *&utt, EST_String &relation1_name, EST_Relation *&relation1, RelArgType type1, EST_String &relation2_name, EST_Relation *&relation2, RelArgType type2 ) { unpack_module_args(args, utt); unpack_relation_arg(utt, CDR_to1(args)?CAR2(args):NIL, relation1_name, relation1, type1); unpack_relation_arg(utt, CDR_to2(args)?CAR3(args):NIL, relation2_name, relation2, type2); } void unpack_module_args(LISP args, EST_Utterance *&utt, EST_String &relation1_name, EST_Relation *&relation1, RelArgType type1, EST_String &relation2_name, EST_Relation *&relation2, RelArgType type2, EST_String &relation3_name, EST_Relation *&relation3, RelArgType type3 ) { unpack_module_args(args, utt); unpack_relation_arg(utt, CDR_to1(args)?CAR2(args):NIL, relation1_name, relation1, type1); unpack_relation_arg(utt, CDR_to2(args)?CAR3(args):NIL, relation2_name, relation2, type2); unpack_relation_arg(utt, CDR_to3(args)?CAR4(args):NIL, relation3_name, relation3, type3); } void unpack_module_args(LISP args, EST_Utterance *&utt, EST_String &relation1_name, EST_Relation *&relation1, RelArgType type1, EST_String &relation2_name, EST_Relation *&relation2, RelArgType type2, EST_String &relation3_name, EST_Relation *&relation3, RelArgType type3, EST_String &relation4_name, EST_Relation *&relation4, RelArgType type4 ) { unpack_module_args(args, utt); unpack_relation_arg(utt, CDR_to1(args)?CAR2(args):NIL, relation1_name, relation1, type1); unpack_relation_arg(utt, CDR_to2(args)?CAR3(args):NIL, relation2_name, relation2, type2); unpack_relation_arg(utt, CDR_to3(args)?CAR4(args):NIL, relation3_name, relation3, type3); unpack_relation_arg(utt, CDR_to4(args)?CAR5(args):NIL, relation4_name, relation4, type4); } void unpack_module_args(LISP args, EST_Utterance *&utt, EST_String &relation1_name, EST_Relation *&relation1, RelArgType type1, EST_String &relation2_name, EST_Relation *&relation2, RelArgType type2, EST_String &relation3_name, EST_Relation *&relation3, RelArgType type3, EST_String &relation4_name, EST_Relation *&relation4, RelArgType type4, EST_String &relation5_name, EST_Relation *&relation5, RelArgType type5 ) { unpack_module_args(args, utt); unpack_relation_arg(utt, CDR_to1(args)?CAR2(args):NIL, relation1_name, relation1, type1); unpack_relation_arg(utt, CDR_to2(args)?CAR3(args):NIL, relation2_name, relation2, type2); unpack_relation_arg(utt, CDR_to3(args)?CAR4(args):NIL, relation3_name, relation3, type3); unpack_relation_arg(utt, CDR_to4(args)?CAR5(args):NIL, relation4_name, relation4, type4); unpack_relation_arg(utt, CDR_to5(args)?CAR6(args):NIL, relation5_name, relation5, type5); } festival/src/modules/base/parameters.cc0000644000175000017500000001136407262605650016637 0ustar awbawb /************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Richard Caley (rjc@cstr.ed.ac.uk) */ /* Date: Tue Aug 26 1997 */ /* ------------------------------------------------------------------- */ /* Utility routines for easy access to parameters from C++ modules. */ /* */ /*************************************************************************/ #include "module_support.h" // implemented as a call to scheme so that redefineing how parameters // are accessed in scheme will affect things as required. Of course // this isn't as efficient as it might be, but we aren't going to be // doing this inside any loops LISP lisp_parameter_get(const EST_String parameter_name) { LISP parameter_get = siod_get_lval("Parameter.get", "Parameter.get not defined"); LISP parameter = rintern(parameter_name); LISP sexp = cons(parameter_get, cons(quote(parameter), NIL)); LISP val=NIL; gc_protect(&sexp); CATCH_ERRORS() { cerr << "error getting parameter " << parameter_name << "\n"; siod_reset_prompt(); gc_unprotect(&sexp); return NIL; } val = leval(sexp, NIL); END_CATCH_ERRORS(); gc_unprotect(&sexp); return val; } int int_parameter_get(const EST_String parameter, int def) { LISP lval = lisp_parameter_get(parameter); if (lval == NIL) return def; if (!FLONUMP(lval)) { cerr << "non numeric value for parameter " << parameter << "\n"; return 0; } return get_c_int(lval); } float float_parameter_get(const EST_String parameter, float def) { LISP lval = lisp_parameter_get(parameter); if (lval == NIL) return def; if (!FLONUMP(lval)) { cerr << "non numeric value for parameter " << parameter << "\n"; return 0.0; } return get_c_float(lval); } bool bool_parameter_get(const EST_String parameter) { LISP lval = lisp_parameter_get(parameter); return lval != NIL; } EST_String string_parameter_get(const EST_String parameter, EST_String def) { LISP lval = lisp_parameter_get(parameter); if (lval == NIL) return def; if (!SYMBOLP(lval) && !STRINGP(lval)) { cerr << "non string value for parameter " << parameter << "\n"; return 0; } return get_c_string(lval); } festival/src/modules/base/ff.cc0000644000175000017500000007776712442332411015076 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : May 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* Basic builtin features */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "modules.h" static EST_String stressname("stress"); static EST_Val val_string0("0"); static EST_Val val_string1("1"); static EST_Val val_int0(0); static EST_Val val_int1(1); static EST_Val default_val_float(0.0); static EST_Val ff_addr(EST_Item *i) { char a[1024]; // The address of the contents so that the same item from different views // have the same address sprintf(a,"%p",i->contents()); return EST_Val(a); } static EST_Val ff_segment_duration(EST_Item *s) { EST_Item *n = as(s,"Segment"); if (n == 0) { cerr << "Asked for segment duration of item not in Segment relation." << endl; festival_error(); } if (n->prev() == 0) return EST_Val(s->F("end", 0)); else return EST_Val(s->F("end", 0)-(n->prev()->F("end",0))); } static EST_Val ff_syllable_duration(EST_Item *s) { EST_Item *n = as(s,"SylStructure"); if (n == 0) { cerr << "Asked for syllable duration of item not in SylStructure relation." << endl; festival_error(); } else { EST_Item *fd = daughter1(n); EST_Item *ld = fd->last(); if (ld == 0) return val_int0; EST_Item *ps = as(fd,"Segment")->prev(); if (ps == 0) return ld->F("end",0); else return EST_Val(ld->F("end",0)-ps->F("end",0)); } // dummy for stupid VC++ compiler { EST_Val junk; return junk; } } static EST_Val ff_word_duration(EST_Item *s) { EST_Item *n = as(s,"SylStructure"); if (n == 0) { cerr << "Asked for word duration of item not in SylStructure relation." << endl; festival_error(); } else { EST_Item *fd = daughter1(daughter1(n)); EST_Item *ld = daughtern(daughtern(n)); if (ld == 0) return val_int0; EST_Item *ps = as(fd,"Segment")->prev(); if (ps == 0) return ld->F("end",0); else return EST_Val(ld->F("end",0)-ps->F("end",0)); } // dummy for stupid VC++ compiler { EST_Val junk; return junk; } } static EST_Val ff_seg_start(EST_Item *s) { EST_Item *n = as(s,"Segment"); if (n->prev() == 0) return default_val_float; else return n->prev()->F("end",0); } static EST_Val ff_syl_start(EST_Item *s) { EST_Item *n = as(s,"SylStructure"); if (daughter1(n) == 0) return default_val_float; else return ff_seg_start(daughter1(n)); } static EST_Val ff_word_start(EST_Item *s) { EST_Item *n = as(s,"SylStructure"); if (daughter1(daughter1(n)) == 0) return default_val_float; else return ff_seg_start(daughter1(daughter1(n))); } static EST_Val ff_seg_end(EST_Item *s) { return s->F("end",0); } static EST_Val ff_seg_mid(EST_Item *s) { return EST_Val(((float)ff_seg_start(s)+(float)ff_seg_end(s))/2.0); } static EST_Val ff_syl_end(EST_Item *s) { EST_Item *n = as(s,"SylStructure"); if (daughtern(n) == 0) return default_val_float; else return ff_seg_end(daughtern(n)); } static EST_Val ff_word_end(EST_Item *s) { EST_Item *n = as(s,"SylStructure"); if (daughtern(n) == 0) return default_val_float; else return ff_syl_end(daughtern(n)); } static EST_Val ff_position_type(EST_Item *s) { /* Position of syllable in this word: initial, mod, final */ EST_Item *nn = as(s,"SylStructure"); if (nn == 0) // its not really a syllable return EST_Val("single"); else if (nn->next() == 0) { if (nn->prev() == 0) return EST_Val("single"); else return EST_Val("final"); } else if (nn->prev() == 0) return EST_Val("initial"); else return EST_Val("mid"); } static EST_Val ff_word_break(EST_Item *w) { /* Break index of word */ EST_Item *nn = as(w,"Phrase"); static EST_Val val4 = EST_Val(4); static EST_Val val3 = EST_Val(3); static EST_Val val2 = EST_Val(2); if ((nn == 0) || (nn->next() != 0)) return val_int1; else { EST_Item *p = parent(nn); if (p) { if (p->name() == "BB") return val4; else if (p->name() == "B") return val3; else if (p->name() == "mB") return val2; else return EST_Val(p->name()); } else return val_int1; } } static EST_Val ff_syl_break(EST_Item *s) { // 0 internal syl end, 1 word end, 4 phrase end (ToBI 3 and 4) EST_Item *nn = as(s,"SylStructure"); if (nn == 0) return val_int1; // no sylstructure so maybe its standalone else if (nn->next() != 0) // word internal return val_int0; else if (parent(nn) == 0) // not in a word -- strange return val_int1; else return ff_word_break(parent(nn)); // take it from the word } static EST_Val ff_old_syl_break(EST_Item *s) { // 0 internal syl end, 1 word end, 4 phrase end (ToBI 3 and 4) // 2's and threes are promoted to 4s EST_Item *nn = as(s,"SylStructure"); static EST_Val val4 = EST_Val(4); if (nn == 0) return val_int1; // no sylstructure so maybe its standalone else if (nn->next() != 0) // word internal return val_int0; else if (parent(nn) == 0) // not in a word -- strange return val_int1; else { EST_Val v = ff_word_break(parent(nn)); if ((v == 3) || (v == 2)) return val4; else return v; } } static EST_Val ff_syl_accented(EST_Item *s) { // t if syllable is accented or not EST_Item *nn = as(s,"Intonation"); if ((nn == 0) || (daughter1(nn) == 0)) return val_int0; else return val_int1; } static EST_Val ff_tobi_accent(EST_Item *s) { // First tobi accent related to syllable EST_Item *nn = as(s,"Intonation"); EST_Item *p; for (p=daughter1(nn); p; p=p->next()) if (p->name().contains("*")) return EST_Val(p->name()); return EST_Val("NONE"); } static EST_Val ff_tobi_endtone(EST_Item *s) { // First tobi endtone (phrase accent or boundary tone) EST_Item *nn = as(s,"Intonation"); EST_Item *p; for (p=daughter1(nn); p; p=p->next()) { EST_String l = p->name(); if ((l.contains("%")) || (l.contains("-"))) return EST_Val(p->name()); } return EST_Val("NONE"); } static EST_Val ff_syl_accent(EST_Item *s) { // (first) accent or NONE on given syllable EST_Item *nn = as(s,"Intonation"); if (daughter2(nn)) return EST_Val("multi"); else if (daughter1(nn)) return EST_Val(daughter1(nn)->name()); else return EST_Val("NONE"); } static EST_Val ff_syl_numphones(EST_Item *s) { // Number of phones in syllable EST_Item *nn = as(s,"SylStructure"); return EST_Val(daughter1(nn)->length()); } static EST_Val ff_word_numsyls(EST_Item *s) { // Number of syllable in word EST_Item *nn = as(s,"SylStructure"); return EST_Val(daughter1(nn)->length()); } static EST_Val ff_syl_onsetsize(EST_Item *s) { // number of segments in the onset EST_Item *nn = as(s,"SylStructure"); EST_Item *p; int size; for (p=daughter1(nn),size=0; p; p=p->next(),size++) if (ph_is_vowel(p->name())) return EST_Val(size); return EST_Val(size); } static EST_Val ff_syl_vowel(EST_Item *s) { // the vowel in the syllable EST_Item *nn = as(s,"SylStructure"); EST_Item *p; int size; for (p=daughter1(nn),size=0; p; p=p->next(),size++) if (ph_is_vowel(p->name())) return EST_Val(p->name()); // no vowel return EST_Val("novowel"); } static EST_Val ff_syl_codasize(EST_Item *s) { // number of segments in the coda EST_Item *nn = as(s,"SylStructure"); EST_Item *p; int size; for (p=daughter1(nn)->last(),size=1; p; p=p->prev(),size++) if (ph_is_vowel(p->name())) return EST_Val(size); return EST_Val(size); } static EST_Val ff_syl_pc_unvox(EST_Item *s) { // Returns percentage of syllable from start to first voiced phone EST_Item *nn = as(s,"SylStructure"); EST_Item *p,*ps; float unvox,start = 0; if (daughter1(nn) == 0) return val_int0; // no segments in syllable else if ((ps = as(daughter1(nn),"Segment")->prev()) != 0) start = ps->F("end",0); unvox = start; for (p=daughter1(nn); p != 0; p=p->next()) { if ((ph_is_vowel(p->name())) || (ph_is_voiced(p->name()))) break; unvox = p->F("end",0); } return EST_Val((int)(((unvox-start)*100)/ (daughtern(nn)->F("end",0)-start))); } static EST_Val ff_syl_vowel_start(EST_Item *s) { // Returns start time of vowel in syllable (or start of syllable) EST_Item *nn = as(s,"SylStructure"); EST_Item *p; for (p=daughter1(nn); p != 0; p=p->next()) { if (ph_is_vowel(p->name())) return EST_Val(ff_seg_start(p)); } // There isn't a vowel, so just take start of syl return EST_Val(ff_syl_start(p)); } static EST_Val ff_seg_onsetcoda(EST_Item *s) { // onset if seg in onset, coda otherwise (vowel is in coda) EST_Item *nn = as(s,"SylStructure"); EST_Item *p; for (p=nn->next(); p; p=p->next()) if (ph_is_vowel(p->name())) return EST_Val("onset"); return EST_Val("coda"); } static EST_Val ff_seg_onset_stop(EST_Item *s) { // 1 if onset of the syllable attached to this segment has a stop EST_Item *nn = as(s,"SylStructure")->first(); for ( ; nn ; nn=nn->next()) { if (ph_is_vowel(nn->name())) return val_string0; if (ph_is_stop(nn->name())) return val_string1; } return val_string0; } static EST_Val ff_seg_coda_fric(EST_Item *s) { // 1 if coda of the syllable attached to this segment has a fricative EST_Item *nn = as(s,"SylStructure")->last(); for ( ; nn ; nn=nn->prev()) { if (ph_is_vowel(nn->name())) return val_string0; if (ph_is_fricative(nn->name())) return val_string1; } return val_string0; } static EST_Val ff_seg_pos_in_syl(EST_Item *s) { // position of segment in syllable EST_Item *nn = as(s,"SylStructure"); EST_Item *p; int pos=0; for (p=nn->first(); p; p=p->next(),pos++) if (p == nn) return EST_Val(pos); // don't think you can get here return EST_Val(pos); } static EST_Val ff_seg_syl_initial(EST_Item *s) { // 1 if seg is syllable initial, 0 otherwise. EST_Item *nn = as(s,"SylStructure"); if (nn->prev() == 0) return val_string1; else return val_string0; } static EST_Val ff_seg_syl_final(EST_Item *s) { // 1 if seg is syllable initial, 0 otherwise. EST_Item *nn = as(s,"SylStructure"); if (nn->next() == 0) return val_string1; else return val_string0; } static EST_Val ff_syl_pos_in_word(EST_Item *s) { // position of syllable in word EST_Item *nn = as(s,"SylStructure"); EST_Item *p; int pos=0; for (p=nn->first(); p; p=p->next(),pos++) if (p == nn) return EST_Val(pos); // don't think you can get here return EST_Val(pos); } static EST_Val ff_pos_in_phrase(EST_Item *s) { // position of word in phrase EST_Item *nn = as(s,"Phrase"); EST_Item *p; int pos=0; for (p=nn->first(); p; p=p->next(),pos++) if (p == nn) return EST_Val(pos); // don't think you can get here return EST_Val(pos); } static EST_Val ff_num_break(EST_Item *s) { // 1 if this word is at the end of a number group and followed by // a new number group EST_Item *nn = as(s,"Token"); if ((nn->next() == 0) && (parent(nn)->name().matches(RXdouble)) && (parent(nn)->next()->name().matches(RXdouble))) return val_string1; else return val_string0; } static EST_Val ff_words_out(EST_Item *s) { return EST_Val(as(s,"Phrase")->length()); } static EST_Val ff_syl_midpitch(EST_Item *s) { // pitch of mid vowel in syllable EST_Item *nn = as(s,"SylStructure"); EST_Item *p; for (p=daughter1(nn); p; p = p->next()) { if (ph_is_vowel(p->name())) return ffeature(p,"R:Target.daughter1.f0"); } // must be a silence or a syllabic consonant return default_val_float; } static EST_Val ff_syl_startpitch(EST_Item *s) { // pitch at start of syllable // average of first segment and previous segment target (if exists) float pt = ffeature(s,"R:SylStructure.daughter1.R:Segment.p.R:Target.daughter1.f0"); float tt = ffeature(s,"R:SylStructure.daughter1.R:Segment.R:Target.daughter1.f0"); if (pt < 0.1) return EST_Val(tt); else if (tt < 0.1) return EST_Val(pt); else return EST_Val((tt+pt)/2.0); } static EST_Val ff_syl_endpitch(EST_Item *s) { // pitch at start of syllable // average of first segment and previous segment target (if exists) float nt = ffeature(s,"R:SylStructure.daughtern.R:Segment.n.R:Target.daughter1.f0"); float tt = ffeature(s,"R:SylStructure.daughtern.R:Segment.R:Target.daughter1.f0"); if (nt < 0.1) return EST_Val(tt); else if (tt < 0.1) return EST_Val(nt); else return EST_Val((tt+nt)/2.0); } static EST_Val ff_seg_pitch(EST_Item *s) { // Return interpolated pitch at mid-point of s EST_Item *t,*lastt; float spoint,deltaf0,deltatime; float smid = ff_seg_mid(s); EST_Utterance *u = get_utt(s); for (lastt=t=u->relation("Target")->first_leaf(); next_leaf(t) != 0; t=next_leaf(t)) { if (smid <= t->F("pos",0)) break; lastt=t; } if (lastt == 0) return EST_Val((float)0.0); deltaf0 = t->F("f0",0)-lastt->F("f0",0); deltatime = t->F("pos",0) - lastt->F("pos",0); if (deltatime <= 0) spoint = lastt->F("f0",0); else spoint = lastt->F("f0",0) + (deltaf0*((smid-lastt->F("pos",0))/deltatime)); if (spoint > 35) return EST_Val(spoint); else return EST_Val((float)0.0); } static EST_Val ff_syl_in(EST_Item *s) { // Number of syllables to since last phrase break EST_Item *nn = as(s,"Syllable"); // The first syllable in the phrase EST_Item *fsyl = as(daughter1(as(parent(s,"SylStructure"),"Phrase")->first(),"SylStructure"), "Syllable"); EST_Item *p; int count; for (count=0,p=nn; p != 0; p=p->prev(),count++) if (p == fsyl) return EST_Val(count); return EST_Val(count); } static EST_Val ff_syl_out(EST_Item *s) { // Number of syllables since last phrase break EST_Item *nn = as(s,"Syllable"); // The last syllable in the phrase EST_Item *lsyl = as(daughtern(as(parent(s,"SylStructure"),"Phrase")->last(),"SylStructure"), "Syllable"); EST_Item *p; int count; for (count=0,p=nn; p != 0; p=p->next(),count++) if (p == lsyl) return EST_Val(count); return EST_Val(count); } static EST_Val ff_ssyl_in(EST_Item *s) { // Number of stressed syllables since last phrase break EST_Item *nn = as(s,"Syllable"); EST_Item *fsyl = as(daughter1(as(parent(s,"SylStructure"),"Phrase")->first(),"SylStructure"), "Syllable"); EST_Item *p; int count; if (nn == fsyl) return val_int0; for (count=0,p=nn->prev(); (p != 0) && (p != fsyl); p = p->prev()) if (p->F(stressname,0) == 1) count ++; return EST_Val(count); } // This function is modified version of ff_ssyl_in(). // If first syllable in the current phrase is stressed syllable, // ff_ssyl_in() does not count it as stressed syllable. // In the following function, this problem is fixed. static EST_Val ff_ssyl_in_modified_version(EST_Item *s) { // Number of stressed syllables since last phrase break EST_Item *nn = as(s,"Syllable"); EST_Item *fsyl = as(daughter1(as(parent(s,"SylStructure"),"Phrase")->first(),"SylStructure"), "Syllable"); EST_Item *p; int count; if (nn == fsyl) return val_int0; for (count=0,p=nn->prev(); (p != 0); p = p->prev()) { if (p->F(stressname,0) == 1) count ++; if (p == fsyl) break; } return EST_Val(count); } static EST_Val ff_ssyl_out(EST_Item *s) { // Number of stressed syllables to next phrase break EST_Item *nn = as(s,"Syllable"); // The last syllable in the phrase EST_Item *lsyl = as(daughtern(as(parent(s,"SylStructure"),"Phrase")->last(),"SylStructure"), "Syllable"); EST_Item *p; int count; if (nn == lsyl) return val_int0; for (count=0,p=nn->next(); (p != 0); p=p->next()) { if (p->F(stressname,0) == 1) count ++; if (p == lsyl) break; } return EST_Val(count); } static EST_Val ff_asyl_in(EST_Item *s) { // Number of accented syllables since last phrase break EST_Item *nn = as(s,"Syllable"); // The first syllable in the phrase EST_Item *fsyl = as(daughter1(as(parent(s,"SylStructure"),"Phrase")->first(),"SylStructure"), "Syllable"); EST_Item *p; int count; if (nn == fsyl) return val_int0; for (count=0,p=nn->prev(); (p != 0) && (p != fsyl); p = p->prev()) if (ff_syl_accented(p) == 1) count ++; return EST_Val(count); } // This function is modified version of ff_asyl_in(). // If first syllable in the current phrase is accented syllable, // ff_asyl_in() does not count it as accented syllable. // In the following function, this problem is fixed. static EST_Val ff_asyl_in_modified_version(EST_Item *s) { // Number of accented syllables since last phrase break EST_Item *nn = as(s,"Syllable"); // The first syllable in the phrase EST_Item *fsyl = as(daughter1(as(parent(s,"SylStructure"),"Phrase")->first(),"SylStructure"), "Syllable"); EST_Item *p; int count; if (nn == fsyl) return val_int0; for (count=0,p=nn->prev(); (p != 0); p = p->prev()) { if (ff_syl_accented(p) == 1) count ++; if (p == fsyl) break; } return EST_Val(count); } static EST_Val ff_asyl_out(EST_Item *s) { // Number of accented syllables to next phrase break EST_Item *nn = as(s,"Syllable"); // The last syllable in the phrase EST_Item *lsyl = as(daughtern(as(parent(s,"SylStructure"),"Phrase")->last(),"SylStructure"), "Syllable"); EST_Item *p; int count; if (nn == lsyl) return val_int0; for (count=0,p=nn->next(); (p != 0); p=p->next()) { if (ff_syl_accented(p) == 1) count ++; if (p == lsyl) break; } return EST_Val(count); } static EST_Val ff_last_accent(EST_Item *s) { // Number of syllables since last accented syllable EST_Item *nn = as(s,"Syllable"); EST_Item *p; int count; for (count=0,p=nn->prev(); p != 0; p=p->prev(),count++) if (ff_syl_accented(p) == 1) return EST_Val(count); return EST_Val(count); } static EST_Val ff_next_accent(EST_Item *s) { // Number of syllables to next accented syllable EST_Item *nn = as(s,"Syllable"); EST_Item *p; int count; for (count=0,p=nn->next(); p != 0; p=p->next(),count++) if (ff_syl_accented(p) == 1) return EST_Val(count); return EST_Val(count); } static EST_Val ff_sub_phrases(EST_Item *s) { // Number of non-major phrase breaks since last major phrase break EST_Item *nn = parent(parent(s,"SylStructure"),"Phrase"); EST_Item *p; int count; for (count=0,p=nn->prev(); p != 0; p=p->prev()) { if (p->name() == "BB") return EST_Val(count); count ++; } return EST_Val(count); } void festival_ff_init(void) { festival_def_nff("segment_duration","Segment",ff_segment_duration, "Segment.segment_duration\n\ The duration of the given stream item calculated as the end of this\n\ item minus the end of the previous item in the Segment relation."); festival_def_nff("syllable_duration","Syllable",ff_syllable_duration, "Syllable.syllable_duration\n\ The duration of the given stream item calculated as the end of last\n\ daughter minus the end of previous item in the Segment relation of the\n\ first duaghter."); festival_def_nff("word_duration","Word",ff_word_duration, "Word.word_duration\n\ The duration of the given stream item. This is defined as the end of\n\ last segment in the last syllable (via the SylStructure relation) minus\n\ the segment immediate preceding the first segment in the first syllable."); festival_def_nff("segment_start","Segment",ff_seg_start, "Segement.segment_start\n\ The start time of the given segment."); festival_def_nff("segment_mid","Segment",ff_seg_mid, "Segement.segment_mid\n\ The middle time of the given segment."); festival_def_nff("syllable_start","Syllable",ff_syl_start, "Syllable.syllable_start\n\ The start time of the given syllable."); festival_def_nff("word_start","Word",ff_word_start, "Word.word_start\n\ The start time of the given word."); festival_def_nff("segment_end","Segment",ff_seg_end, "Segment.segment_end\n\ The end time of the given segment."); festival_def_nff("syllable_end","Syllable",ff_syl_end, "Syllable.syllable_end\n\ The end time of the given syllable."); festival_def_nff("word_end","Word",ff_word_end, "Word.word_end\n\ The end time of the given word."); festival_def_nff("addr","ANY",ff_addr, "ANY.addr\n\ Returned by popular demand, returns the address of given item that\n\ is guaranteed unique for this session."); festival_def_nff("accented","Syllable",ff_syl_accented, "Syllable.accented\n\ Returns 1 if syllable is accented, 0 otherwise. A syllable is\n\ accented if there is at least one IntEvent related to it."); festival_def_nff("syl_accent","Syllable",ff_syl_accent, "Syllable.syl_accent\n\ Returns the name of the accent related to the syllable. NONE is returned\n\ if there are no accents, and multi is returned if there is more than one."); festival_def_nff("tobi_accent","Syllable",ff_tobi_accent, "Syllable.tobi_accent\n\ Returns the ToBI accent related to syllable. ToBI accents are\n\ those which contain a *. NONE is returned if there are none. If\n\ there is more than one ToBI accent related to this syllable the\n\ first one is returned."); festival_def_nff("tobi_endtone","Syllable",ff_tobi_endtone, "Syllable.tobi_endtone\n\ Returns the ToBI endtone related to syllable. ToBI end tones are\n\ those IntEvent labels which contain a % or a - (i.e. end tones or\n\ phrase accents). NONE is returned if there are none. If\n\ there is more than one ToBI end tone related to this syllable the\n\ first one is returned."); festival_def_nff("syl_onsetsize","Syllable",ff_syl_onsetsize, "Syllable.syl_onsetsize\n\ Returns the number of segments before the vowel in this syllable. If\n\ there is no vowel in the syllable this will return the total number\n\ of segments in the syllable."); festival_def_nff("syl_vowel","Syllable",ff_syl_vowel, "Syllable.syl_vowel\n\ Returns the name of the vowel within this syllable. Note this is not\n\ the general form you probably want. You can't refer to ph_* features \n\ of this. Returns \"novowel\" is no vowel can be found."); festival_def_nff("syl_codasize","Syllable",ff_syl_codasize, "Syllable.syl_codasize\n\ Returns the number of segments after the vowel in this syllable. If\n\ there is no vowel in the syllable this will return the total number\n\ of segments in the syllable." ); festival_def_nff("seg_onsetcoda","Segment",ff_seg_onsetcoda, "Segment.seg_onsetcoda\n\ Returns onset if this segment is before the vowel in the syllable it\n\ is contained within. Returns coda if it is the vowel or after. If\n\ the segment is not in a syllable it returns onset."); festival_def_nff("seg_onset_stop","Segment",ff_seg_onset_stop, "Segment.seg_onset_stop\n\ Returns 1 if onset of the syllable this segment is in contains a stop.\n\ 0 otherwise."); festival_def_nff("seg_coda_fric","Segment",ff_seg_coda_fric, "Segment.seg_coda_fric\n\ Returns 1 if coda of the syllable this segment is in contains a fricative.\n\ 0 otherwise."); festival_def_nff("syl_numphones","Syllable",ff_syl_numphones, "Syllable.syl_numphones\n\ Returns number of phones in syllable."); festival_def_nff("syl_pc_unvox","Syllable",ff_syl_pc_unvox, "Syllable.syl_pc_unvox\n\ Percentage of total duration of unvoiced segments from\n\ start of syllable. (i.e. percentage to start of first voiced segment)"); festival_def_nff("syl_vowel_start","Syllable",ff_syl_vowel_start, "Syllable.syl_vowel_start\n\ Start position of vowel in syllable. If there is no vowel the start\n\ position of the syllable is returned."); festival_def_nff("syl_midpitch","Syllable",ff_syl_midpitch, "Syllable.syl_midpitch\n\ Pitch at the mid vowel of this syllable."); festival_def_nff("syl_startpitch","Syllable",ff_syl_startpitch, "Syllable.syl_startpitch\n\ Pitch at the start of this syllable."); festival_def_nff("syl_endpitch","Syllable",ff_syl_endpitch, "Syllable.syl_endpitch\n\ Pitch at the end of this syllable."); festival_def_nff("seg_pitch","Segment",ff_seg_pitch, "Segment.seg_pitch\n\ Pitch at the middle of this segment."); festival_def_nff("syl_in","Syllable",ff_syl_in, "Syllable.syl_in\n\ Returns number of syllables since last phrase break. This is 0 if\n\ this syllable is phrase initial."); festival_def_nff("syl_out","Syllable",ff_syl_out, "Syllable.syl_out\n\ Returns number of syllables to next phrase break. This is 0 if\n\ this syllable is phrase final."); festival_def_nff("ssyl_in","Syllable",ff_ssyl_in, "Syllable.ssyl_in\n\ Returns number of stressed syllables since last phrase break, not\n\ including this one."); festival_def_nff("ssyl_in_modified_version","Syllable",ff_ssyl_in_modified_version, "Syllable.ssyl_in_modified_version\n\ Returns number of stressed syllables since last phrase break, not\n\ including this one."); festival_def_nff("ssyl_out","Syllable",ff_ssyl_out, "Syllable.ssyl_out\n\ Returns number of stressed syllables to next phrase break, not including\n\ this one."); festival_def_nff("asyl_in","Syllable",ff_asyl_in, "Syllable.asyl_in\n\ Returns number of accented syllables since last phrase break, not\n\ including this one. Accentedness is as defined by the syl_accented\n\ feature."); festival_def_nff("asyl_in_modified_version","Syllable",ff_asyl_in_modified_version, "Syllable.asyl_in_modified_version\n\ Returns number of accented syllables since last phrase break, not\n\ including this one. Accentedness is as defined by the syl_accented\n\ feature."); festival_def_nff("asyl_out","Syllable",ff_asyl_out, "Syllable.asyl_out\n\ Returns number of accented syllables to the next phrase break, not\n\ including this one. Accentedness is as defined by the syl_accented\n\ feature."); festival_def_nff("last_accent","Syllable",ff_last_accent, "Syllable.last_accent\n\ Returns the number of syllables since last accented syllable."); festival_def_nff("next_accent","Syllable",ff_next_accent, "Syllable.next_accent\n\ Returns the number of syllables to the next accented syllable."); festival_def_nff("sub_phrases","Syllable",ff_sub_phrases, "Syllable.sub_phrases\n\ Returns the number of non-major phrase breaks since last major\n\ phrase break. Major phrase breaks are 4, as returned by syl_break,\n\ minor phrase breaks are 2 and 3."); festival_def_nff("syl_break","Syllable",ff_syl_break, "Syllable.syl_break\n\ The break level after this syllable. Word internal is syllables\n\ return 0, non phrase final words return 1. Final syllables in \n\ phrase final words return the name of the phrase they are related to.\n\ Note the occasional \"-\" that may appear of phrase names is removed\n\ so that this feature function returns a number in the range 0,1,2,3,4."); festival_def_nff("old_syl_break","Syllable",ff_old_syl_break, "Syllable.old_syl_break\n\ Like syl_break but 2 and 3 are promoted to 4 (to be compatible with\n\ some older models."); festival_def_nff("pos_in_syl","Segment",ff_seg_pos_in_syl, "Segment.pos_in_syl\n\ The position of this segment in the syllable it is related to. The index\n\ counts from 0. If this segment is not related to a syllable this \n\ returns 0."); festival_def_nff("syl_initial","Segment",ff_seg_syl_initial, "Segment.syl_initial\n\ Returns 1 if this segment is the first segment in the syllable it\n\ is related to, or if it is not related to any syllable."); festival_def_nff("syl_final","Segment",ff_seg_syl_final, "Segment.syl_final\n\ Returns 1 if this segment is the last segment in the syllable it\n\ is related to, or if it is not related to any syllable."); festival_def_nff("pos_in_word","Syllable",ff_syl_pos_in_word, "Syllable.pos_in_word\n\ The position of this syllable in the word it is related to. The index\n\ counts from 0. If this syllable is not related to a word then 0 is\n\ returned."); festival_def_nff("word_numsyls","Word",ff_word_numsyls, "Word.word_numsyls\n\ Returns number of syllables in a word."); festival_def_nff("word_break","Word",ff_word_break, "Word.word_break\n\ The break level after this word. Non-phrase final words return 1\n\ Phrase final words return the name of the phrase they are in."); festival_def_nff("pos_in_phrase","Word",ff_pos_in_phrase, "Word.pos_in_phrase\n\ The position of this word in the phrase this word is in."); festival_def_nff("num_break","Word",ff_num_break, "Word.num_break\n\ 1 if this is the last word in a numeric token and it is followed by\n\ a numeric token."); festival_def_nff("words_out","Word",ff_words_out, "Word.words_out\n\ Number of words to end of this phrase."); festival_def_nff("position_type","Syllable",ff_position_type, "Syllable.position_type\n\ The type of syllable with respect to the word it it related to. This\n\ may be any of: single for single syllable words, initial for word\n\ initial syllables in a poly-syllabic word, final for word final\n\ syllables in poly-syllabic words, and mid for syllables within \n\ poly-syllabic words."); } festival/src/modules/base/pos.cc0000644000175000017500000001702011465010357015262 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : August 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Various part-of-speech predciting modules */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "lexicon.h" static EST_VTCandidate *pos_candlist(EST_Item *s,EST_Features &f); static EST_VTPath *pos_npath(EST_VTPath *p,EST_VTCandidate *c,EST_Features &f); static double find_np_prob(EST_VTPath *p,int n,int *state); static EST_Ngrammar *pos_ngram = 0; static EST_String zeroString("0"); static int p_word = 0; // arbitrary numbers static int n_word = 1; LISP FT_Classic_POS_Utt(LISP utt) { // Predict part of speech for word stream EST_Utterance *u = get_c_utt(utt); LISP pos_lex_name, pos_ngram_name; LISP lastlex, pos_p_start_tag, pos_pp_start_tag; *cdebug << "Classic POS module\n"; pos_lex_name = siod_get_lval("pos_lex_name",NULL); if (pos_lex_name == NIL) return utt; // not set so ignore it pos_ngram_name = siod_get_lval("pos_ngram_name","no pos ngram name"); pos_p_start_tag = siod_get_lval("pos_p_start_tag","no prev start tag"); pos_pp_start_tag = siod_get_lval("pos_pp_start_tag","no prev prev start tag"); lastlex = lex_select_lex(pos_lex_name); if ((pos_ngram = get_ngram(get_c_string(pos_ngram_name))) == 0) { cerr << "POS: no ngram called \"" << get_c_string(pos_ngram_name) << "\" defined" << endl; festival_error(); } p_word = pos_ngram->get_vocab_word(get_c_string(pos_p_start_tag)); n_word = pos_ngram->get_vocab_word(get_c_string(pos_pp_start_tag)); EST_Viterbi_Decoder v(pos_candlist,pos_npath,pos_ngram->num_states()); v.initialise(u->relation("Word")); v.search(); v.result("pos_index"); lex_select_lex(lastlex); EST_Item *w; EST_String pos; LISP l; // Map pos tagset to desired set LISP pos_map = siod_get_lval("pos_map",NULL); for (w=u->relation("Word")->first(); w != 0; w = w->next()) { // convert pos index into string value pos = pos_ngram->get_vocab_word(w->f("pos_index").Int()); w->set("pos",pos); for (l=pos_map; l != NIL; l=cdr(l)) if (siod_member_str(pos,car(car(l))) != NIL) { w->set("pos",get_c_string(car(cdr(car(l))))); break; } } return utt; } static EST_VTCandidate *pos_candlist(EST_Item *s,EST_Features &f) { // Return list of possible pos based on a priori probabilities LISP pd,l; EST_Item *token; EST_VTCandidate *c; EST_VTCandidate *all_c = 0; EST_String actual_pos; (void)f; if (((actual_pos = s->S("pos","0")) != "0") || (((token = parent(s,"Token")) != 0) && ((actual_pos = token->S("pos","0")) != "0"))) { // There is an explicit pos specified, so respect it pd = cons(make_param_float(actual_pos,1.0),NIL); c = new EST_VTCandidate; c->name = pos_ngram->get_vocab_word(actual_pos); c->score = 1.0; c->s = s; c->next = 0; return c; } LISP e = lex_lookup_word(s->name(),NIL); pd = car(cdr(e)); if (pd == NIL) { const char *chr = s->name(); if (strchr("0123456789",chr[0]) != NULL) e = lex_lookup_word("_number_",NIL); // I *know* there is an entry else e = lex_lookup_word("_OOV_",NIL); // I *know* there is an entry pd = car(cdr(e)); } // Build a candidate for each entry in prob distribution for (l=pd; l != NIL; l=cdr(l)) { c = new EST_VTCandidate; c->name = pos_ngram->get_vocab_word(get_c_string(car(car(l)))); c->score = get_c_float(car(cdr(car(l)))); c->s = s; c->next = all_c; all_c = c; } return all_c; } static EST_VTPath *pos_npath(EST_VTPath *p,EST_VTCandidate *c,EST_Features &f) { // Build a potential new path from previous path and this candidate EST_VTPath *np = new EST_VTPath; // static EST_String lscorename("lscore"); double prob; double lprob; (void)f; np->c = c; np->from = p; int n = c->name.Int(); prob = find_np_prob(p,n,&np->state); if (prob == 0) lprob = log(0.00000001); else lprob = log(prob); // np->set_feature(lscorename,(c->score+lprob)); if (p==0) np->score = (c->score+lprob); else np->score = (c->score+lprob) + p->score; return np; } static double find_np_prob(EST_VTPath *p,int n,int *state) { int oldstate=0; if (p==0) { // This could be done once before the search is called int order = pos_ngram->order(); EST_IVector window(order); int i; window.a_no_check(order-1) = n; window.a_no_check(order-2) = p_word; for (i = order-3; i>=0; i--) window.a_no_check(i) = n_word; oldstate = pos_ngram->find_state_id(window); } else oldstate = p->state; *state = pos_ngram->find_next_state_id(oldstate,n); const EST_DiscreteProbDistribution &pd = pos_ngram->prob_dist(oldstate); if (pd.samples() == 0) return 0; else return (double)pd.probability(n); } festival/src/modules/base/phrasify.cc0000644000175000017500000005767012442333465016331 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : August 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Phrase break prediction */ /* */ /* Modified Jul 2011: aup added support for CART phrasing (phrasyn) */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "modules.h" static void phrasing_none(EST_Utterance *u); static void phrasing_by_cart(EST_Utterance *u); static void phrasing_by_probmodels(EST_Utterance *u); static void phrasing_by_fa(EST_Utterance *u); static void phrasing_by_cart_viterbi(EST_Utterance *u); // Addition by AUP static void phrasing_by_cart_probmodels_combined(EST_Utterance *u); // Addition by AUP static EST_VTCandidate *cart_bb_candlist(EST_Item *s, EST_Features &f); // Addition by AUP static EST_VTCandidate *bb_candlist(EST_Item *s,EST_Features &f); static EST_VTPath *bb_npath(EST_VTPath *p,EST_VTCandidate *c,EST_Features &f); static double find_b_prob(EST_VTPath *p,int n,int *state); // Used in various default value cases static int B_word = 0; static int BB_word = 0; static int NB_word = 0; static int pos_p_start_tag = 0; static int pos_pp_start_tag = 0; static int pos_n_start_tag = 0; static double gscale_s = 1.0; static double gscale_p = 0.0; static EST_Ngrammar *bb_ngram = 0; static EST_Ngrammar *bb_pos_ngram = 0; static LISP bb_tags = NIL; static LISP pos_map = NIL; static LISP phrase_type_tree = NIL; static LISP bb_unigrams = NIL; /* Interslice */ static EST_Track *bb_track = 0; LISP FT_Classic_Phrasify_Utt(LISP utt) { // Predict and add phrasing to an utterance EST_Utterance *u = get_c_utt(utt); LISP phrase_method = ft_get_param("Phrase_Method"); *cdebug << "Phrasify module\n"; *cdebug << "Using method: " << get_c_string(phrase_method) << endl; if (u->relation_present("Phrase")) return utt; // already specified else if (phrase_method == NIL) phrasing_none(u); // all one phrase else if (streq("prob_models",get_c_string(phrase_method))) phrasing_by_probmodels(u); else if (streq("prob_cart_combined", get_c_string(phrase_method))) phrasing_by_cart_probmodels_combined(u); else if (streq("cart_tree",get_c_string(phrase_method))) phrasing_by_cart(u); else if (streq("forced_align",get_c_string(phrase_method))) phrasing_by_fa(u); else { cerr << "PHRASIFY: unknown phrase method \"" << get_c_string(phrase_method) << endl; festival_error(); } return utt; } static void phrasing_none(EST_Utterance *u) { // All in a single phrase EST_Item *w,*phr=0; u->create_relation("Phrase"); for (w=u->relation("Word")->first(); w != 0; w = w->next()) { if (phr == 0) phr = add_phrase(u); append_daughter(phr,"Phrase",w); if (w->next() == 0) { w->set("pbreak","B"); phr->set_name("4"); phr = 0; } } } static void phrasing_by_cart(EST_Utterance *u) { EST_Item *w,*phr=0; LISP tree; EST_Val pbreak; u->create_relation("Phrase"); tree = siod_get_lval("phrase_cart_tree","no phrase cart tree"); for (w=u->relation("Word")->first(); w != 0; w = w->next()) { if (phr == 0) phr = add_phrase(u); append_daughter(phr,"Phrase",w); pbreak = wagon_predict(w,tree); w->set("pbreak",pbreak.string()); if ((pbreak == "B") || (pbreak == "BB")) { phr->set_name((EST_String)pbreak); phr = 0; } } } static void pbyp_get_params(LISP params) { EST_String bb_pos_name,bb_pos_filename,bb_name,bb_break_filename; EST_String bb_track_name; LISP l1; bb_pos_name = get_param_str("pos_ngram_name",params,""); bb_pos_filename = get_param_str("pos_ngram_filename",params,""); if ((bb_pos_ngram = get_ngram(bb_pos_name,bb_pos_filename)) == 0) { cerr << "PHRASIFY: no ngram called \"" << bb_pos_name << "\" defined." << endl; festival_error(); } gscale_s = get_param_float("gram_scale_s",params,1.0); gscale_p = get_param_float("gram_scale_p",params,0.0); pos_map = get_param_lisp("pos_map",params,NIL); bb_name = get_param_str("break_ngram_name",params,""); bb_break_filename = get_param_str("break_ngram_filename",params,""); *cdebug << "File: "<load(bb_track_name) != format_ok) { delete bb_track; cerr << "PHRASE: failed to load FA track " << bb_track_name << endl; festival_error(); } } l1 = siod_get_lval("pos_p_start_tag",NULL); if (l1 != NIL) pos_p_start_tag = bb_pos_ngram->get_vocab_word(get_c_string(l1)); l1 = siod_get_lval("pos_pp_start_tag",NULL); if (l1 != NIL) pos_pp_start_tag = bb_pos_ngram->get_vocab_word(get_c_string(l1)); l1 = siod_get_lval("pos_n_start_tag",NULL); if (l1 != NIL) pos_n_start_tag = bb_pos_ngram->get_vocab_word(get_c_string(l1)); } static void phrasing_by_probmodels(EST_Utterance *u) { // Predict phrasing using POS and prob models of B distribution EST_Item *w,*phr=0; EST_String pbreak; int num_states; pbyp_get_params(siod_get_lval("phr_break_params",NULL)); gc_protect(&bb_tags); for (w=u->relation("Word")->first(); w != 0; w = w->next()) { // Set up tag index for pos ngram EST_String lpos = map_pos(pos_map,w->f("pos").string()); w->set("phr_pos",lpos); w->set("pos_index",bb_pos_ngram->get_vocab_word(lpos)); } B_word = bb_ngram->get_vocab_word("B"); NB_word = bb_ngram->get_vocab_word("NB"); BB_word = bb_ngram->get_vocab_word("BB"); num_states = bb_ngram->num_states(); EST_Viterbi_Decoder v(bb_candlist,bb_npath,num_states); v.initialise(u->relation("Word")); v.search(); v.result("pbreak_index"); // Given predicted break, go through and add phrases u->create_relation("Phrase"); for (w=u->relation("Word")->first(); w != 0; w = w->next()) { w->set("pbreak",bb_ngram-> get_vocab_word(w->f("pbreak_index").Int())); if (phr == 0) phr = add_phrase(u); append_daughter(phr,"Phrase",w); if (phrase_type_tree != NIL) { EST_Val npbreak = wagon_predict(w,phrase_type_tree); w->set("pbreak",npbreak.string()); // may reset to BB } pbreak = (EST_String)w->f("pbreak"); if (pbreak == "B") w->set("blevel",3); else if (pbreak == "mB") w->set("blevel",2); if ((pbreak == "B") || (pbreak == "BB") || (pbreak == "mB")) { phr->set_name((EST_String)pbreak); phr = 0; } } gc_unprotect(&bb_tags); bb_tags = NIL; } static void phrasing_by_cart_viterbi(EST_Utterance *u) { EST_Item *w, *phr = 0; LISP tree; LISP answer; int num_states; EST_String pbreak; *cdebug << "Inside phrasing_by_cart_viterbi"<get_vocab_word("B"); NB_word = bb_ngram->get_vocab_word("NB"); BB_word = bb_ngram->get_vocab_word("BB"); *cdebug << "States: "<num_states()<order()<num_states(); EST_Viterbi_Decoder v(cart_bb_candlist,bb_npath,num_states); v.initialise(u->relation("Word")); v.search(); int x = v.result("pbreak_index"); *cdebug << "Viterbi result ended in "<create_relation("Phrase"); for (w=u->relation("Word")->first(); w != 0; w = w->next()) { w->set("pbreak",bb_ngram-> get_vocab_word(w->f("pbreak_index").Int())); if (phr == 0) phr = add_phrase(u); append_daughter(phr,"Phrase",w); if (phrase_type_tree != NIL) { EST_Val npbreak = wagon_predict(w,phrase_type_tree); w->set("pbreak",npbreak.string()); // may reset to BB *cdebug << "Reset: " << npbreak << endl; ; } pbreak = (EST_String)w->f("pbreak"); if (pbreak == "B") w->set("blevel",3); else if (pbreak == "mB") w->set("blevel",2); if ((pbreak == "B") || (pbreak == "BB") || (pbreak == "mB")) { phr->set_name((EST_String)pbreak); phr = 0; } } gc_unprotect(&bb_tags); bb_tags = NIL; } static EST_VTCandidate *cart_bb_candlist(EST_Item *s, EST_Features &f) { double prob=1.0; double divisor = 1.0; EST_VTCandidate *all_c = 0; EST_VTCandidate *c; LISP l; LISP m; LISP answer, tree; tree = siod_get_lval("phrase_cart_tree", "no phrase cart tree"); answer = wagon_pd(s, car(tree)); //*cdebug << get_c_string(answer) <next() == 0) // end of utterances so force a break { EST_VTCandidate *c = new EST_VTCandidate; c->s = s; c->name = B_word; c->score = log(0.95); // very very likely, but not absolute c->next = all_c; all_c = c; return all_c; } m = bb_unigrams; for (l=bb_tags; l !=0; l=cdr(l)) { if(m == NIL) divisor = 1.0; else { divisor = get_c_float(car(m)); m = cdr(m); } c = new EST_VTCandidate; c->s = s; c->name = bb_ngram->get_vocab_word(get_c_string(car(l))); prob = get_param_float(get_c_string(car(l)), answer, 0.001); if (prob == 0) prob = 0.0000001; if (prob == 1) prob = 0.9999999; // Divide by unigram probability to get "Reverse" score //*cdebug << "Prob: "<score = log(prob) - log(divisor); s->set("phrase_score", c->score); c->next = all_c; all_c = c; } return all_c; } static void phrasing_by_cart_probmodels_combined(EST_Utterance *u) { *cdebug << "Using Phrasing Method: prob_cart_combined"<order()); (void)f; int tag; if (bb_pos_ngram->order() == 4) { window[1] = s->I("pos_index",0); if (s->prev() != 0) window[0] = s->prev()->I("pos_index",0); else window[0] = pos_p_start_tag; if (s->next() != 0) window[2] = s->next()->I("pos_index",0); else window[2] = pos_n_start_tag; *cdebug << window[0] << " " << window[1] << " " << window[2] << " " ; } else if (bb_pos_ngram->order() == 3) { window[0] = s->I("pos_index",0); if (s->next() != 0) window[1] = s->next()->I("pos_index",0); else window[1] = pos_n_start_tag; } else if (bb_pos_ngram->order() == 5) { // This is specific for some set of pos tagsets window[2] = s->I("pos_index",0); if (s->prev() != 0) { window[1] = s->prev()->I("pos_index",0); } else { window[1] = pos_p_start_tag; } if (s->next() != 0) { window[3] = s->next()->I("pos_index",0); if (s->next()->next() != 0) window[0] = s->next()->next()->I("pos_index",0); else window[0] = 0; } else { window[3] = pos_n_start_tag; window[0] = 0; } } else { cerr << "PHRASIFY: can't deal with ngram of size " << bb_pos_ngram->order() << endl; festival_error(); } double prob=1.0; EST_VTCandidate *all_c = 0; EST_Val labelled_brk = ffeature(s,"R:Token.parent.pbreak"); if ((labelled_brk != "0") && (ffeature(s,"R:Token.n.name") == "0")) // last word in token { // there is a labelled break on the token so respect it EST_VTCandidate *c = new EST_VTCandidate; c->s = s; c->name = bb_ngram->get_vocab_word(labelled_brk.string()); c->score = log(0.95); // very very likely, but not absolute c->next = all_c; all_c = c; // but then if you give only one option ... } else if (s->next() == 0) // end of utterances so force a break { EST_VTCandidate *c = new EST_VTCandidate; c->s = s; c->name = B_word; c->score = log(0.95); // very very likely, but not absolute c->next = all_c; all_c = c; } else if (s->name() == ".end_utt") { // This is a quick check to see if forcing "." to B is worth it EST_VTCandidate *c = new EST_VTCandidate; c->s = s; c->name = B_word; c->score = log(0.95); // very very likely, but not absolute c->next = all_c; all_c = c; } else if (siod_get_lval("break_non_bayes",NULL) != NIL) { /* This uses the "wrong" formula to extract the probability */ /* Extract P(B | context) rather than P(context | B) as below */ /* This gives worse results as well as not following Bayes */ /* equations */ EST_VTCandidate *c; LISP l; for (l=bb_tags; l != 0; l=cdr(l)) { c = new EST_VTCandidate; c->s = s; tag = bb_ngram->get_vocab_word(get_c_string(car(l))); c->name = tag; window[bb_pos_ngram->order()-1] = tag; const EST_DiscreteProbDistribution &pd = bb_pos_ngram->prob_dist(window); if (pd.samples() == 0) { if (tag == B_word) prob = 0.2; else prob = 0.8; } else prob = pd.probability(tag); if (prob == 0) c->score = log(0.0000001); else c->score = log(prob); c->next = all_c; all_c = c; } } else { // Standard Bayes model EST_VTCandidate *c; LISP l; for (l=bb_tags; l != 0; l=cdr(l)) { c = new EST_VTCandidate; c->s = s; tag = bb_ngram->get_vocab_word(get_c_string(car(l))); c->name = tag; window[bb_pos_ngram->order()-1] = tag; prob = bb_pos_ngram->reverse_probability(window); // If this word came from inside a token reduce the // probability of a break if ((ffeature(s,"R:Token.n.name") != "0") && ((s->as_relation("Token")->first()->length()) < 7)) { float weight = ffeature(s,"pbreak_scale"); if (weight == 0) weight = 0.5; if (tag == B_word) prob *= weight; else prob = 1.0-((1.0-prob)*weight); } if (prob == 0) c->score = log(0.0000001); else c->score = log(prob); s->set("phrase_score",c->score); c->next = all_c; all_c = c; } } return all_c; } static EST_VTPath *bb_npath(EST_VTPath *p,EST_VTCandidate *c,EST_Features &f) { EST_VTPath *np = new EST_VTPath; (void)f; // static EST_String lscorename("lscore"); double prob; double lprob,lang_prob; np->c = c; np->from = p; int n = c->name.Int(); prob = find_b_prob(p,n,&np->state); if (np->state == -1) prob = find_b_prob(p,n,&np->state); if (prob == 0) lprob = log(0.00000001); else lprob = log(prob); lang_prob = (1.0 * c->score) + gscale_p; lang_prob = c->score; //*cdebug << "Addition: "<<(lang_prob+lprob)<set_feature(lscorename,lang_prob+lprob); if (p==0) np->score = (lang_prob+lprob); else np->score = (lang_prob+lprob) + p->score; return np; } static double find_b_prob(EST_VTPath *p,int n,int *state) { int oldstate=0; double prob; /* //1-gram LM enforced *state = 0; if(n == B_word) return 0.06; else return 0.94; */ /* //0-gram LM enforced *state = 0; return 0.5; */ if (p == 0) { int order = bb_ngram->order(); EST_IVector window(order); int i; window.a_no_check(order-1) = n; window.a_no_check(order-2) = B_word; for (i=order-3; i>=0; i--) window.a_no_check(i) = NB_word; oldstate = bb_ngram->find_state_id(window); } else oldstate = p->state; const EST_DiscreteProbDistribution &pd = bb_ngram->prob_dist(oldstate); if (pd.samples() == 0) prob = 0; else prob = (double)pd.probability(n); // This is too specific if (n == B_word) prob *= gscale_s; *state = bb_ngram->find_next_state_id(oldstate,n); //*cdebug << "oldstate: "<state; /* if (streq("of", ( p && p->c && p->c->s ? (const char *)ffeature(p->c->s,"name").String() : "null"))) printf("ding\n"); */ // Prob is the prob that time since last break could be // generated by the duration model // Skip over break if we're at one for (i = oldstate; i < bb_track->num_frames(); i++) { if (bb_track->a(i,0) == ATOTH_NBREAK) break; } // time since last break in words with std for (wstddev=wtime=0.0,d=p; d ; d=d->from) { wtime += ffeature(d->c->s,"word_duration").Float(); wstddev += ffeature(d->c->s,"lisp_word_stddev").Float(); if (bb_track->a(d->state,0) == ATOTH_BREAK) break; } // time since last break in acoustics for (atime=0.01,j=i; j>0; j--) { if (bb_track->a(j,0) == ATOTH_BREAK) break; atime += bb_track->t(j) - ( i == 0 ? 0 : bb_track->t(j-1)); } // Find best state for give time if (wstddev == 0) i++; else if (n == B_word) { /* cost of having a break here */ /* extend acoustics until next break */ for (; i < bb_track->num_frames(); i++) { if (bb_track->a(i,0) == ATOTH_BREAK) break; atime += bb_track->t(i) - ( i == 0 ? 0 : bb_track->t(i-1)); } z = fabs((atime-wtime)/wstddev); } else { /* cost of having a non-break here */ for ( i++,z = fabs((atime-wtime)/wstddev); (i < bb_track->num_frames()) && (bb_track->a(i,0) == ATOTH_NBREAK); i++) { atime += bb_track->t(i) - ( i == 0 ? 0 : bb_track->t(i-1)); /* printf("better atime %f wtime %f wstddev %f z %f new z %f\n", atime,wtime,wstddev,z, fabs((atime-wtime)/wstddev)); */ if (fabs((atime-wtime)/wstddev) > (float)z) break; else z = fabs((atime-wtime)/wstddev); } } if (d && d->c && d->c->s && (d->c->s->next()->next()) == NULL) { /* must be in final state */ printf("must be in final state\n"); if (i != bb_track->num_frames()) z = 0.000001; } if (z == 0) printf("z == 0"); /* number hack */ prob = (2.0 - z)/2.0; /* prob = z; */ if (prob < 0.000001) prob = 0.000001; else if (prob > 0.999999) prob = 0.999999; // prob of atime given (wtime,wstddev) printf("%d %d %f %f %f %f %s %s %f\n",oldstate,i,atime,wtime,wstddev,z, ( p && p->c && p->c->s ? (const char *)ffeature(p->c->s,"name").String() : "null"), ( n == B_word ? "B" : "NB"), prob ); if (i >= bb_track->num_frames()) i = bb_track->num_frames() - 1; *state = i; return prob; } static EST_VTPath *bb_fapath(EST_VTPath *p,EST_VTCandidate *c,EST_Features &f) { EST_VTPath *np = new EST_VTPath; (void)f; // static EST_String lscorename("lscore"); double prob; double lprob,lang_prob; np->c = c; np->from = p; int n = c->name.Int(); prob = find_b_faprob(p,n,&np->state); if (prob == 0) lprob = log(0.00000001); else lprob = log(prob); lang_prob = (1.0 * c->score) + gscale_p; lang_prob = c->score; // np->set_feature(lscorename,lang_prob+lprob); if (p==0) np->score = (lang_prob+lprob); else np->score = (lang_prob+lprob) + p->score; return np; } static void phrasing_by_fa(EST_Utterance *u) { // Predict phrasing using POS and prob models of B distribution EST_Item *w,*phr=0; EST_String pbreak; int num_states; pbyp_get_params(siod_get_lval("phr_break_params",NULL)); gc_protect(&bb_tags); for (w=u->relation("Word")->first(); w != 0; w = w->next()) { // Set up tag index for pos ngram EST_String lpos = map_pos(pos_map,w->f("pos").string()); w->set("phr_pos",lpos); w->set("pos_index",bb_pos_ngram->get_vocab_word(lpos)); } num_states = bb_track->num_frames(); EST_Viterbi_Decoder v(bb_candlist,bb_fapath,num_states); v.initialise(u->relation("Word")); v.search(); v.result("pbreak_index"); // Given predicted break, go through and add phrases u->create_relation("Phrase"); for (w=u->relation("Word")->first(); w != 0; w = w->next()) { w->set("pbreak",bb_ngram-> get_vocab_word(w->f("pbreak_index").Int())); if (phr == 0) phr = add_phrase(u); append_daughter(phr,"Phrase",w); if (phrase_type_tree != NIL) { EST_Val npbreak = wagon_predict(w,phrase_type_tree); w->set("pbreak",npbreak.string()); // may reset to BB } pbreak = (EST_String)w->f("pbreak"); if (pbreak == "B") w->set("blevel",3); else if (pbreak == "mB") w->set("blevel",2); if ((pbreak == "B") || (pbreak == "BB") || (pbreak == "mB")) { phr->set_name((EST_String)pbreak); phr = 0; } } gc_unprotect(&bb_tags); bb_tags = NIL; } EST_Item *add_phrase(EST_Utterance *u) { EST_Item *item = u->relation("Phrase")->append(); item->set_name("phrase"); return item; } festival/src/modules/base/word.cc0000644000175000017500000002145711465010357015445 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* From words to syllables and segments using the lexicon */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "lexicon.h" #include "modules.h" static EST_Item *add_syllable(EST_Utterance *u, int stress); static LISP specified_word_pronunciation(EST_Item *w, LISP lpos); LISP FT_Classic_Word_Utt(LISP utt) { // Look up words in lexicon and create syllable and segment streams EST_Utterance *u = get_c_utt(utt); EST_Item *w; LISP entry,s,p,lpos; EST_String pos; EST_Item *syl,*seg; EST_Relation *SylStructure; *cdebug << "Word module\n"; u->create_relation("Syllable"); u->create_relation("Segment"); SylStructure = u->create_relation("SylStructure"); for (w=u->relation("Word")->first(); w != 0; w = w->next()) { lpos = NIL; pos = (EST_String)ffeature(w,"hg_pos"); // explicit homograph pos disambiguation if (pos == "0") pos = (EST_String)ffeature(w,"pos"); if (pos != "0") lpos = rintern(pos); // Check if there is an explicitly given pronunciation before // going to the lexicon if ((entry = specified_word_pronunciation(w,lpos)) == NIL) entry = lex_lookup_word(w->name(),lpos); if (lpos == NIL) w->set("pos",get_c_string(car(cdr(entry)))); SylStructure->append(w); for (s=car(cdr(cdr(entry))); s != NIL; s=cdr(s)) { syl = add_syllable(u,get_c_int(car(cdr(car(s))))); append_daughter(w,"SylStructure",syl); for (p=car(car(s)); p != NIL; p=cdr(p)) { seg = add_segment(u,get_c_string(car(p))); append_daughter(syl,"SylStructure",seg); } } } return utt; } LISP FT_Unilex_Word_Utt(LISP utt) { // This tries to be a bit cleverer than Classic_Word in dealing with full and reduced forms of words. // Look up words in lexicon and create syllable and segment streams EST_Utterance *u = get_c_utt(utt); EST_Item *w; LISP entry,entry2,s,p,s2,p2,lpos,lexpos; EST_String pos, vowel_form,sname,s2name; EST_Item *syl,*seg; EST_Relation *SylStructure; *cdebug << "Word module\n"; u->create_relation("Syllable"); u->create_relation("Segment"); SylStructure = u->create_relation("SylStructure"); for (w=u->relation("Word")->first(); w != 0; w = w->next()) { lpos = NIL; pos = EST_String(ffeature(w,"hg_pos")); // explicit homograph pos disambiguation if (pos == "0") pos = EST_String(ffeature(w,"pos")); if (pos != "0") lpos = rintern(pos); // Check if there is an explicitly given pronunciation before // going to the lexicon if ((entry = specified_word_pronunciation(w,lpos)) == NIL) entry = lex_lookup_word(w->name(),lpos); lexpos = car(cdr(entry)); // deal with full/reduced specification in pos as a list. entry2 = NIL; if (! atomp(lexpos)) { if ( (vowel_form = get_c_string(car(cdr(lexpos)))) == "full") { entry2 = lex_lookup_word(w->name(),cons(rintern("reduced"),NIL)); if (lpos == NIL) w->set("pos",get_c_string(car(lexpos))); } } else if (lpos == NIL) w->set("pos",get_c_string(lexpos)); SylStructure->append(w); if (entry2) // compare full and reduced form entries for (s=car(cdr(cdr(entry))),s2=car(cdr(cdr(entry2))) ; s != NIL; s=cdr(s)) { syl = add_syllable(u,get_c_int(car(cdr(car(s))))); append_daughter(w,"SylStructure",syl); for (p=car(car(s)),p2=car(car(s2)); p != NIL; p=cdr(p)) { seg = add_segment(u,get_c_string(car(p))); append_daughter(syl,"SylStructure",seg); if(p2 != NIL) { sname = get_c_string(car(p)); s2name = get_c_string(car(p2)); if (sname != s2name) { seg->set("reducable",1); seg->set("fullform",sname); seg->set("reducedform",s2name); } p2=cdr(p2); } } if(s2 != NIL) s2 = cdr(s2); } else for (s=car(cdr(cdr(entry))); s != NIL; s=cdr(s)) { syl = add_syllable(u,get_c_int(car(cdr(car(s))))); append_daughter(w,"SylStructure",syl); for (p=car(car(s)); p != NIL; p=cdr(p)) { seg = add_segment(u,get_c_string(car(p))); append_daughter(syl,"SylStructure",seg); } } } return utt; } static LISP specified_word_pronunciation(EST_Item *w, LISP lpos) { // If there is a phoneme feature on w or the Token related to // w use that as the pronunciation. Note the value will be a string // from which a list can be read. EST_String p; if (((p = (EST_String)ffeature(w,"phonemes")) != "0") || ((p = (EST_String)ffeature(w,"R:Token.parent.phonemes")) != "0")) { LISP phones = read_from_lstring(strintern(p)); return cons(strintern(w->name()), cons(lpos, cons(lex_syllabify(phones),NIL))); } else return NIL; } EST_Item *add_word(EST_Utterance *u, const EST_String &name) { EST_Item *item = u->relation("Word")->append(); item->set_name(name); return item; } EST_Item *add_word(EST_Utterance *u, LISP word) { // Build a Word Ling_Item from the Lisp description, which may // contain other features LISP f; EST_Item *si_word; int has_name = FALSE; if (consp(word)) { // feature form si_word = add_word(u,""); for (f=word; f != NIL; f=cdr(f)) { if (streq("name",get_c_string(car(car(f))))) { has_name = TRUE; si_word->set_name(get_c_string(car(cdr(car(f))))); } else si_word->set(get_c_string(car(car(f))), get_c_string(car(cdr(car(f))))); } if (!has_name) { cerr << "add_word: word has description but no name" << endl; cerr << " " << siod_sprint(word) << endl; festival_error(); } } else // just the name si_word = add_word(u,get_c_string(word)); return si_word; } static EST_Item *add_syllable(EST_Utterance *u, int stress) { EST_Item *item = u->relation("Syllable")->append(); item->set_name("syl"); item->set("stress",stress); return item; } EST_Item *add_segment(EST_Utterance *u, const EST_String &s) { EST_Item *item = u->relation("Segment")->append(); item->set_name(s); return item; } festival/src/modules/base/postlex.cc0000644000175000017500000001212511465010357016160 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : February 1997 */ /*-----------------------------------------------------------------------*/ /* */ /* Post-lexical rules: vowel reduction and contraction, R deleteion */ /* */ /* All this is far too specific, and should be parameterized better */ /* -- and gradually it is ... */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "modules.h" static void vowel_reduction(EST_Utterance *u); static void r_reduction(EST_Utterance *u); static void vowel_reduce(EST_Item *syl,LISP vow_table); LISP FT_PostLex_Utt(LISP utt) { // Do vowel reduction, destructively changes vowel segment values EST_Utterance *u = get_c_utt(utt); vowel_reduction(u); r_reduction(u); return utt; } static void r_reduction(EST_Utterance *u) { // R reduction for mrpa (British English) EST_Item *s,*t; LISP r_red_tree; if (!streq(get_c_string(ft_get_param("PhoneSet")),"mrpa")) return; r_red_tree = siod_get_lval("postlex_mrpa_r_cart_tree",NULL); if (r_red_tree == NIL) return; for (s=u->relation("Segment")->first(); s != 0; s = t) { t = s->next(); if (wagon_predict(s,r_red_tree) == "delete") s->unref_all(); } } static void vowel_reduction(EST_Utterance *u) { EST_Item *s; LISP red_tree, full_vow_table, vow_table=NIL; red_tree = siod_get_lval("postlex_vowel_reduce_cart_tree", NULL); full_vow_table = siod_get_lval("postlex_vowel_reduce_table",NULL); vow_table = car(cdr(siod_assoc_str(get_c_string(ft_get_param("PhoneSet")), full_vow_table))); if ((vow_table == NIL) || (red_tree == NIL)) return; // ain't anything to do for (s=u->relation("Syllable")->first(); s != 0; s = s->next()) { if (wagon_predict(s,red_tree) == "1") vowel_reduce(s,vow_table); } } static void vowel_reduce(EST_Item *syl,LISP vow_table) { // Reduce vowel in syl by looking it up in vow_table for // appropriate vowel mapping EST_Item *seg; LISP vreduce=NIL; for (seg=daughter1(syl,"SylStructure"); seg; seg=seg->next()) { if (ph_is_vowel(seg->name())) { vreduce = siod_assoc_str(seg->name(),vow_table); if (vreduce != NIL) seg->set_name(get_c_string(car(cdr(vreduce)))); return; // ignore any secondary vowels in syllable (should only be one) } } } festival/src/modules/base/phrinfo.cc0000644000175000017500000001133110126474330016123 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : June 1997 */ /*-----------------------------------------------------------------------*/ /* */ /* An input method for much more general info on tokens */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "modules.h" #include "text.h" static EST_Item *make_phrase_node(EST_Utterance *u, const EST_String &name, LISP feats); static EST_Item *make_token_node(EST_Utterance *u, const EST_String &name, LISP feats); void create_phraseinput(EST_Utterance *u) { // Build from phrase input form (phrase, tokens, segs) LISP l,ptree,t; EST_Item *phrase,*token; ptree = utt_iform(*u); u->create_relation("Phrase"); u->create_relation("Token"); for (l=ptree; l != NIL; l=cdr(l)) { if (streq("Phrase",get_c_string(car(car(l))))) { phrase = make_phrase_node(u,"Phrase",car(cdr(car(l)))); for (t=cdr(cdr(car(l))); t != NIL; t=cdr(t)) { if (consp(car(t))) token = make_token_node(u,get_c_string(car(car(t))), car(cdr(car(t)))); else token = make_token_node(u,get_c_string(car(t)),NIL); append_daughter(phrase,token); } } else // no explicit phrase marker { cerr << "PhrInfo: malformed input form." << endl; festival_error(); } } } static EST_Item *make_phrase_node(EST_Utterance *u, const EST_String &name, LISP feats) { // Create a phrase node with name and features EST_Item *p; p = add_phrase(u); p->set_name(name); add_item_features(p,feats); return p; } static EST_Item *make_token_node(EST_Utterance *u, const EST_String &name, LISP feats) { // Create a token node with name and features EST_Token t = name; EST_Item *li = add_token(u,t); LISP f; for (f=feats; f != NIL; f=cdr(f)) { const char *nname = get_c_string(car(car(f))); if (streq(nname,"punctuation")) li->set("punc",get_c_string(car(cdr(car(f))))); else li->set(nname,get_c_string(car(cdr(car(f))))); } return li; } festival/src/modules/Duration/Makefile0000644000175000017500000000570707262605650016504 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### # Contains implementation of various duration prediction # # modules # # # ########################################################################### TOP=../../.. DIRNAME=src/modules/Duration H = durationP.h SRCS = dur_aux.cc duration.cc Klatt.cc OBJS = $(SRCS:.cc=.o) FILES = Makefile $(SRCS) $(H) LOCAL_INCLUDES = -I../include INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/Duration/dur_aux.cc0000644000175000017500000001454310126474330017010 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Basic duration utilities common between different methods */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "durationP.h" float dur_get_stretch(void) { LISP lstretch = ft_get_param("Duration_Stretch"); float stretch; if (lstretch == NIL) stretch = 1.0; else stretch = get_c_float(lstretch); if (stretch < 0.1) { cerr << "Duration_Stretch: is too small (" << stretch << ") ingnoring it\n"; stretch = 1.0; } return stretch; } float dur_get_stretch_at_seg(EST_Item *s) { float global_stretch = dur_get_stretch(); EST_Item *nn = parent(parent(parent(s,"SylStructure")),"Token"); EST_Item *syl = parent(s,"SylStructure"); float local_stretch = 0.0; float syl_stretch = 0.0; float seg_stretch = 0.0; float stretch = 1.0; if (nn) local_stretch = ffeature(nn,"dur_stretch").Float(); if (syl) syl_stretch = ffeature(syl,"dur_stretch").Float(); seg_stretch = ffeature(s,"dur_stretch").Float(); if (local_stretch != 0.0) stretch *= local_stretch; if (syl_stretch != 0.0) stretch *= syl_stretch; if (seg_stretch != 0.0) stretch *= seg_stretch; return stretch*global_stretch; } void festival_Duration_init(void) { festival_def_utt_module("Duration_Averages",FT_Duration_Ave_Utt, "(Duration_Averages UTT)\n\ Label all segments with their average duration found from the assoc\n\ list of phone names to averages in phoneme_durations. This module is\n\ called through the module Duration when the Parameter Duration_Method\n\ is set to Averages. [see Average durations]"); festival_def_utt_module("Duration_Default",FT_Duration_Def_Utt, "(Duration_Default UTT)\n\ Label all segments with a fixed duration of 100ms. This module is\n\ called through the module Duration when the Parameter Duration_Method\n\ is unset or set to Default. [see Default durations]"); festival_def_utt_module("Duration_Tree_ZScores", FT_Duration_Tree_ZScores_Utt, "(Duration_Tree_ZScores UTT)\n\ Uses the CART tree in duration_cart_tree to predict z scores duration\n\ values for each segment in UTT. The z scores are converted back to\n\ absolute values by the assoc list of phones to means and standard\n\ deviations in the variable duration_ph_info. This module is called\n\ through the module Duration when the Parameter Duration_Method is set\n\ to Tree_ZScores. This method modifies its predicted durations by the\n\ factor set in the Parameter Duration_Stretch (if set).\n\ [see CART durations]"); festival_def_utt_module("Duration_Tree",FT_Duration_Tree_Utt, "(Duration_Tree UTT)\n\ Uses the CART tree in duration_cart_tree to predict absolute durations\n\ for each segment in UTT. This module is called through the module\n\ Duration when the Parameter Duration_Method is set to Tree. This\n\ method modifies its predicted durations by the factor set in the\n\ Parameter Duration_Stretch (if set). [see CART durations]"); festival_def_utt_module("Duration_Klatt",FT_Duration_Klatt_Utt, "(Duration_Klatt UTT)\n\ This uses an implementation of the Klatt Duration rules to predict\n\ durations for each segment in UTT. It uses the information in\n\ duration_klatt_params for mean and lower bound for each phone. This\n\ module is called through the module Duration when the Parameter\n\ Duration_Method is set to Klatt. This method modifies its predicted \n\ durations by the factor set in the Parameter Duration_Stretch (if set).\n\ [see Klatt durations]"); } festival/src/modules/Duration/duration.cc0000644000175000017500000001410111465020400017144 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Duration averages and default and tree */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "durationP.h" LISP FT_Duration_Ave_Utt(LISP utt) { // Predict average duration on segments EST_Utterance *u = get_c_utt(utt); EST_Item *s; float end=0.0, dur; LISP ph_durs,ldur; float stretch; *cdebug << "Duration Average module\n"; ph_durs = siod_get_lval("phoneme_durations","no phoneme durations"); for (s=u->relation("Segment")->first(); s != 0; s = s->next()) { ldur = siod_assoc_str(s->name(),ph_durs); stretch = dur_get_stretch_at_seg(s); if (ldur == NIL) { cerr << "Phoneme: " << s->name() << " have no default duration " << endl; dur = 0.100; } else dur = get_c_float(car(cdr(ldur))); end += (dur*stretch); s->set("end",end); } return utt; } LISP FT_Duration_Def_Utt(LISP utt) { // Predict fixed duration on segments EST_Utterance *u = get_c_utt(utt); EST_Item *s; float end=0.0; float stretch; *cdebug << "Duration Default module\n"; for (s=u->relation("Segment")->first(); s != 0; s = s->next()) { stretch = dur_get_stretch_at_seg(s); end += 0.100*stretch; s->set("end",end); } return utt; } LISP FT_Duration_Tree_Utt(LISP utt) { // Predict duration on segments using CART tree EST_Utterance *u = get_c_utt(utt); EST_Item *s; float end=0.0, dur,stretch; LISP tree; EST_Val pdur; *cdebug << "Duration Tree module\n"; tree = siod_get_lval("duration_cart_tree","no duration cart tree"); for (s=u->relation("Segment")->first(); s != 0; s = s->next()) { pdur = wagon_predict(s,tree); stretch = dur_get_stretch_at_seg(s); if (pdur == 0.0) { cerr << "Phoneme: " << s->name() << " tree predicted 0.0 changing it" << endl; dur = 0.050; } else dur = (float)pdur; dur *= stretch; end += dur; s->set("end",end); } return utt; } #define PH_AVE(X) (get_c_float(car(cdr(X)))) #define PH_STD(X) (get_c_float(car(cdr(cdr(X))))) LISP FT_Duration_Tree_ZScores_Utt(LISP utt) { // Predict duration on segments using CART tree EST_Utterance *u = get_c_utt(utt); EST_Item *s; float end=0.0, dur,stretch; LISP tree,dur_info,ph_info; float pdur; float ave, std; *cdebug << "Duration Tree ZScores module\n"; tree = siod_get_lval("duration_cart_tree","no duration cart tree"); dur_info = siod_get_lval("duration_ph_info","no duration phone info"); for (s=u->relation("Segment")->first(); s != 0; s = s->next()) { pdur = wagon_predict(s,tree); ph_info = siod_assoc_str(s->name(),dur_info); stretch = dur_get_stretch_at_seg(s); if (ph_info == NIL) { cerr << "Phoneme: " << s->name() << " has no duration info\n"; ave = 0.080; std = 0.020; } else { ave = PH_AVE(ph_info); std = PH_STD(ph_info); } if ((pdur > 3) || (pdur < -3)) { // cerr << "Duration tree extreme for " << s->name() << // " " << pdur << endl; pdur = ((pdur < 0) ? -3 : 3); } s->set("dur_factor",pdur); dur = ave + (pdur*std); dur *= stretch; if (dur < 0.010) dur = 0.010; // just in case it goes wrong end += dur; s->set("end",end); } return utt; } festival/src/modules/Duration/Klatt.cc0000644000175000017500000002675611465010356016432 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Paul Taylor */ /* Date : July 1995 */ /*-----------------------------------------------------------------------*/ /* */ /* Klatt Duration Rules */ /* */ /*=======================================================================*/ /* This is an implementation of the Klatt rule system as described in chapter 9 of "From text to speech: The MITalk system", Allen, Hunnicutt and Klatt. The function klatt_seg_dur() calculates a duration for each segment in the input. It does this by calling a number of rules (named 1 to 11) as defined in the MITalk book. Most rules return a number which modifies the inherenent duration of each segment. The original rules are set up so as to return a percentage, here the system retursn a floating point value which I think is neater. */ #include #include "festival.h" #include "durationP.h" static void klatt_dur_debug(EST_Item *s); static float rule2(EST_Item *seg); static float rule3(EST_Item *seg); static float rule4(EST_Item *seg); static float rule5(EST_Item *seg); static float rule6(EST_Item *seg); static float rule7(EST_Item *seg); static float rule8(EST_Item *seg); static float rule9(EST_Item *seg); static float rule10(EST_Item *seg); static float rule9a(EST_Item *seg); static float sub_rule9a(const EST_String &ph); static int klatt_seg_dur(EST_Item *seg); static float min_dur(EST_Item *s_seg); static float inher_dur(EST_Item *s_seg); int onset(EST_Item *seg); static LISP klatt_params = NIL; static int debug = 0; LISP FT_Duration_Klatt_Utt(LISP utt) { // Predict fixed duration on segments EST_Utterance *u = get_c_utt(utt); EST_Item *s; *cdebug << "Duration Klatt module\n"; klatt_params = siod_get_lval("duration_klatt_params", "no klatt duration params"); for (s=u->relation("Segment")->first(); s != 0; s = s->next()) klatt_seg_dur(s); return utt; } static int klatt_seg_dur(EST_Item *seg) { float min; float fact = 1.0; float start, dur; float duration_speed = dur_get_stretch_at_seg(seg); start = ffeature(seg,"segment_start"); if (ph_is_silence(seg->name())) dur = 0.250 * duration_speed; else { if (debug) klatt_dur_debug(seg); fact *= rule2(seg) * rule3(seg) * rule4(seg) * rule5(seg) * rule6(seg) * rule7(seg) * rule8(seg) * rule9(seg) * rule10(seg); min = (rule7(seg) != 1.0) ? min_dur(seg)/2: min_dur(seg); dur = ((((inher_dur(seg) - min) * fact) + min) / 1000.0) * duration_speed; } seg->set("end",start + dur); return 0; } static float min_dur(EST_Item *seg) { LISP p = siod_assoc_str(seg->name(),klatt_params); if (p == NIL) { cerr << "Klatt_Duration: no minimum duration for \"" << seg->name() << "\"\n"; festival_error(); } return get_c_float(car(cdr(cdr(p)))); } static float inher_dur(EST_Item *seg) { LISP p = siod_assoc_str(seg->name(),klatt_params); if (p == NIL) { cerr << "Klatt_Duration: no minimum duration for \"" << seg->name() << "\"\n"; festival_error(); } return get_c_float(car(cdr(p))); } static int word_final(EST_Item *seg) { // True if this segment is the last in a word EST_Item *nn = seg->as_relation("SylStructure"); if (nn->next() || (parent(nn)->next())) return FALSE; else return TRUE; } static int syl_final(EST_Item *seg) { // True if this segment is the last in a syllable EST_Item *nn = seg->as_relation("SylStructure"); if (nn->next()) return FALSE; else return TRUE; } static int word_initial(EST_Item *seg) { // True if this segment is the first in a word EST_Item *nn = seg->as_relation("SylStructure"); if (nn->prev() || parent(nn)->prev()) return FALSE; else return TRUE; } static int phrase_initial(EST_Item *seg) { // True if this segment is the first in a phrase if (word_initial(seg)) { EST_Item *nn = parent(parent(seg,"SylStructure")); if (as(nn,"Phrase")->prev()) return FALSE; else return TRUE; } return FALSE; } int onset(EST_Item *seg) { if (ffeature(seg,"onsetcoda") == "onset") return 1; else return 0; } int coda(EST_Item *seg) { if (ffeature(seg,"onsetcoda") == "coda") return 1; else return 0; } static float rule2(EST_Item *seg) { // clause final lengthening if (coda(seg)) { int b = ffeature(seg,"R:SylStructure.parent.syl_break"); if ((b > 1) && (b < 4)) return 1.4; } return 1.0; } static float rule3(EST_Item *seg) { // Non-phrase-final shortening // syllabic segments are shortened by 60 if not in a phrase-final syllable int b = ffeature(seg,"R:SylStructure.parent.syl_break"); if ((b < 2) && ph_is_syllabic(seg->name())) return 0.6; // A phrase-final postvocalic liquid or nasal is lengthened by 140 if ((b == 4) && (ph_is_liquid(seg->name()) || ph_is_nasal(seg->name()))) return(1.4); return 1.0; } static float rule4(EST_Item *seg) { // Non-word-final shortening int b = ffeature(seg,"R:SylStructure.parent.syl_break"); // Syllabic segments are shortened by 85 if not in a word-final syllable if ((b == 0) && ph_is_syllabic(seg->name())) return(0.85); return 1.0; } static float rule5(EST_Item *seg) { // Polysyllabic Shortening int num_syls = ffeature(seg,"R:SylStructure.parent.parent.num_syls"); // Syllabic segments in a polysyllabic word are shortened by 80. if ((num_syls > 1) && ph_is_syllabic(seg->name())) return 0.8; return 1.0; } static float rule6(EST_Item *seg) { // Non-initial-consonant shortening if (!word_initial(seg) && (ph_is_consonant(seg->name()))) return 0.85; return 1.0; } static float rule7(EST_Item *seg) { // Unstressed shortening if (ffeature(seg,"R:SylStructure.parent.stress") == 1) return 1.0; if (ph_is_syllabic(seg->name())) { if (word_initial(seg) || word_final(seg)) return 0.7; else return 0.5; } if (onset(seg) && ph_is_liquid(seg->name())) // or glide... return 0.1; return 0.7; } // Lengthening for emphasis static float rule8(EST_Item *seg) { if (!ph_is_vowel(seg->name())) return 1.0; if (ffeature(seg,"R:SylStructure.parent.accented") == 1) return 1.4; return 1.0; } // this is really rule 9b, but its eaiser to make it call rule 9a static float rule9(EST_Item *seg) { // Postvocalic context of vowels */ int b = ffeature(seg,"R:SylStructure.parent.syl_break"); if (b > 1) return (0.7 + (0.3 * rule9a(seg))); else return rule9a(seg); } static float rule9a(EST_Item *seg) { // Postvocalic context of vowels EST_Item *s_next,*s_next_next; if (ph_is_vowel(seg->name())) { if (syl_final(seg)) return 1.2; s_next = seg->next(); if ((s_next) && (syl_final(s_next))) return sub_rule9a(s_next->name()); s_next_next = s_next->next(); if ((ph_is_sonorant(s_next->name())) && (s_next_next) && (ph_is_obstruent(s_next_next->name()))) return sub_rule9a(s_next_next->name()); } else if (onset(seg)) return 1.0; else if (ph_is_sonorant(seg->name())) { if (syl_final(seg)) return 1.2; s_next = seg->next(); if (ph_is_obstruent(s_next->name())) return sub_rule9a(s_next->name()); } return 1.0; } // sub rule, independent of seg position static float sub_rule9a(const EST_String &ph) { if (ph_is_voiced(ph)) { if (ph_is_fricative(ph)) return 1.6; else if (ph_is_stop(ph)) return 1.2; else if (ph_is_nasal(ph)) return 0.85; else return 1.0; } else if (ph_is_stop(ph)) return 0.7; else return 1.0; } // Shortening in clusters static float rule10(EST_Item *seg) { int b = ffeature(seg,"R:SylStructure.parent.syl_break"); if (syl_final(seg) && (b > 1)) return 1.0; else { if (ph_is_vowel(seg->name())) { if (ph_is_vowel(seg->next()->name())) return 1.20; else if ((!phrase_initial(seg)) && (ph_is_vowel(seg->prev()->name()))) return 0.70; else return 1.0; } else if (ph_is_consonant(seg->next()->name())) if (!phrase_initial(seg) && (ph_is_consonant(seg->prev()->name()))) return 0.5; else return 0.7; else if (!phrase_initial(seg) && (ph_is_consonant(seg->prev()->name()))) return 0.7; } return 1.0; } static void klatt_dur_debug(EST_Item *seg) { float f; if ((f = rule2(seg))!= 1.0) cout << "Fired rule 2 " << f << endl; if ((f = rule3(seg))!= 1.0) cout << "Fired rule 3 " << f << endl; if ((f = rule4(seg))!= 1.0) cout << "Fired rule 4 " << f << endl; if ((f = rule5(seg))!= 1.0) cout << "Fired rule 5 " << f << endl; if ((f = rule6(seg))!= 1.0) cout << "Fired rule 6 " << f << endl; if ((f = rule7(seg))!= 1.0) cout << "Fired rule 7 " << f << endl; if ((f = rule8(seg))!= 1.0) cout << "Fired rule 8 " << f << endl; if ((f = rule9(seg))!= 1.0) cout << "Fired rule 9 " << f << endl; if ((f = rule10(seg))!= 1.0) cout << "Fired rule 10" << f << endl; return; } festival/src/modules/Duration/durationP.h0000644000175000017500000000636107262605650017157 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Shared duration utilities */ /* */ /*=======================================================================*/ #ifndef __DURATION_H__ #define __DURATION_H__ LISP FT_Duration_Ave_Utt(LISP args); LISP FT_Duration_Def_Utt(LISP args); LISP FT_Duration_Tree_Utt(LISP args); LISP FT_Duration_Tree_ZScores_Utt(LISP args); LISP FT_Duration_Klatt_Utt(LISP args); float dur_get_stretch(void); float dur_get_stretch_at_seg(EST_Item *s); #endif /* __DURATION_H__ */ festival/src/modules/Intonation/Makefile0000644000175000017500000000572507262605650017041 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### # # # Contains implementations of various intonation theories # # # ########################################################################### TOP=../../.. DIRNAME=src/modules/Intonation H = SRCS = int_aux.cc duffint.cc simple.cc gen_int.cc int_tree.cc OBJS = $(SRCS:.cc=.o) FILES = Makefile $(SRCS) $(H) LOCAL_INCLUDES = -I../include INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/Intonation/int_aux.cc0000644000175000017500000002274510220237411017340 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Basic intonation utilities common between different */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "intonation.h" #include "modules.h" #include "lexicon.h" static EST_String IntEventname("IntEvent"); static EST_String Targetname("Target"); EST_Item *add_target(EST_Utterance *u,EST_Item *seg, float pos, float val) { // Check time is NOT the same as the last target, as this causes problems... float last_time; EST_Item* last_item = u->relation(Targetname)->last_leaf(); if (last_item) last_time = last_item->f("pos"); else last_time = -1.0; // no last time. if(last_time == pos) { pos += 0.001; *cdebug << "Repeated f0 target time, fix your generation function!\n"; } if (seg->as_relation(Targetname) == 0) u->relation(Targetname)->append(seg); EST_Item *item = append_daughter(seg,Targetname); item->set("f0",val); item->set("pos",pos); return item; } EST_Item *add_IntEvent(EST_Utterance *u,EST_Item *syl, const EST_String &label) { if (syl->as_relation("Intonation") == 0) u->relation("Intonation")->append(syl); EST_Item *item = u->relation(IntEventname)->append(); item->set_name(label); append_daughter(syl,"Intonation",item); return item; } void targets_to_f0(EST_Relation &targ, EST_Track &f0, const float shift) { float prev_f0=0.0; float prev_pos=0, m; EST_Item *s; int i; f0.resize(int(ceil(targ.last_leaf()->F("pos",0) / shift)), 1); f0.fill_time(shift); s = targ.first_leaf(); // fill with zeros until first target; for (i = 0; i < f0.num_frames(); ++i) { if (f0.t(i) > s->F("pos",0)) break; f0.a(i) = 0.0; } prev_pos = s->F("pos",0); prev_f0 = s->F("f0",0); s = next_leaf(s); for (m=0.0,i = 0; i < f0.num_frames(); ++i) { if (s && f0.t(i) > s->F("pos")) { prev_pos = s->F("pos"); prev_f0 = s->F("f0"); s = next_leaf(s); if (s == 0) break; m = (s->F("f0") - prev_f0)/ (s->F("pos") - prev_pos); } f0.a(i) = (m * (f0.t(i) - prev_pos)) + prev_f0; } for ( ; i < f0.num_frames(); ++i) f0.a(i) = 0.0; } LISP FT_us_targets_to_f0(LISP lutt) { EST_Utterance *utt = get_c_utt(lutt); EST_Track *f0 = new EST_Track; utt->create_relation("f0"); EST_Item *f = utt->relation("f0")->append(); f->set("name", "f0"); f->set_val("f0", est_val(f0)); targets_to_f0(*utt->relation("Target"), *f0, 0.01); return lutt; } LISP FT_Intonation_Default_Utt(LISP args); LISP FT_Int_Targets_Default_Utt(LISP args); LISP FT_Intonation_Simple_Utt(LISP args); LISP FT_Int_Targets_Simple_Utt(LISP args); LISP FT_Intonation_Tree_Utt(LISP args); LISP FT_Int_Targets_LR_Utt(LISP args); LISP FT_Int_Targets_LR_5_Utt(LISP args); LISP FT_Int_Targets_General_Utt(LISP utt); LISP FT_Int_Targets_Relation_Utt(LISP utt, LISP relname); void festival_Intonation_init(void) { festival_def_utt_module("Intonation_Default",FT_Intonation_Default_Utt, "(Intonation_Default UTT)\n\ this method is such a bad intonation module that it does nothing at all.\n\ This utterance module is called when the Parameter Int_Method is not\n\ set or set to Default. This module is called through the Intonation\n\ module. [see Default intonation]"); init_subr_2("Int_Targets_Relation", FT_Int_Targets_Relation_Utt, "(Int_Targets_Relation UTT)"); init_subr_1("targets_to_f0", FT_us_targets_to_f0, "(targets_to_f0 UTT)\n\ Make f0 relation, and place an f0 contour in it, using F0 targets\n\ from the Target Relation\n"); festival_def_utt_module("Int_Targets_Default",FT_Int_Targets_Default_Utt, "(Int_Targets_Default UTT)\n\ This module creates two Targets causing a simple downward continuous\n\ F0 through the whole utterance. The code is in an appropriate named file\n\ called duffint. This module is called when the Parameter\n\ Int_Method is not set or set to Default. This module is called through\n\ the Int_Targets module. Optional parameters for a start value (default\n\ 130) and end value (default 110) may be set in the variable\n\ diffint_params. This can be used to generate a monotone intonation\n\ with a setting like (set! duffint_params '((start 100) (end 100))).\n\ [see Default intonation]"); festival_def_utt_module("Intonation_Simple",FT_Intonation_Simple_Utt, "(Intonation_Simple)\n\ Assign accents to each content word, creating an IntEvent stream. This \n\ utterance module is called when the Parameter Int_Method is set to \n\ Simple. This module is called through the Intonation module.\n\ [see Simple intonation]"); festival_def_utt_module("Int_Targets_Simple",FT_Int_Targets_Simple_Utt, "(Int_Targets_Simple UTT)\n\ Naively add targets for hat shaped accents for each accent in the \n\ IntEvent stream. This module is called when the Parameter Int_Method is\n\ set to Simple. This module is called through the Int_Targets module.\n\ [see Simple intonation]"); festival_def_utt_module("Int_Targets_General",FT_Int_Targets_General_Utt, "(Int_Targets_General UTT)\n\ Add targets based on the functions defined in int_general_params. This\n\ method allows quite detailed control over the general of targets per\n\ syllable, see manual for details and examples. This module is called\n\ when the Parameter Int_Method is set to General. This module is called\n\ through the Int_Targets module. [see General intonation]"); festival_def_utt_module("Intonation_Tree",FT_Intonation_Tree_Utt, "(Intonation_Tree UTT)\n\ Use the CART trees in int_tone_cart_tree and int_accent_cart_tree to\n\ create an IntEvent stream of tones and accents related to syllables.\n\ This module is called through the Intonation module and is selected\n\ when the Parameter Int_Method is ToBI. [see Tree intonation]"); festival_def_utt_module("Int_Targets_LR",FT_Int_Targets_LR_Utt, "(Int_Targets_LR UTT)\n\ Predict Target F0 points using linear regression from factors such as\n\ accent, tone, stress, position in phrase etc. This utterance module is\n\ called through the module Int_Targets when the Parameter Int_Method is\n\ set to ToBI, even though this technique is not restricted to the ToBI\n\ labelling system. [see Tree intonation]"); festival_def_utt_module("Int_Targets_5_LR",FT_Int_Targets_LR_5_Utt, "(Int_Targets_5_LR UTT)\n\ Predict Target F0 points using linear regression from factors such as\n\ accent, tone, stress, position in phrase etc. This utterance module is\n\ called through the module Int_Targets when the Parameter Int_Method is\n\ set to ToBI, even though this technique is not restricted to the ToBI\n\ labelling system. [see Tree intonation]"); } festival/src/modules/Intonation/duffint.cc0000644000175000017500000001157010126474330017332 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Duff intonation */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "intonation.h" LISP FT_Intonation_Default_Utt(LISP utt) { return utt; } LISP FT_Int_Targets_Default_Utt(LISP utt) { // Predict intonation from labels etc (producing F0 contour) EST_Utterance *u = get_c_utt(utt); EST_Item *s; EST_Relation *seg; LISP params; float start,end; *cdebug << "Intonation duff module\n"; // should create some random targets params = siod_get_lval("duffint_params",NULL); start = get_param_float("start",params,130.0); end = get_param_float("end",params,110.0); u->create_relation("Target"); seg = u->relation("Segment"); if (seg->length() == 0) return utt; add_target(u,seg->first(),0,start); s = seg->last(); add_target(u,s,(float)ffeature(s,"segment_end"),end); return utt; } LISP FT_Int_Targets_Relation_Utt(LISP utt, LISP relname) { // Predict intonation from labels etc (producing F0 contour) EST_Utterance *u = get_c_utt(utt); EST_Track *pm = 0; LISP params; float start,end; int n_frames; *cdebug << "Intonation duff module\n"; // should create some random targets params = siod_get_lval("duffint_params",NULL); start = get_param_float("start",params,130.0); end = get_param_float("end",params,110.0); pm = track(u->relation(get_c_string(relname))->head()->f("coefs")); float pp = 1.0/start; // float end_time = ((float)pm->num_frames()) * pp; float end_time = pm->end(); n_frames = (int)(ceil)(end_time/pp); cout << "n_frames: " << n_frames << endl; cout << "end_time: " << end_time << endl; EST_Track *f0 = new EST_Track; f0->resize(n_frames, 1); f0->fill_time(0.01); float m = (end-start) /end_time; float c = start; for (int i = 0; i < n_frames; ++ i) f0->a(i) = (m * ((float) i) * 0.01) + c; u->create_relation("f0"); EST_Item *item = u->relation("f0")->append(); item->set_val("f0", est_val(f0)); return utt; } festival/src/modules/Intonation/simple.cc0000644000175000017500000001345211465010356017166 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black (and Paul Taylor) */ /* Date : April 199[4|6] */ /*-----------------------------------------------------------------------*/ /* */ /* Simple intonation prediction: a hat shape on each content word */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "intonation.h" static void add_targets(EST_Utterance *u,EST_Item *syl, float baseline,float peak); LISP FT_Intonation_Simple_Utt(LISP utt) { // Predict some accents EST_Utterance *u = get_c_utt(utt); EST_Item *s; LISP accent_tree; EST_Val paccent; *cdebug << "Simple intonation module" << endl; accent_tree = siod_get_lval("int_accent_cart_tree","no accent tree"); u->create_relation("IntEvent"); u->create_relation("Intonation"); for (s=u->relation("Syllable")->first(); s != 0; s = s->next()) { paccent = wagon_predict(s,accent_tree); if (paccent != "NONE") add_IntEvent(u,s,paccent.string()); } return utt; } LISP FT_Int_Targets_Simple_Utt(LISP utt) { // Predict F0 targets EST_Utterance *u = get_c_utt(utt); EST_Item *s, *p, start_word, end_word; float start,end,duration; float baseline, decline; EST_Item *start_syl, *end_syl; LISP simple_params; float f0_mean, f0_std; *cdebug << "Simple int targets module" << endl; // Create some down step accents simple_params = siod_get_lval("int_simple_params","no simple params"); f0_mean = get_param_int("f0_mean",simple_params,110); f0_std = get_param_int("f0_std",simple_params,25); u->create_relation("Target"); for (p=u->relation("Phrase")->first(); p != 0 ; p=p->next()) { baseline = f0_mean + (f0_std * 0.6); start = ffeature(p,"R:Phrase.daughter1.word_start"); end = ffeature(p,"R:Phrase.daughtern.word_end"); duration = end - start; decline = f0_std / duration; start_syl = daughter1(daughter1(p),"SylStructure"); end_syl = daughtern(daughtern(p),"SylStructure"); if (start_syl) add_target(u,daughter1(start_syl,"SylStructure"), ffeature(start_syl,"R:SylStructure.daughter1.segment_start"), baseline); for (s=start_syl->as_relation("Syllable"); s != end_syl->next(); s = s->next()) { if (ffeature(s,"accented") == 1) add_targets(u,s,baseline,f0_std); baseline -= decline*(ffeature(s,"syllable_duration").Float()); } if (end_syl) add_target(u,daughtern(end_syl,"SylStructure"), ffeature(end_syl,"R:SylStructure.daughtern.segment_end"), f0_mean-f0_std); } return utt; } static void add_targets(EST_Utterance *u,EST_Item *syl, float baseline,float peak) { // Add a down stepped accent at this point EST_Item *first_seg = daughter1(syl,"SylStructure"); EST_Item *end_seg = daughter1(syl,"SylStructure"); EST_Item *t=0,*vowel_seg; add_target(u,first_seg,ffeature(first_seg,"segment_start"),baseline); vowel_seg = end_seg; // by default for (t = first_seg; t != 0; t = t->next()) if (ph_is_vowel(t->name())) { vowel_seg = t; break; } add_target(u,vowel_seg,ffeature(vowel_seg,"segment_mid"),baseline+peak); add_target(u,end_seg,ffeature(end_seg,"segment_end"),baseline); } festival/src/modules/Intonation/gen_int.cc0000644000175000017500000001300211465010356017307 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* A general intonation method for implementing various simple rule */ /* intonation systems. It allows a list of targets to be predicted */ /* in a way fully specified by the user without changing the C/C++ code */ /* This was specifically designed to replace the simple intonation mode */ /* monotone mode, and implemented generic ToBI type labels. */ /* */ /* This was to help Gregor Moehler do German ToBI as well as get a */ /* we can use for a rule-based English ToBI for comparison with trained */ /* versions */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "intonation.h" static void check_targs(EST_Utterance *u); static EST_Item *find_nearest_seg(EST_Utterance *u,float pos); LISP FT_Int_Targets_General_Utt(LISP utt) { // Predict F0 targets EST_Utterance *u = get_c_utt(utt); EST_Item *s; EST_Item *seg; EST_Relation *targrel; LISP gen_params, targets, t; LISP tfunc; // a lisp function that returns list of targets and values // Create some down step accents gen_params = siod_get_lval("int_general_params", "no general intonation simple params"); tfunc = get_param_lisp("targ_func",gen_params,NIL); if (tfunc == NIL) { cerr << "Int Target General: no target function specified" << endl; festival_error(); } targrel = u->create_relation("Target"); for (s=u->relation("Syllable")->first(); s != 0 ; s=s->next()) { targets = leval(cons(tfunc,cons(utt,cons(siod(s),NIL))),NIL); // Add the given targets for (t=targets; t != NIL; t=cdr(t)) { seg = find_nearest_seg(u,get_c_float(car(car(t)))); add_target(u,seg,get_c_float(car(car(t))), get_c_float(car(cdr(car(t))))); } } check_targs(u); return utt; } static EST_Item *find_nearest_seg(EST_Utterance *u,float pos) { // Find the segment that this target falls within. // This naively searchs from the start of the segments, // this is not very efficient EST_Item *seg; for (seg=u->relation("Segment")->first(); seg != 0;seg=seg->next()) { if (seg->F("end") >= pos) return seg; } cerr << "Int Target General: target past end of segments at " << pos << endl; festival_error(); return NULL; } static void check_targs(EST_Utterance *u) { // Check targets are in order EST_Item *t; float l = 0.0; for (t=u->relation("Target")->first_leaf(); t != 0;t=next_leaf(t)) { if (t->F("pos") < l) { cerr << "Int Target General: targets out of order" << endl; festival_error(); } l = t->F("pos"); } } festival/src/modules/Intonation/int_tree.cc0000644000175000017500000003306111465010356017504 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : May 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Tree-based prediction of intonation. Uses accent and end */ /* tone prediction trees, could be ToBI could be something */ /* else, its up to the trees to decide ... */ /* */ /* Accents and boundaries are predicted by CART tree while */ /* the F0 targets are predicted by linear regression (as */ /* described in Black and Hunt ICSLP96) */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "intonation.h" enum lr_tpos {tp_start, tp_left, tp_mid, tp_right, tp_end}; static EST_String accent_specified(EST_Item *s); static EST_String tone_specified(EST_Item *s); static int after_pause(EST_Item *s); static int before_pause(EST_Item *s); static EST_Item *vowel_seg(EST_Item *syl); static void init_int_lr_params(void); static void add_target_at(EST_Utterance *u, EST_Item *seg, float val,lr_tpos pos); static float apply_lr_model(LISP model, EST_FVector &feats); static void find_feat_values(EST_Item *s, LISP model,EST_FVector &feats); static LISP Intonation_Endtone_Tree_Utt(LISP utt); // ... mh 99-08-06 static LISP Intonation_Accent_Tree_Utt(LISP utt); static float target_f0_mean = 0.0; static float target_f0_std = 1.0; static float model_f0_mean = 0.0; static float model_f0_std = 1.0; #define MZSCORE(X) (((X)-model_f0_mean)/model_f0_std) #define UNTZSCORE(X) (((X)*target_f0_std)+target_f0_mean) #define MAP_F0(X) (UNTZSCORE(MZSCORE(X))) LISP FT_Intonation_Tree_Utt(LISP utt) { // For each syllable predict intonation events. Potentially // two forms, accents and ent tones EST_Utterance *u = get_c_utt(utt); u->create_relation("IntEvent"); u->create_relation("Intonation"); utt = Intonation_Endtone_Tree_Utt(utt); utt = Intonation_Accent_Tree_Utt(utt); return utt; } LISP Intonation_Accent_Tree_Utt(LISP utt) { // For each syllable predict intonation events. // here only accents EST_Utterance *u = get_c_utt(utt); EST_Item *s; EST_String paccent; LISP accent_tree; accent_tree = siod_get_lval("int_accent_cart_tree","no accent tree"); for (s=u->relation("Syllable")->first(); s != 0; s=s->next()) { if ((paccent = accent_specified(s)) == "0") // check if pre-specified paccent = (EST_String)wagon_predict(s,accent_tree); if (paccent != "NONE") add_IntEvent(u,s,paccent); } return utt; } LISP Intonation_Endtone_Tree_Utt(LISP utt) { // For each syllable predict intonation events. // here only endtones EST_Utterance *u = get_c_utt(utt); EST_Item *s; EST_String ptone; LISP endtone_tree; endtone_tree = siod_get_lval("int_tone_cart_tree","no tone cart tree"); for (s=u->relation("Syllable")->first(); s != 0; s=s->next()) { if ((ptone = tone_specified(s)) == "0") ptone = (EST_String)wagon_predict(s,endtone_tree); if (ptone != "NONE") add_IntEvent(u,s,ptone); } return utt; } static EST_String accent_specified(EST_Item *s) { // If there is an explicit accent specifed on the related token // If there is check the syllable to see if its stress or a singleton EST_Item *word = parent(s,"SylStructure"); if (!word) return "0"; EST_Item *token = parent(word,"Token"); EST_String paccent("0"); if (token) paccent = (EST_String)ffeature(token,"accent"); if (paccent == "0") { paccent = (EST_String)ffeature(word,"accent"); if (paccent == "0") return paccent; } if (ffeature(s,"stress") == "1") { // only goes on first stressed syllable EST_Item *p; for (p=as(s,"SylStructure")->prev(); p != 0; p=p->prev()) if (ffeature(s,"stress") == "1") return "NONE"; // specified but not on this syllable return paccent; // first stressed syl in word } else if (daughter1(word)->length() == 1) return paccent; else return "NONE"; // pre-specified but inappropriate syllable in word } static EST_String tone_specified(EST_Item *s) { // If there is an explicit accent specifed on the related token // If there is check the syllable to see if its strees or a singleton EST_Item *ss = s->as_relation("SylStructure"); EST_Item *word = parent(ss); if (!word) return "0"; EST_Item *token = parent(word,"Token"); EST_String ptone("0"); if (token) ptone = (EST_String)ffeature(token,"tone"); if (ptone == "0") { ptone = (EST_String)ffeature(word,"tone"); if (ptone == "0") return ptone; } if (ss->next() == 0) // final syllable in word return ptone; else return "NONE"; // pre-specified but inappropriate syllable in word } LISP FT_Int_Targets_LR_Utt(LISP utt) { // Predict F0 targets using Linear regression EST_Utterance *u = get_c_utt(utt); EST_Item *s; float pstart, pmid, pend; LISP start_lr, mid_lr, end_lr; init_int_lr_params(); // Note the models must *all* be the same size start_lr = siod_get_lval("f0_lr_start","no f0 start lr model"); mid_lr = siod_get_lval("f0_lr_mid","no f0 mid lr model"); end_lr = siod_get_lval("f0_lr_end","no f0 end lr model"); u->create_relation("Target"); pend = 0; EST_FVector feats; feats.resize(siod_llength(start_lr)); for (s=u->relation("Syllable")->first(); s != 0; s=s->next()) { find_feat_values(s,start_lr,feats); pstart = apply_lr_model(start_lr,feats); pstart = MAP_F0(pstart); if (after_pause(s)) add_target_at(u,daughter1(s,"SylStructure"),pstart,tp_start); else add_target_at(u,daughter1(s,"SylStructure"), (pstart+pend)/2.0,tp_start); pmid = apply_lr_model(mid_lr,feats); pmid = MAP_F0(pmid); add_target_at(u,vowel_seg(s),pmid,tp_mid); pend = apply_lr_model(end_lr,feats); pend = MAP_F0(pend); if (before_pause(s)) add_target_at(u,daughtern(s,"SylStructure"),pend,tp_end); } return utt; } LISP FT_Int_Targets_LR_5_Utt(LISP utt) { // Predict F0 targets using Linear regression // This version uses 5 points rather than 3. EST_Utterance *u = get_c_utt(utt); EST_Item *s; float pstart, pleft, pmid, pright, pend; LISP start_lr, left_lr, mid_lr, right_lr, end_lr; init_int_lr_params(); // Note the models must *all* be the same size start_lr = siod_get_lval("f0_lr_start","no f0 start lr model"); left_lr = siod_get_lval("f0_lr_left","no f0 left lr model"); mid_lr = siod_get_lval("f0_lr_mid","no f0 mid lr model"); right_lr = siod_get_lval("f0_lr_right","no f0 right lr model"); end_lr = siod_get_lval("f0_lr_end","no f0 end lr model"); u->create_relation("Target"); pend = 0; EST_FVector feats; feats.resize(siod_llength(start_lr)); for (s=u->relation("Syllable")->first(); s != 0; s=s->next()) { find_feat_values(s,start_lr,feats); pstart = apply_lr_model(start_lr,feats); pstart = MAP_F0(pstart); if (after_pause(s)) add_target_at(u,daughter1(s,"SylStructure"),pstart,tp_start); else add_target_at(u,daughter1(s,"SylStructure"), (pstart+pend)/2.0,tp_start); pleft = apply_lr_model(left_lr,feats); pleft = MAP_F0(pleft); add_target_at(u,vowel_seg(s),pleft,tp_left); pmid = apply_lr_model(mid_lr,feats); pmid = MAP_F0(pmid); add_target_at(u,vowel_seg(s),pmid,tp_mid); pright = apply_lr_model(right_lr,feats); pright = MAP_F0(pright); add_target_at(u,vowel_seg(s),pright,tp_right); pend = apply_lr_model(end_lr,feats); pend = MAP_F0(pend); if (before_pause(s)) add_target_at(u,daughtern(s,"SylStructure"),pend,tp_end); } return utt; } #define FFEATURE_NAME(X) (get_c_string(car(X))) #define FFEATURE_WEIGHT(X) (get_c_float(car(cdr(X)))) #define FFEATURE_MAPCLASS(X) (car(cdr(cdr(X)))) static void find_feat_values(EST_Item *s, LISP model,EST_FVector &feats) { EST_Val v = 0.0; int i; LISP f; const char *ffeature_name, *last_name=""; feats[0] = 1; for (i=1,f=cdr(model); CONSP(f); f=CDR(f),i++) { ffeature_name = FFEATURE_NAME(CAR(f)); if (!streq(ffeature_name,last_name)) v = ffeature(s,ffeature_name); if (siod_llength(CAR(f)) == 3) { // A map class is specified if (siod_member_str(v.string(),FFEATURE_MAPCLASS(CAR(f))) != NIL) feats[i] = 1; else feats[i] = 0; } else feats[i] = (float)v; last_name = ffeature_name; } } static float apply_lr_model(LISP model, EST_FVector &feats) { float answer = FFEATURE_WEIGHT(car(model)); int i; LISP f; for(i=1,f=cdr(model); iF("end")), val); else if (pos == tp_end) add_target(u,seg,seg->F("end"),val); else { cerr << "add_target_at: unknown position type\n"; festival_error(); } } static int after_pause(EST_Item *s) { // TRUE if segment immediately previous to this is a silence EST_Item *p; if (s->prev() == 0) return TRUE; EST_Item *ss = s->as_relation("SylStructure"); if (s->prev() == ss->prev()) return FALSE; p = daughter1(ss)->as_relation("Segment")->prev(); if (p == 0) return TRUE; else if (ph_is_silence(p->name())) return TRUE; else return FALSE; } static int before_pause(EST_Item *s) { // TRUE is segment immediately after this is a silence if (s->next() == 0) return TRUE; EST_Item *ss = s->as_relation("SylStructure"); EST_Item *n = daughtern(ss)->as_relation("Segment")->next(); if (ph_is_silence(n->name())) return TRUE; else return FALSE; } static EST_Item *vowel_seg(EST_Item *syl) { // return related to vowel segment EST_Item *p; for (p=daughter1(syl,"SylStructure"); p != 0; p=p->next()) if (ph_is_vowel(p->name())) return p; // No vowel found, so return first daughter. return daughter1(syl,"SylStructure"); } festival/src/modules/Text/Makefile0000644000175000017500000000530207262605650015632 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../../.. DIRNAME=src/modules/Text H = tokenP.h NOOPTSRCS = text_modes.cc SRCS = text_aux.cc token.cc text.cc tok_ext.cc \ token_pos.cc xxml.cc $(NOOPTSRCS) OBJS = $(SRCS:.cc=.o) FILES = Makefile $(SRCS) $(H) INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/Text/text_aux.cc0000644000175000017500000000653110126474330016337 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Basic text utilities */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "text.h" EST_Item *add_token(EST_Utterance *u,EST_Token &t) { // Add a token stream item to the end of Token relation EST_Item *item = u->relation("Token")->append(); item->set_name(t.string()); if (t.punctuation() != "") item->set("punc",t.punctuation()); item->set("whitespace",t.whitespace()); item->set("prepunctuation",t.prepunctuation()); return item; } festival/src/modules/Text/token.cc0000644000175000017500000005613511465010356015624 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : November 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Tokenizing */ /* */ /* This provides tokenizing methods for tokens into words. All that */ /* special rules stuff for analysizing numbers, dates, acronyms etc. */ /* Much of this is still too specific and although easy to add to it */ /* be better if the rules could be specified externally */ /* */ /* Note only English tokenization has any substance at present */ /* */ /*=======================================================================*/ #include using namespace std; #include "festival.h" #include "lexicon.h" #include "modules.h" #include "text.h" #include "tokenP.h" static EST_Regex numpointnum("[0-9]*\\.[0-9]+"); static EST_Regex RXintcommaed("[0-9][0-9]?[0-9]?,\\([0-9][0-9][0-9],\\)*[0-9][0-9][0-9]\\(\\.[0-9]+\\)?"); static EST_Regex RXintord("[0-9]*\\(1st\\|2nd\\|3rd\\|[0-9]th\\)"); static EST_Regex RXdottedabbrev("\\([A-Za-z]\\.\\)+[A-Za-z]\\.?"); static EST_Regex RXapostropheS(".*'[sS]$"); static EST_String PunctuationChars("'`.,:;!?{}[]()-\""); static EST_Regex RXpunctuation("\\(\\]\\|[-[.,!?]\\)+"); static EST_String remove_punct(const EST_String &tok); static int only_punc(const EST_String &tok); static LISP num_2_words(int iword); static LISP say_num_as_words(const EST_String &num); static LISP word_it(EST_Item *t,const EST_String tok); static LISP builtin_word_it(EST_Item *token, EST_String tok); static LISP say_as_letters(const EST_String &word); static LISP say_num_as_ordinal(const EST_String &num); static LISP say_num_as_year(const EST_String &num); static LISP say_as_digits(const EST_String &word); LISP FT_English_Token_Utt(LISP utt); LISP FT_Welsh_Token_Utt(LISP utt); LISP FT_Spanish_Token_Utt(LISP utt); LISP FT_Any_Token_Utt(LISP utt); static LISP user_token_to_word_func = NIL; LISP FT_Welsh_Token_Utt(LISP utt) { return FT_Any_Token_Utt(utt); } LISP FT_Spanish_Token_Utt(LISP utt) { (void)utt; cerr << "TOKEN: Spanish tokenization not yet supported\n"; festival_error(); // never happens return NULL; } LISP FT_Any_Token_Utt(LISP utt) { // Language independent EST_Token to Word module. Uses user specified // token to word function of simply creates a word for each token. EST_Utterance *u = get_c_utt(utt); LISP words,w; EST_Item *t; EST_Item *new_word; user_token_to_word_func = siod_get_lval("token_to_words",NULL); u->create_relation("Word"); for (t=u->relation("Token")->first(); t != 0; t = t->next()) { if (user_token_to_word_func != NIL) { words = word_it(t,t->name()); for (w=words; w != NIL; w=cdr(w)) { new_word = add_word(u,car(w)); append_daughter(t,"Token",new_word); } } else { // No user token_to_word function so just do it directly new_word = add_word(u,t->name()); append_daughter(t,"Token",new_word); } } user_token_to_word_func = NIL; // reset this return utt; } LISP FT_English_Token_Utt(LISP utt) { // This module generates a word stream from a token stream // Tokens may go to zero or more words, tokens retain information // about their punctuation and spacing on the page. // Preceeding and succeeding punctuation become words EST_Utterance *u = get_c_utt(utt); EST_Item *t; LISP words,w,eou_tree,l; EST_Item *new_word; *cdebug << "Token module (English)" << endl; eou_tree = siod_get_lval("eou_tree","No end of utterance tree"); user_token_to_word_func = siod_get_lval("token_to_words",NULL); u->create_relation("Word"); for (t=u->relation("Token")->first(); t != 0; t = t->next()) { words = word_it(t,t->name()); // Initial punctuation becomes words new_word = 0; if ((t->f("prepunctuation") != "0") && (t->f("prepunctuation") != "")) { l = symbolexplode(strintern(t->f("prepunctuation").string())); for (w=l; w != NIL; w=cdr(w)) { new_word = add_word(u,car(w)); append_daughter(t,"Token",new_word); } } // Words become words for (w=words; w != NIL; w=cdr(w)) { new_word = add_word(u,car(w)); append_daughter(t,"Token",new_word); } // final word gets punctuation marking if ((new_word != 0) && (ffeature(t,"punc") != "0")) { if ((ffeature(t,"punc") == ".") && (wagon_predict(t,eou_tree) == 0)) { // It wasn't a really punctuation mark t->set("punc","0"); } else { l = symbolexplode(strintern(ffeature(t,"punc").string())); for (w=l; w != NIL; w=cdr(w)) { new_word = add_word(u,car(w)); append_daughter(t,"Token",new_word); } } } } user_token_to_word_func = NIL; return utt; } static LISP word_it(EST_Item *token, const EST_String tok) { // The user may specify their own addition;a token to word rules // through the variable user_token_to_word_func if so we must // call that, which may or may not call the builtin version // The builtin version if bound in LISP so that recursion works // properly // This takes a LISP utt as an argument as creating a new wraparound // will cause gc to fail. LISP tok_string = strcons(tok.length(),tok); if (user_token_to_word_func != NIL) // check user's rules return leval(cons(user_token_to_word_func, cons(siod(token), cons(tok_string,NIL))),NIL); else return builtin_word_it(token,tok); } static LISP builtin_word_it(EST_Item *token, EST_String tok) { // Return a list of words for this token EST_String token_pos; if (tok == "") return NIL; else if (in_current_lexicon(downcase(tok),NIL)) // if in lexicon use as is { if ((tok != token->name()) && // mainly to catch internal "a" (tok.length() == 1)) { LISP let_pos = siod_get_lval("token.letter_pos",NULL); return cons(cons(make_param_str("name",tok), cons(make_param_lisp("pos",let_pos),NIL)), NIL); } else return cons(strintern(tok),NIL); } else if ((token_pos = (EST_String)ffeature(token,"token_pos")) == "ordinal") return say_num_as_ordinal(tok); else if (token_pos == "year") return say_num_as_year(tok); else if ((token_pos == "digits") || (tok.matches(make_regex("0[0-9]+")))) return say_as_digits(tok); else if (tok.matches(RXint)) return say_num_as_words(tok); else if (tok.matches(RXintord)) return say_num_as_ordinal(tok.at(0,tok.length()-2)); else if (tok.matches(RXintcommaed)) // containing commas at thousands { if (tok.contains(".")) return word_it(token,remove_punct(tok.before("."))+ "."+tok.after(".")); else return say_num_as_words(remove_punct(tok)); } else if (tok.matches(RXapostropheS)) { return append(word_it(token,tok.at(0,tok.length()-2)), cons(strintern("'s"),NIL)); } else if (tok.matches(numpointnum)) { EST_String afterpoint = tok.after("."); LISP ap = NIL; int i; for (i=0; i < afterpoint.length(); i++) ap = append(say_num_as_words(afterpoint.at(i,1)),ap); return append(say_num_as_words(tok.before(".")), cons(strintern("point"),reverse(ap))); } else if ((tok.matches(make_regex("[A-Z][A-Z]+"))) && ((!tok.contains(make_regex("[AEIOUY]"))) || ((!tok.contains(make_regex("[^AEIOU][AEIOU][^AEIOU]"))) && (tok.length() < 5)))) // an acronym return say_as_letters(tok); else if (tok.matches(RXdottedabbrev)) return say_as_letters(remove_punct(tok)); else if ((tok.matches(RXalpha)) && !(tok.matches(make_regex(".*[AEIOUYaeiouy].*")))) return say_as_letters(tok); // no vowels so spell it else if (tok.matches(RXalpha)) // as is, some sort of word return cons(strintern(tok),NIL); else if (only_punc(tok)) return stringexplode(tok); else if (tok.contains("-")) return append(word_it(token,tok.before("-")), word_it(token,tok.after("-"))); else if (tok.contains(".")) // internet address { LISP a=NIL; EST_String tok2 = tok; for ( ; tok2.contains("."); tok2 = tok2.after(".")) a = append(a, append(word_it(token,tok2.before(".")), cons(strintern("dot"),NIL))); a = append(a,word_it(token,tok2)); return a; } else if (tok.contains("/")) return append(word_it(token,tok.before("/")), cons(strintern("slash"), word_it(token,tok.after("/")))); else if (tok.contains("&")) return append(word_it(token,tok.before("&")), cons(strintern("ampersand"), word_it(token,tok.after("&")))); else if (tok.contains("_")) return append(word_it(token,tok.before("_")), cons(strintern("underscore"), word_it(token,tok.after("_")))); else if (tok.contains("'")) return word_it(token,tok.before("'")+tok.after("'")); else if (tok.contains("`")) return append(word_it(token,tok.before("`")), word_it(token,tok.after("`"))); else if (tok.contains("\"")) return append(word_it(token,tok.before("\"")), word_it(token,tok.after("\""))); else if (tok.contains(",")) return append(word_it(token,tok.before(",")), word_it(token,tok.after(","))); else if (tok.contains("(")) return append(word_it(token,tok.before("(")), word_it(token,tok.after("("))); else if (tok.contains(")")) return append(word_it(token,tok.before(")")), word_it(token,tok.after(")"))); else if (tok.matches(make_regex("^[^a-zA-Z].+"))) // incrementally remove return append(say_as_letters(tok.at(0,1)),// num/symbol from front word_it(token,tok.at(1,tok.length()-1))); else if (tok.matches(make_regex(".+[^a-zA-Z]$"))) // incrementally remove rear return append(word_it(token,tok.at(0,tok.length()-1)), say_as_letters(tok.at((int)tok.length()-1,1))); else // could try harder return say_as_letters(remove_punct(tok)); } static LISP say_as_digits(const EST_String &word) { // Should be string of digits, but I wont require it // This isn't really correct for telephone numbers (oh/zero/double) LISP l; LISP lets = stringexplode(word); LISP let_pos = siod_get_lval("token.letter_pos",NULL); for (l=lets; l != NIL; l=cdr(l)) { if (streq(get_c_string(car(l)),"0")) CAR(l) = strintern("zero"); else if (streq(get_c_string(car(l)),"1")) CAR(l) = strintern("one"); else if (streq(get_c_string(car(l)),"2")) CAR(l) = strintern("two"); else if (streq(get_c_string(car(l)),"3")) CAR(l) = strintern("three"); else if (streq(get_c_string(car(l)),"4")) CAR(l) = strintern("four"); else if (streq(get_c_string(car(l)),"5")) CAR(l) = strintern("five"); else if (streq(get_c_string(car(l)),"6")) CAR(l) = strintern("six"); else if (streq(get_c_string(car(l)),"7")) CAR(l) = strintern("seven"); else if (streq(get_c_string(car(l)),"8")) CAR(l) = strintern("eight"); else if (streq(get_c_string(car(l)),"9")) CAR(l) = strintern("nine"); else CAR(l) = cons(make_param_lisp("name",car(l)), cons(make_param_lisp("pos",let_pos),NIL)); } return lets; } static LISP say_as_letters(const EST_String &word) { // Explode letters in word and say them, marking them as nouns // This is particularly designed so that A/a doesn't come out as // the determiner a which in typically schwa'd LISP l; LISP lets = stringexplode(word); LISP let_pos = siod_get_lval("token.letter_pos",NULL); for (l=lets; l != NIL; l=cdr(l)) { EST_String name = EST_String(get_c_string(car(l))); if (name.matches(make_regex("[0-9]"))) CAR(l) = car(say_as_digits(get_c_string(car(l)))); // else if (name.matches(make_regex("[^a-zA-Z]"))) // // Not sure, probably a bug to get here // CAR(l) = cons(make_param_str("name","symbol"), // cons(make_param_lisp("pos",let_pos),NIL)); else CAR(l) = cons(make_param_lisp("name",car(l)), cons(make_param_lisp("pos",let_pos),NIL)); } return lets; } static int only_punc(const EST_String &tok) { // If this token consists solely of punctuation chars // If this is true I'm probably suppose to say some of them int i; EST_String np; const char *tokch = tok; for (i=0; i 9) { if (num(0) == '-') return cons(strintern("minus"),say_as_digits(num.after("-"))); else return say_as_digits(num); } else return num_2_words(atoi(num)); } static LISP say_num_as_year(const EST_String &num) { int iword = atoi(num); if (num.length() > 4) return say_num_as_words(num); else if (num.matches(make_regex("00"))) { return cons(strintern("o"), cons(strintern("o"),NIL)); } else if (num.matches(make_regex("0[0-9]"))) { return cons(strintern("o"), num_2_words(iword)); } else if (iword < 100) return num_2_words(iword); else if ((iword % 1000) < 10) { if ((iword % 1000) == 0) return append(num_2_words(iword/1000), cons(strintern("thousand"),NIL)); else return append(num_2_words(iword/1000), cons(strintern("thousand"), cons(strintern("and"), num_2_words(iword%1000)))); } else if ((iword % 100) == 0) return append(num_2_words(iword/100), cons(strintern("hundred"),NIL)); else if ((iword % 100) < 10) return append(num_2_words(iword/100), cons(strintern("o"), num_2_words(iword%100))); else return append(num_2_words(iword/100), num_2_words(iword%100)); } static LISP num_2_words(int iword) { // Convert number of list of words int tens, units; LISP s_tens, lang_stype=NIL; if (iword < 0) return cons(strintern("minus"),num_2_words(-iword)); else if (iword < 20) switch (iword) { case 0: return cons(strintern("zero"),NIL); case 1: return cons(strintern("one"),NIL); case 2: return cons(strintern("two"),NIL); case 3: return cons(strintern("three"),NIL); case 4: return cons(strintern("four"),NIL); case 5: return cons(strintern("five"),NIL); case 6: return cons(strintern("six"),NIL); case 7: return cons(strintern("seven"),NIL); case 8: return cons(strintern("eight"),NIL); case 9: return cons(strintern("nine"),NIL); case 10: return cons(strintern("ten"),NIL); case 11: return cons(strintern("eleven"),NIL); case 12: return cons(strintern("twelve"),NIL); case 13: return cons(strintern("thirteen"),NIL); case 14: return cons(strintern("fourteen"),NIL); case 15: return cons(strintern("fifteen"),NIL); case 16: return cons(strintern("sixteen"),NIL); case 17: return cons(strintern("seventeen"),NIL); case 18: return cons(strintern("eighteen"),NIL); case 19: return cons(strintern("nineteen"),NIL); default: return cons(siod_get_lval("token.unknown_word_name",NULL), NIL); } else if (iword < 100) { tens = iword / 10; units = iword % 10; switch (tens) { case 2: s_tens = strintern("twenty"); break; case 3: s_tens = strintern("thirty"); break; case 4: s_tens = strintern("forty"); break; case 5: s_tens = strintern("fifty"); break; case 6: s_tens = strintern("sixty"); break; case 7: s_tens = strintern("seventy"); break; case 8: s_tens = strintern("eighty"); break; case 9: s_tens = strintern("ninety"); break; default: return cons(siod_get_lval("token.unknown_word_name",NULL), NIL); } if (units != 0) return cons(s_tens,num_2_words(units)); else return cons(s_tens,NIL); } else if (iword < 1000) { lang_stype = ft_get_param("Language"); if (streq("americanenglish",get_c_string(lang_stype))) return append(num_2_words(iword/100), cons(strintern("hundred"), (((iword % 100) != 0) ? num_2_words(iword % 100) : NIL))); else return append(num_2_words(iword/100), cons(strintern("hundred"), (((iword % 100) != 0) ? cons(strintern("and"), num_2_words(iword % 100)) : NIL))); } #if 0 // We don't depend on this hack now. else if ((iword > 1910) && (iword < 2000)) // hacky date condition return append(num_2_words(iword/100), num_2_words(iword%100)); #endif else if (iword < 1000000) return append(num_2_words(iword/1000), cons(strintern("thousand"), (((iword % 1000) != 0) ? ((((iword % 1000)/100) == 0) ? cons(strintern("and"),num_2_words(iword % 1000)): num_2_words(iword % 1000)) : NIL))); else if (iword >= 1000000) return append(num_2_words(iword/1000000), cons(strintern("million"), (((iword % 1000000) != 0) ? num_2_words(iword % 1000000) : NIL))); else return cons(strintern("bignum"),NIL); } static LISP l_word_it(LISP token, LISP tok) { // Lisp wrap around for word_it EST_Item *t = get_c_item(token); EST_String tok_name = get_c_string(tok); return builtin_word_it(t,tok_name); } void festival_token_init(void) { festival_def_utt_module("Token_English",FT_English_Token_Utt, "(Token_English UTT)\n\ Build a Word stream from the Token stream, for English (American and\n\ British English), analyzing compound words, numbers, etc. as tokens\n\ into words."); festival_def_utt_module("Token_Welsh",FT_Welsh_Token_Utt, "(Token_Welsh UTT)\n\ Build a Word stream from the Token stream, for Welsh, analyzing\n\ compound words, numbers etc as tokens into words."); festival_def_utt_module("Token_Spanish",FT_Spanish_Token_Utt, "(Token_Spanish UTT)\n\ Build a Word stream from the Token stream, for Castillian Spanish,\n\ analyzing compound words, numbers etc as tokens into words."); festival_def_utt_module("Token_Any",FT_Any_Token_Utt, "(Token_Any UTT)\n\ Build a Word stream from the Token stream, in a language independent way,\n\ which means that all simple tokens should be in the lexicon, or analysed\n\ by letter to sound rules."); festival_def_utt_module("Token_POS",FT_Token_POS_Utt, "(Token_POS UTT)\n\ Assign feature token_pos to tokens thats match CART trees in the\n\ variable token_pos_cart_trees. These are used for gross level pos\n\ such as identifying how numbers should be pronunced."); init_subr_2("builtin_english_token_to_words",l_word_it, "(english_token_to_words TOKENSTREAM TOKENNAME)\n\ Returns a list of words expanded from TOKENNAME. Note that as this\n\ function may be called recursively TOKENNAME may not be the name of\n\ TOKENSTREAM."); } festival/src/modules/Text/text.cc0000644000175000017500000002512411465010356015462 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Basic text utilities */ /* */ /* This seems to be the only language specific part that cannot be */ /* reasonably parameterized. I'd like to change this but I'm not sure */ /* of the best way. Language-specific token processing module */ /* generating Words (lexical items) from Tokens are current written as */ /* FT_*_Token_Utt functions. A language-independent one is available */ /* FT_Any_Token_Utt which depends heavily on the lexicon can be used */ /* when you don't have the language specific version. */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "text.h" static void tts_raw_token(EST_Item *t); static void tts_raw_utt(LISP utt); LISP FT_Text_Utt(LISP utt) { // Parse text into words EST_Utterance *u = get_c_utt(utt); EST_String text; EST_TokenStream ts; LISP ws,punc,scs; EST_Token tok; *cdebug << "Text module\n"; text = get_c_string(utt_iform(*u)); u->create_relation("Token"); ts.open_string(text); ts.set_SingleCharSymbols(EST_Token_Default_SingleCharSymbols); ts.set_PunctuationSymbols(EST_Token_Default_PunctuationSymbols); ts.set_PrePunctuationSymbols(EST_Token_Default_PrePunctuationSymbols); if ((ws = siod_get_lval("token.whitespace",NULL)) == NIL) ts.set_WhiteSpaceChars(EST_Token_Default_WhiteSpaceChars); else ts.set_WhiteSpaceChars(get_c_string(ws)); if ((punc = siod_get_lval("token.punctuation",NULL)) == NIL) ts.set_PunctuationSymbols(EST_Token_Default_PunctuationSymbols); else ts.set_PunctuationSymbols(get_c_string(punc)); if ((punc = siod_get_lval("token.prepunctuation",NULL)) == NIL) ts.set_PrePunctuationSymbols(EST_Token_Default_PrePunctuationSymbols); else ts.set_PrePunctuationSymbols(get_c_string(punc)); if ((scs = siod_get_lval("token.singlecharsymbols",NULL)) == NIL) ts.set_SingleCharSymbols(EST_Token_Default_SingleCharSymbols); else ts.set_SingleCharSymbols(get_c_string(scs)); for (ts >> tok; tok.string() != ""; ts >> tok) add_token(u,tok); return utt; } LISP tts_file(LISP filename,LISP mode) { LISP user_text_modes,t_mode; user_text_modes = siod_get_lval("tts_text_modes",NULL); if ((mode == NIL) || (streq(get_c_string(mode),"text")) || (streq(get_c_string(mode),"fundamental"))) tts_file_raw(filename); // Simple text file else { t_mode = siod_assoc_str(get_c_string(mode),user_text_modes); if (t_mode == NIL) { // Attempt to load it leval(cons(rintern("request"), cons(strintern(EST_String(get_c_string(mode))+ "-mode"),NIL)),NIL); // get it again, and see if its defined user_text_modes = siod_get_lval("tts_text_modes",NULL); } t_mode = siod_assoc_str(get_c_string(mode),user_text_modes); if (t_mode == NIL) { cerr << "tts_file: can't find mode description \"" << get_c_string(mode) << "\" using raw mode instead" << endl; tts_file_raw(filename); // so read it as simple text file } else tts_file_user_mode(filename,car(cdr(t_mode))); } return NIL; } void tts_file_raw(LISP filename) { // Say the contents of a named file EST_TokenStream ts; LISP ws,prepunc,punc,scs; LISP lutt,eou_tree; LISP stream = NULL; stream = fopen_c(get_c_string(filename), "rb"); if (ts.open(stream->storage_as.c_file.f, FALSE) == -1) { cerr << "tts_file: can't open file \"" << filename << "\"\n"; festival_error(); } ts.set_SingleCharSymbols(EST_Token_Default_SingleCharSymbols); ts.set_PunctuationSymbols(EST_Token_Default_PunctuationSymbols); ts.set_PrePunctuationSymbols(EST_Token_Default_PrePunctuationSymbols); if ((ws = siod_get_lval("token.whitespace",NULL)) == NIL) ts.set_WhiteSpaceChars(EST_Token_Default_WhiteSpaceChars); else ts.set_WhiteSpaceChars(get_c_string(ws)); if ((punc = siod_get_lval("token.punctuation",NULL)) == NIL) ts.set_PunctuationSymbols(EST_Token_Default_PunctuationSymbols); else ts.set_PunctuationSymbols(get_c_string(punc)); if ((prepunc = siod_get_lval("token.prepunctuation",NULL)) == NIL) ts.set_PrePunctuationSymbols(EST_Token_Default_PrePunctuationSymbols); else ts.set_PrePunctuationSymbols(get_c_string(prepunc)); if ((scs = siod_get_lval("token.singlecharsymbols",NULL)) == NIL) ts.set_SingleCharSymbols(EST_Token_Default_SingleCharSymbols); else ts.set_SingleCharSymbols(get_c_string(scs)); eou_tree = siod_get_lval("eou_tree","No end of utterance tree set"); lutt = tts_chunk_stream(ts,tts_raw_token,tts_raw_utt,eou_tree,0); // The last one is returned because the chunker doesn't know if this // is truly the end of an utterance or not, but here we do know. tts_raw_utt(lutt); ts.close(); if (stream) fclose_l(stream); } static void tts_raw_token(EST_Item *t) { // Do something to token, in this case nothing (void)t; } static void tts_raw_utt(LISP utt) { // Do (simple) tts on this utt LISP lutt; // There are some pessimal cases when the utterance is empty if ((utt == NIL) || (get_c_utt(utt)->relation("Token")->length() == 0)) return; // in this case do nothing. lutt = quote(utt); lutt = cons(rintern("apply_hooks"), cons(rintern("tts_hooks"), cons(lutt,NIL))); lutt = cons(rintern("set!"), cons(rintern("utt_tts"), cons(lutt,NIL))); // Synth and Play it lutt = leval(lutt,NIL); user_gc(NIL); } LISP new_token_utt(void) { // An empty utterance ready to take Tokens EST_Utterance *u = new EST_Utterance; u->f.set("type","Tokens"); u->create_relation("Token"); return siod(u); } LISP tts_chunk_stream(EST_TokenStream &ts, TTS_app_tok app_tok, TTS_app_utt app_utt, LISP eou_tree, LISP utt) { // Get tokens from ts and cummulate them in u. // Apply app_tok to each token // Apply app_utt to each utt signalled // Return untermitated utterance potentially for next call // Uses the wagon tree eou_tree to predict utterance termination on // penultimate token. EST_Item *tok, *ebo; EST_Token t; if (utt == NIL) utt = new_token_utt(); EST_Utterance *u = get_c_utt(utt); while (!ts.eof()) { t = ts.get(); tok = add_token(u,t); app_tok(tok); // do what you do with the token ebo = as(tok,"Token")->prev(); // end but one token if ((ebo != 0) && (wagon_predict(ebo,eou_tree) == 1)) { // Remove that extra token remove_item(tok,"Token"); app_utt(utt); // do what you do with the utt utt = new_token_utt(); u = get_c_utt(utt); add_token(u,t); // add that last token to the new utt. } } return utt; } #if 0 LISP memon(void) { printf("memon\n"); putenv("MALLOC_TRACE=mallfile"); mtrace(); return NIL; } LISP memoff(void) { muntrace(); printf("memoff\n"); return NIL; } #endif void festival_Text_init(void) { festival_token_init(); festival_def_utt_module("Text",FT_Text_Utt, "(Text UTT)\n\ From string in input form tokenize and create a token stream."); init_subr_2("tts_file",tts_file, "(tts_file FILE MODE)\n\ Low level access to tts function, you probably want to use the function\n\ tts rather than this one. Render data in FILE as speech. Respect\n\ MODE. Currently modes are defined through the variable tts_text_modes."); #if 0 init_subr_0("memon",memon, "(tts_file FILE MODE)"); init_subr_0("memoff",memoff, "(tts_file FILE MODE)"); #endif init_subr_3("extract_tokens",extract_tokens, "(extract_tokens FILE TOKENS OUTFILE)\n\ Find all occurrences of TOKENS in FILE and output specified context around\n\ the token. Results are appended to OUTFILE, if OUTFILE is nil, output\n\ goes to stdout."); } festival/src/modules/Text/tok_ext.cc0000644000175000017500000001462411465010356016156 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : February 1997 */ /*-----------------------------------------------------------------------*/ /* */ /* EST_Token extract */ /* */ /* Extract tokens and related features from text files. */ /* */ /* This is designed to find examples of particular tokens from large */ /* text corpora (10s millions of words). This initial use is for */ /* Resolving homographs using David Yarowsky's techniques. */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "lexicon.h" #include "text.h" static int rhc = 10; static int lhc = 10; static void search_file(const EST_String &filename, LISP tokens, LISP ofile); static EST_Item *next_token(EST_TokenStream &ts, EST_Relation &ps, EST_Item *s); static void append_token(EST_Relation &ps, const EST_Token &s); static void output_find(const EST_String &filename, EST_Item *s,LISP v,LISP tfeats, FILE* fd); LISP extract_tokens(LISP file, LISP tokens, LISP ofile) { // Extract occurrences of tokens in file search_file(get_c_string(file),tokens, ofile); return NIL; } static void search_file(const EST_String &filename, LISP tokens, LISP ofile) { // Step through the file token by token with a window of tokens // output info when current token matchs on of those required EST_TokenStream ts; EST_Relation ps; EST_Item *s = 0; LISP l,v; FILE *ofd; if (ts.open(filename) == -1) { cerr << "Extract_tokens: can't open file \"" << filename << "\" for reading\n"; festival_error(); } ts.set_PunctuationSymbols(EST_Token_Default_PunctuationSymbols); ts.set_PrePunctuationSymbols(EST_Token_Default_PrePunctuationSymbols); if (ofile == NIL) ofd = stdout; else if ((ofd = fopen(get_c_string(ofile),"a")) == NULL) { cerr << "extract_tokens: cannot open \"" << get_c_string(ofile) << "\" for appending" << endl; festival_error(); } for (s = next_token(ts,ps,s); s != 0; s=next_token(ts,ps,s)) { for (l=tokens; l != NIL; l=cdr(l)) // if (s->name().matches(make_regex(get_c_string(car(car(l)))))) { v = leval(cons(car(car(l)),cons(siod(s),NIL)),NIL); if (v != NIL) output_find(filename,s,v,car(l),ofd); } } ts.close(); if (ofd != stdout) fclose(ofd); } static void output_find(const EST_String &filename, EST_Item *s,LISP v,LISP tfeats, FILE* fd) { // Found a match so output info LISP t; fprintf(fd,"%s %s ",get_c_string(v), (const char *)filename); for (t=cdr(tfeats); t != NIL; t=cdr(t)) fprintf(fd,"%s ",(const char *) ffeature(s,get_c_string(car(t))).string()); fprintf(fd,"\n"); } static EST_Item *next_token(EST_TokenStream &ts, EST_Relation &ps, EST_Item *s) { // return next EST_Token as stream item extending right hand context // and deleting left hand context as required. EST_Item *ns; int i; if (s == 0) { // at start so fill rhc for (i=0; i < lhc; i++) append_token(ps,EST_Token("*lhc*")); append_token(ps,ts.get()); ns = ps.last(); for (i=0; i < rhc; i++) append_token(ps,ts.get()); return ns; } // As token can be "" if there is trailing whitespace before eof // I ignore it. if ((!ts.eof()) && (ts.peek() != "")) append_token(ps,ts.get()); remove_item(ps.first(),"Token"); return s->next(); } static void append_token(EST_Relation &ps, const EST_Token &t) { // Append s as stream cell on end of ps EST_Item *item = ps.append(); item->set_name(t.string()); item->set("filepos",t.filepos()); } festival/src/modules/Text/token_pos.cc0000644000175000017500000000745511465010356016506 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : March 1997 */ /*-----------------------------------------------------------------------*/ /* */ /* This assigns a form of part of speech to tokens. It is typically use */ /* to assign gross level pos for things before they are converted to */ /* words, particularly, numbers (ordinals, digitized, years, numbers) */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "tokenP.h" LISP FT_Token_POS_Utt(LISP utt) { // This module assigns token_pos feature to each token based on the // Assoc list of Regex to CART trees in token_pos_cart_trees. EST_Utterance *u = get_c_utt(utt); EST_Item *t; LISP trees,l; trees = siod_get_lval("token_pos_cart_trees",NULL); if (trees == NIL) return utt; for (t=u->relation("Token")->first(); t != 0; t = t->next()) { if (t->f("token_pos","0") == "0") for (l=trees; l != NIL; l=cdr(l)) // find a tree that matches { if (t->name().matches(make_regex(get_c_string(car(car(l)))))) { t->set_val("token_pos", wagon_predict(t,car(cdr(car(l))))); break; } } } return utt; } festival/src/modules/Text/xxml.cc0000644000175000017500000002332607713311505015471 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : August 1997 */ /*-----------------------------------------------------------------------*/ /* */ /* There are just too many different versions of sgml-based mark up */ /* and none of them are stable so this is allows arbitrary */ /* of tags to Lisp functions so any of them can be implemented in Lisp */ /* That is people can worry about the actual content later and do not */ /* need to change the C++ */ /* */ /* Of course once I give you this functionality you'll just want more */ /* */ /*=======================================================================*/ #include "EST_unix.h" #include "festival.h" #include "text.h" #include "lexicon.h" static LISP xxml_get_attribute(const EST_String &remainder); static char *xxml_process_line(const char *line); static void tts_xxml_token(EST_Item *t); static void tts_xxml_utt(LISP lutt); static LISP xxml_word_features = NIL; static LISP xxml_token_hooks = NIL; void tts_file_xxml(LISP filename) { // For stml ssml rsml and jsml etc // filename contains *output* from something like nsgml EST_String inname = get_c_string(filename); EST_String line, type, remainder; EST_TokenStream ts; LISP atts, element_defs; LISP utt = NIL; // for cummulation of tokens if (ts.open(inname) == -1) { cerr << "xxml: unable to open output from SGML parser" << endl; festival_error(); } ts.set_WhiteSpaceChars(" \t\r\n"); ts.set_SingleCharSymbols(""); ts.set_PunctuationSymbols(""); ts.set_PrePunctuationSymbols(""); element_defs = siod_get_lval("xxml_elements",NULL); atts = NIL; if (ts.peek() != get_c_string(car(car(element_defs)))) { cerr << "xxml parse error: " << get_c_string(filename) << " Expected " << get_c_string(car(car(element_defs))) << " but found " << ts.peek() << endl; festival_error(); } while (ts.peek() != get_c_string(car(car(cdr(element_defs))))) { if (ts.eof()) { cerr << "xxml parse error: unexpected end of file \n"; festival_error(); } line = (EST_String)ts.get_upto_eoln(); type = line.at(0,1); remainder = line.after(0); if (type == "-") { // Segments into utterances as it goes along utt = xxml_get_tokens(remainder, siod_get_lval("xxml_word_features",NULL), utt); } else if (type == "A") // general attribute { atts = cons(xxml_get_attribute(remainder),atts); } else if ((type == "(") || (type == ")")) { utt = xxml_call_element_function(type+remainder,atts, element_defs,utt); atts = NIL; } else { cerr << "xxml parse error: unexpected token found " << line << endl; festival_error(); } } // Last call (should synthesize trailing tokens utt = xxml_call_element_function(ts.get().string(),atts,element_defs,utt); ts.close(); } LISP xxml_call_element_function(const EST_String &element, LISP atts, LISP elements, LISP utt) { // Form the call to the defined element function, with the attributes // and the utterance, returns the utterance LISP def,l; def = siod_assoc_str(element,elements); if (def != NIL) { // You get two arguments, ATTLIST and UTTERANCE l = cons( make_param_lisp("ATTLIST", cons(rintern("quote"),cons(atts,NIL))), cons( make_param_lisp("UTT", cons(rintern("quote"),cons(utt,NIL))), NIL)); return leval(cons(rintern("let"), cons(l,cdr(cdr(def)))),NIL); } else // no definition to do nothing return utt; } static LISP xxml_get_attribute(const EST_String &remainder) { EST_TokenStream ts; LISP tokens=NIL,att=NIL; EST_String name; EST_Token t; ts.open_string(remainder); name = (EST_String)ts.get(); if ((t=ts.get()) == "IMPLIED") att = cons(rintern(name),cons(NIL,NIL)); else if (t == "TOKEN") { EST_Token v = ts.get(); att = cons(rintern(name),cons(cons(rintern(v.string()),NIL),NIL)); } else if (t == "CDATA") { while (!ts.eof()) tokens = cons(rintern(ts.get().string()),tokens); att = cons(rintern(name),cons(reverse(tokens),NIL)); } else { cerr << "XXML: unknow attribute type " << remainder << endl; festival_error(); } ts.close(); return att; } static char *xxml_process_line(const char *line) { // STML (sgml) data line have a number of special escape characters // this undoes them, namely "\\n" to "\n" char *procline = walloc(char,strlen(line)+1); int i,j; for (i=j=0; line[i] != '\0'; j++,i++) { if (line[i] == '\\') { i++; if (line[i] == 'n') procline[j] = '\n'; else if (line[i] == '\\') procline[j] = '\\'; else if ((line[i] == '0') || // its an octal number (line[i] == '1')) { int k,oct = 0; for (k=0; k < 3; k++,i++) oct = (oct*8)+(line[i]-'0'); procline[j] = oct; i--; } else { procline[j] = line[i]; // no change i--; } } else procline[j] = line[i]; // no change } procline[j] = '\0'; return procline; } static void tts_xxml_token(EST_Item *t) { // Add xxml_word features to t LISP a; for (a=xxml_word_features; a != NIL; a=cdr(a)) if ((car(cdr(car(a))) != NIL) && (!streq("NAME",get_c_string(car(car(a)))))) { if (cdr(cdr(car(a))) == NIL) t->set(get_c_string(car(car(a))), get_c_string(car(cdr(car(a))))); else { // Its more complex than a single atom so save the list t->set(get_c_string(car(car(a))), siod_sprint(car(cdr(car(a))))); } } apply_hooks(xxml_token_hooks,siod(t)); } LISP xxml_get_tokens(const EST_String &line,LISP feats,LISP utt) { // Read from here until end of line collects all the tokens // Note tokens are in reverse order until they are made into an // utterance EST_TokenStream ls; EST_Token t; LISP eou_tree; char *processed_line; processed_line = xxml_process_line(line); ls.open_string(processed_line); ls.set_SingleCharSymbols( get_c_string(siod_get_lval("token.singlecharsymbols", "token.singlecharsymbols unset"))); ls.set_PunctuationSymbols( get_c_string(siod_get_lval("token.punctuation", "token.punctuation unset"))); ls.set_PrePunctuationSymbols( get_c_string(siod_get_lval("token.prepunctuation", "token.prepunctuation unset"))); ls.set_WhiteSpaceChars( get_c_string(siod_get_lval("token.whitespace", "token.whitespace unset"))); eou_tree = siod_get_lval("eou_tree","No end of utterance tree set"); xxml_word_features = feats; xxml_token_hooks = siod_get_lval("xxml_token_hooks",NULL); // Segment and synth as much as appropriate utt = tts_chunk_stream(ls,tts_xxml_token,tts_xxml_utt,eou_tree,utt); return utt; } static void tts_xxml_utt(LISP lutt) { // Build and utterance with these tokens and apply xxml synth function if ((lutt == NIL) || (get_c_utt(lutt)->relation("Token")->length() == 0)) return; // in this case do nothing. leval(cons(rintern("xxml_synth"), cons(quote(lutt),NIL)),NIL); } festival/src/modules/Text/text_modes.cc0000644000175000017500000001470611223441730016651 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : November 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Support for general user tts modes */ /* Each mode consists of user definable parameters for (at least) the */ /* the following */ /* filter external Unix program filter */ /* utterance chunk tree: decision tree to determine end of utterance */ /* punctuation */ /* whitespace */ /* token analysis rule */ /* init function run before mode is applied */ /* exit function run after mode is applied */ /* */ /*=======================================================================*/ #include #include "EST_unix.h" #include "festival.h" #include "text.h" #include "lexicon.h" static void um_apply_filter(const EST_String &filtername, const EST_String &infile, const EST_String &outname); void tts_file_user_mode(LISP filename, LISP params) { volatile EST_String tmpname = make_tmp_filename(); volatile EST_String inname = (EST_String)get_c_string(filename); volatile EST_String filter; volatile EST_TokenStream ts; volatile LISP func; jmp_buf *old_errjmp = est_errjmp; int old_errjmp_ok = errjmp_ok; func = get_param_lisp("init_func",params,NIL); if (func != NIL) leval(cons(func,NIL),NIL); // run initial function if specified errjmp_ok = 1; est_errjmp = walloc(jmp_buf,1); if (setjmp(*est_errjmp)) { cerr << "festival: text modes, caught error and tidying up\n"; if (siod_ctrl_c == TRUE) { wfree(est_errjmp); est_errjmp = old_errjmp; errjmp_ok = old_errjmp_ok; err("forwarded ctrl_c",NIL); } } else { filter.ignore_volatile() = get_param_str("filter",params,""); um_apply_filter(filter.ignore_volatile(),inname.ignore_volatile(),tmpname.ignore_volatile()); if (streq("xxml",get_param_str("analysis_type",params,""))) tts_file_xxml(strintern(tmpname.ignore_volatile())); else if (streq("xml",get_param_str("analysis_type",params,""))) { // As xml support is optional we call it through a LISP // function which wont be defined if its not in this installation leval(cons(rintern("tts_file_xml"),cons(strintern(tmpname.ignore_volatile()),NIL)),NIL); } else tts_file_raw(strintern(tmpname.ignore_volatile())); } wfree(est_errjmp); est_errjmp = old_errjmp; errjmp_ok = old_errjmp_ok; unlink(tmpname.ignore_volatile()); func = get_param_lisp("exit_func",params,NIL); if (func != NIL) leval(cons(func,NIL),NIL); // run end function if specified } void um_apply_filter(const EST_String &filtername, const EST_String &infile, const EST_String &outfile) { // filter the file into standard form EST_String command; if (access(infile,R_OK) != 0) { cerr << "TTS user mode: \"" << infile << "\" cannot be accessed" << endl; festival_error(); } if (filtername == "") { // don't bother forking FILE *fdin, *fdout; char buff[256]; int n; if ((fdin = fopen(infile,"rb")) == NULL) { cerr << "TTS user mode: \"" << infile << "\" cannot be read from" << endl; festival_error(); } if ((fdout = fopen(outfile,"wb")) == NULL) { cerr << "TTS user mode: \"" << outfile << "\" cannot be written to" << endl; festival_error(); } while ((n = fread(buff,1,256,fdin)) > 0) fwrite(buff,1,n,fdout); fclose(fdin); fclose(fdout); } else { command = filtername + " '" + infile + "' > " + outfile; system(command); // should test if this is successful or not } } festival/src/modules/Text/tokenP.h0000644000175000017500000000575007262605650015612 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : March 1997 */ /*-----------------------------------------------------------------------*/ /* */ /* Private declarations for tokens */ /* */ /*=======================================================================*/ #ifndef __TOKENP_H__ #define __TOKENP_H__ LISP FT_Token_POS_Utt(LISP utt); #endif festival/src/modules/UniSyn/Makefile0000644000175000017500000000545707263321403016136 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../../.. DIRNAME=src/modules/UniSyn H = UniSyn.h us_features.h us_synthesis.h TSRCS = CPPSRCS = UniSyn.cc us_prosody.cc us_unit.cc ps_synthesis.cc \ us_mapping.cc us_features.cc $(TSRCS) SRCS = $(CPPSRCS) OBJS = $(CPPSRCS:.cc=.o) FILES = Makefile $(SRCS) $(H) INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules UniSyn.o: $(CXX_COMMAND) $(TDPSOLA_FLAGS) UniSyn.cc -o UniSyn.o festival/src/modules/UniSyn/UniSyn.cc0000644000175000017500000003006611465010356016225 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Paul Taylor */ /* Date: February 1998 */ /* --------------------------------------------------------------------- */ /* Waveform Generation Scheme Interface File */ /* */ /*************************************************************************/ #include "siod.h" #include "EST.h" #include "UniSyn.h" #include "us_synthesis.h" #include "Phone.h" VAL_REGISTER_TYPE(ivector,EST_IVector) VAL_REGISTER_TYPE(wavevector,EST_WaveVector); SIOD_REGISTER_TYPE(wavevector, EST_WaveVector); void map_to_relation(EST_IVector &map, EST_Relation &r, const EST_Track &source_pm, const EST_Track &target_pm); EST_Features *scheme_param(const EST_String& param, const EST_String &path) { EST_Features *f, *p; f = feats(siod_get_lval(param, "Couldn't find scheme paramete named: " + param)); p = (path == "") ? f : &f->A(path); return p; } LISP FT_us_linear_smooth_amplitude( LISP lutt ) { EST_Utterance *utt = get_c_utt( lutt ); us_linear_smooth_amplitude( utt ); return lutt; } static LISP FT_wavevector_get_wave( LISP l_wavevector, LISP l_framenum ) { EST_WaveVector *wv = wavevector( l_wavevector ); int i = get_c_int( l_framenum ); if( i<0 || i>wv->length() ) EST_error( "index out of bounds" ); return siod( &((*wv)[i]) ); } LISP FT_us_unit_concat(LISP lutt) { EST_String window_name; float window_factor; bool window_symmetric; EST_Features *f = scheme_param("Param", "unisyn"); window_name = f->S("window_name"); window_factor = f->F("window_factor"); window_symmetric = (f->I("window_symmetric",1) == 0) ? false : true; us_unit_concat(*get_c_utt(lutt), window_factor, window_name, false, window_symmetric); return lutt; } LISP FT_us_unit_raw_concat(LISP lutt) { us_unit_raw_concat(*get_c_utt(lutt)); return lutt; } LISP FT_us_energy_normalise(LISP lutt, LISP lrname) { EST_Utterance *utt = get_c_utt(lutt); EST_String rname = get_c_string(lrname); us_energy_normalise(*utt->relation(rname)); return lutt; } LISP FT_us_generate_wave(LISP lutt, LISP l_f_method, LISP l_o_method) { EST_String filter_method = get_c_string(l_f_method); EST_String ola_method = get_c_string(l_o_method); EST_Utterance *utt = get_c_utt(lutt); EST_Features *f = scheme_param("Param", "unisyn"); if(f->I("window_symmetric",1) == 0){ ola_method = "asymmetric_window"; } us_generate_wave(*utt, filter_method, ola_method); return lutt; } LISP FT_us_mapping(LISP lutt, LISP method) { us_mapping(*get_c_utt(lutt), get_c_string(method)); return lutt; } LISP FT_us_get_copy_wave(LISP lutt, LISP l_sig_file, LISP l_pm_file, LISP l_seg_file) { EST_Utterance *utt = get_c_utt(lutt); EST_Relation seg; EST_String sig_file = get_c_string(l_sig_file); EST_String seg_file = get_c_string(l_seg_file); EST_String pm_file = get_c_string(l_pm_file); EST_Track *pm = new EST_Track; EST_Wave *sig = new EST_Wave; if (pm->load(pm_file) != format_ok) return NIL; if (sig->load(sig_file) != format_ok) return NIL; if (seg.load(seg_file) != format_ok) return NIL; if (!ph_is_silence(seg.tail()->f("name"))) { EST_Item *n = seg.tail()->insert_after(); n->set("name", ph_silence()); n->set("end", seg.tail()->prev()->F("end") + 0.1); } us_get_copy_wave(*utt, *sig, *pm, seg); return lutt; } LISP FT_f0_to_pitchmarks(LISP lutt, LISP l_f0_name, LISP l_pm_name, LISP l_end_time) { EST_Utterance *utt = get_c_utt(lutt); int num_channels=0; const float default_f0 = 100.0; EST_Relation *f0_rel=0, *pm_rel=0; EST_Track *f0=0, *pm=0; EST_Item *a; float end_time = (l_end_time == NIL) ? -1 : get_c_float(l_end_time); f0_rel = utt->relation(get_c_string(l_f0_name), 1); pm_rel = utt->create_relation(get_c_string(l_pm_name)); f0 = track(f0_rel->head()->f("f0")); pm = new EST_Track; a = pm_rel->append(); a->set_val("coefs", est_val(pm)); a = pm_rel->append(); if (utt->relation_present("SourceCoef")) { EST_Track *source_coef = track(utt->relation("SourceCoef")->head()->f("coefs")); num_channels = source_coef->num_channels(); } f0_to_pitchmarks(*f0, *pm, num_channels, default_f0, end_time); return lutt; } LISP FT_map_to_relation(LISP lutt, LISP lsource_name, LISP ltarget_name, LISP lrel_name) { EST_Utterance *utt = get_c_utt(lutt); EST_Track *source_pm = 0; EST_Track *target_pm = 0; EST_IVector *map = 0; target_pm = track(utt->relation(get_c_string(ltarget_name))->head()->f("coefs")); source_pm = track(utt->relation(get_c_string(lsource_name))->head()->f("coefs")); map = ivector(utt->relation("US_map")->head()->f("map")); utt->create_relation(get_c_string(lrel_name)); map_to_relation(*map, *utt->relation(get_c_string(lrel_name)), *source_pm, *target_pm); return NIL; } void festival_UniSyn_init(void) { proclaim_module("UniSyn"); register_unisyn_features(); init_subr_2( "wavevector.getwave", FT_wavevector_get_wave, "(wavevector.getwave WAVEVECTOR FRAMENUM)\n\ retrieves an EST_Wave frame (int FRAMENUM) from a wavevector."); init_subr_1("us_linear_smooth_amplitude", FT_us_linear_smooth_amplitude, "(us_linear_smooth_amplitude UTT)\n\ Perform linear amplitute smoothing on diphone joins."); init_subr_1("us_unit_raw_concat", FT_us_unit_raw_concat, "(us_init_raw_concat UTT)."); init_subr_2("us_energy_normalise", FT_us_energy_normalise, "(us_ps_synthesis UTT SIGPR)\n\ Synthesize utterance UTT using signal processing technique SIGPR \n\ for the UniSyn pitch-synchronous synthesizer."); init_subr_3("us_generate_wave", FT_us_generate_wave, "(us_td_synthesis UTT FILTER_METHOD OLA_METHOD)\n\ Synthesize utterance UTT using signal processing technique SIGPR \n\ for the UniSyn pitch-synchronous synthesizer."); init_subr_2("us_mapping", FT_us_mapping, "(us_mapping UTT method)\n\ Synthesize utterance UTT using signal processing technique SIGPR \n\ for the UniSyn pitch-synchronous synthesizer."); init_subr_1("us_unit_concat", FT_us_unit_concat, "(us_unit_concat UTT)\n\ Concat coef and wave information in unit stream into a single \n\ Frames structure storing the result in the Frame relation"); init_subr_4("us_f0_to_pitchmarks", FT_f0_to_pitchmarks, "(us_f0_to_pitchmarks UTT F0_relation PM_relation END_TIME)\n\ From the F0 contour in F0_relation, create a set of pitchmarks\n\ in PM_relation. If END_TIME is not nil, Extra pitchmarks will be \n\ created at the default interval up to this point"); init_subr_4("map_to_relation", FT_map_to_relation, "(map_to_relation UTT Source_relation Target_relation new_relation)\n\ From the F0 contour in F0_relation, create a set of pitchmarks\n\ in PM_relation. If END_TIME is not nil, Extra pitchmarks will be \n\ created at the default interval up to this point"); init_subr_4("us_get_copy_wave", FT_us_get_copy_wave, "(warp_utterance UTT (Wavefile Pitchmark_file))\n\ Change waveform to match prosodic specification of utterance."); #ifdef HAVE_US_TDPSOLA_TM us_init_tdpsola(); #endif } /* init_subr_2("us_F0targets_to_pitchmarks", FT_us_F0targets_to_pitchmarks, "(us_F0targets_to_pitchmarks UTT Segment_Relation)\n\ Make set of pitchmarks according to F0 target specification"); LISP FT_merge_pitchmarks(LISP lutt, LISP l_pm1, LISP l_pm2, LISP l_guide_name) { EST_Utterance *utt = get_c_utt(lutt); EST_Track *pm1 = track(utt->relation(get_c_string(l_pm1), 1)->head()->f("coefs", 1)); EST_Track *pm2 = track(utt->relation(get_c_string(l_pm2), 1)->head()->f("coefs", 1)); EST_Relation *guide = utt->relation(get_c_string(l_guide_name), 1); EST_Relation *pm_rel = utt->create_relation("TargetCoefs"); EST_Track *target_pm = new EST_Track; EST_Item *a = pm_rel->append(); a->fset_val("coefs", est_val(target_pm)); merge_pitchmarks(*get_c_utt(lutt), *pm1, *pm2, *target_pm, *guide); return lutt; } LISP FT_warp_pitchmarks(LISP lutt, LISP l_pm_file, LISP l_seg_file) { EST_Utterance *utt = get_c_utt(lutt); EST_String pm_file = get_c_string(l_pm_file); EST_String seg_file = get_c_string(l_seg_file); EST_Track *pm = new EST_Track; EST_Relation seg; if (pm->load(pm_file) != format_ok) return NIL; if (seg.load(seg_file) != format_ok) return NIL; warp_pitchmarks(*utt, pm, seg, *utt->relation("Segment")); return lutt; } init_subr_3("us_warp_pitchmarks", FT_warp_pitchmarks, "(warp_utterance UTT (Wavefile Pitchmark_file))\n\ Change waveform to match prosodic specification of utterance."); LISP FT_us_load_utt_segments(LISP l_utt, LISP l_filename) { EST_String filename = get_c_string(l_filename); EST_Utterance tu; EST_Utterance *u = get_c_utt(l_utt); EST_Item *s, *t; if (tu.load(filename) != format_ok) festival_error(); u->relation("Segment")->clear(); for (s = tu.relation("Segment")->head(); s; s = s->next()) { t = u->relation("Segment")->append(); t->fset("name", s->fS("name")); t->fset("end", s->fS("end")); } return l_utt; } void us_F0targets_to_pitchmarks(EST_Utterance &utt, const EST_String &seg_relation); LISP FT_us_F0targets_to_pitchmarks(LISP lutt, LISP lseg) { EST_String s = (lseg == NIL) ? "" : get_c_string(lseg); us_F0targets_to_pitchmarks(*get_c_utt(lutt), s); return lutt; } */ festival/src/modules/UniSyn/us_prosody.cc0000644000175000017500000003704612445314665017224 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Paul Taylor */ /* Date: 6 Jan 1998 */ /* --------------------------------------------------------------------- */ /* UniSyn prosody manipulation functions */ /* */ /*************************************************************************/ #include "us_synthesis.h" #include "Phone.h" //static void add_end_silences(EST_Relation &segment); //static void add_end_silences(EST_Relation &segment, EST_Relation &target); void pitchmarks_to_f0(EST_Track &pm, EST_Track &fz, float shift) { int i; float period; fz.resize((int)(pm.end()/shift), 1); fz.fill_time(shift); for (i = 0; i < fz.num_frames() -1 ; ++i) { period = get_time_frame_size(pm, pm.index_below(fz.t(i))); fz.a(i) = 1.0 /period; } } void f0_to_pitchmarks(EST_Track &fz, EST_Track &pm, int num_channels, float default_f0, float target_end) { int i; float max = 0.0; float fz_end; // Its impossible to guess the length of the pitchmark array before // hand. Here we find the upper limit and resize at the end for (i = 0; i < fz.num_frames(); ++i) { if (fz.a_no_check(i) < 0) fz.a_no_check(i) = 0; if (fz.a_no_check(i) > 500) fz.a_no_check(i) = fz.a_no_check(i-1); if (fz.a_no_check(i) > max) max = fz.a_no_check(i); #if 0 if (fz.a_no_check(i) < 0) fz.a_no_check(i) = 0; if (fz.a_no_check(i) > 750) fz.a_no_check(i) = 749; if (fz.a_no_check(i) > max) max = fz.a_no_check(i); #endif } // Coefficients will also be placed in here, so its best allocate // space for their channels now fz_end = fz.end(); pm.resize(int(max * (Gof(fz_end, target_end))) + 10, num_channels); int fz_len = fz.length(); float t1 = 0.0; //first pitchmark convention float t2; float f1 = fz.a_no_check(0); //arbitrary init float f2; double area = 0.5; // init value int pm_i = 0; int pm_len = pm.length(); for( int i=0; i= 1.0) && (pm_i < pm_len) ){ area -= 1.0; float discriminant = f2*f2 - 2.0 * area * slope; if (discriminant < 0.0) discriminant = 0.0; pm.t(pm_i++) = t2 - 2.0 * area / (f2 + sqrt (discriminant)); } t1 = t2; f1 = f2; } float default_shift = 1.0 / default_f0; if (target_end > fz_end) for (; t1 < target_end; ++pm_i) t1 = pm.t(pm_i) = t1 + default_shift; pm.resize(pm_i-1, num_channels); } /* Convert an F0 contour into a set of pitchmarks. This is done by the obvious iterative function. Space before the first defined F0 value is filled with regularly space pitchmarks at intervals 1/def_f0. If the target_end value is specified, more default pitchmarks are placed after the end of the last f0 value until time target_end has been reached. */ void f0_to_pitchmarks_orig(EST_Track &fz, EST_Track &pm, int num_channels, float default_f0, float target_end) { int i; float max = 0.0, prev_pm = 0.0, val; float fz_end; // cout << "fz end: " << fz.end() << endl; // cout << "fz n fg: " << fz.num_frames() << endl; // Its impossible to guess the length of the pitchmark array before // hand. Here we find the upper limit and resize at the end for (i = 0; i < fz.num_frames(); ++i) { if (fz.a_no_check(i) < 0) fz.a_no_check(i) = 0; if (fz.a_no_check(i) > 750) fz.a_no_check(i) = 749; if (fz.a_no_check(i) > max) max = fz.a_no_check(i); } // Coefficients will also be placed in here, so its best allocate // space for their channels now fz_end = fz.end(); pm.resize(int(max * (Gof(fz_end, target_end))) + 10, num_channels); // cout << "fz end: " << fz.end() << endl; // cout << "fz n fg: " << fz.num_frames() << endl; // cout << "pmn fg: " << pm.num_frames() << endl; for (i = 0; prev_pm < fz_end; ++i) { val = fz.a(prev_pm) > 0.0 ? fz.a(prev_pm) : default_f0; pm.t(i) = prev_pm + (1.0 / val); prev_pm = pm.t(i); } if (target_end > fz_end) for (; prev_pm < target_end; ++i) { pm.t(i) = prev_pm + (1.0 / default_f0); prev_pm = pm.t(i); } pm.resize(i - 1, num_channels); } // not sure if this is useful void linear_pitchmarks(EST_Track &source_pm, EST_Track &target_pm, float start_f0, float end_f0) { int i; float m, length, pitch; target_pm.resize(source_pm.num_frames(), source_pm.num_channels()); length = (float)source_pm.num_frames() / (end_f0 - start_f0); target_pm.t(0) = 0.0; m = (end_f0 - start_f0) / length; for(i = 1; i < target_pm.num_frames(); ++i) { pitch = (((float)i / (float) target_pm.num_frames()) * (end_f0 - start_f0)) + start_f0; target_pm.t(i) = target_pm.t(i - 1) + (1 /pitch); } } // not sure if this is useful void stretch_f0_time(EST_Track &f0, float stretch, float s_last_time, float t_last_time) { for (int i = 0 ; i < f0.num_frames(); ++i) { // cout << i << " o t:" << f0.t(i) << endl; f0.t(i) = ((f0.t(i) - s_last_time) * stretch) + t_last_time; // cout << i << " m t:" << f0.t(i) << endl; } } // make target F0 from source F0, with same F0 values as original, // but durations specified by target_seg. /* void us_F0targets_to_pitchmarks(EST_Utterance &utt, const EST_String &seg_relation) { utt.create_relation("TargetCoef"); EST_Track *target_coef = new EST_Track; EST_Item *end_seg; int num_channels = 0; float end; if (utt.relation_present("SourceCoef")) { EST_Track *source_coef = track(utt.relation("SourceCoef")->head()->f("coefs")); num_channels = source_coef->num_channels(); } if (seg_relation == "") end_seg = utt.relation("Segment", 1)->last(); else end_seg = utt.relation(seg_relation, 1)->last(); if (end_seg) end = end_seg->F("end"); else end = 0; targets_to_pitchmarks(*(utt.relation("Target")), *target_coef, num_channels,end); EST_Item *item = utt.relation("TargetCoef")->append(); item->set("name", "coef"); item->set_val("coefs",est_val(target_coef)); } void targets_to_pitchmarks(EST_Relation &targ, EST_Track &pitchmarks, int num_channels,float end) { EST_Item *s; float time, f0, prev_time, prev_f0, m, max; int i; // Its impossible to guess the length of the pitchmark array before // hand. Here we find the upper limit and resize at the end for (max = 0.0, s = targ.first_leaf(); s; s = next_leaf(s)) if (s->F("f0") > max) max = s->F("f0"); pitchmarks.resize((int)(max * 1.1 * end)+1, num_channels); prev_time = 0; prev_f0 = targ.first_leaf() ? targ.first_leaf()->F("f0") : 120; pitchmarks.t(0) = 0.0; for (i = 1, s = targ.first_leaf(); s; s = next_leaf(s)) { time = s->f("pos"); f0 = s->F("f0"); if (f0 < 30) // to protect against with duff IntTarget algorithms continue; if (time == prev_time) continue; else if (time < prev_time) { cerr << "UniSyn: warning target in wrong order at " << prev_time; cerr << " ignored" << endl; continue; } m = (f0 - prev_f0) / (time - prev_time); { f0 = (m * (pitchmarks.t(i - 1) - prev_time)) + prev_f0; pitchmarks.t(i) = pitchmarks.t(i - 1) + 1.0/f0; } prev_time = time; prev_f0 = f0; } // Ensure pitch marks go to the end of the utterance // This will effectively mean the last half diphone will be extend over // the whol final segment. This will only be reasonable if the // final segment is a silence. for (; pitchmarks.t(i - 1) < end; ++i) pitchmarks.t(i) = pitchmarks.t(i - 1) + 1.0/prev_f0; pitchmarks.resize(i, pitchmarks.num_channels()); } */ /*static void add_end_silences(EST_Relation &segment, EST_Relation &target) { EST_Item *t, *n; float shift = 0.0; const float pause_duration = 0.1; t = segment.head(); if (!ph_is_silence(t->f("name"))) { n = t->insert_before(); n->set("name", ph_silence()); n->set("dur", pause_duration); shift += pause_duration; } t = segment.tail(); if (!ph_is_silence(t->f("name"))) { n = t->insert_after(); n->set("name", ph_silence()); n->set("dur", pause_duration); shift += pause_duration; } dur_to_end(segment); target.tail()->set("pos", (target.tail()->F("pos") + shift)); } void merge_pitchmarks(EST_Utterance &u, EST_Track &pm1, EST_Track &pm2, EST_Track &target_pm, EST_Relation &guide) { EST_Item *s; float s_end, s_start; int s_i_start, s_i_end; int i, j = 0; (void) u; target_pm.resize(1000000, 0); s_start = 0.0; for (s = guide.head(); s; s = s->next()) { s_end = s->F("end", 1); if (s->fI("use_pm") == 1) { s_i_start = pm1.index_below(s_start); s_i_end = pm1.index_below(s_end); for (i = s_i_start; i < s_i_end; ++i, ++j) target_pm.t(j) = pm1.t(i); } else { s_i_start = pm2.index_below(s_start); s_i_end = pm2.index_below(s_end); for (i = s_i_start; i < s_i_end; ++i, ++j) target_pm.t(j) = pm2.t(i); } s_start = s_end; } } void warp_f0(EST_Track &source_f0, EST_Relation &source_seg, EST_Track &target_f0, EST_Relation &target_seg) { EST_Item *s, *t; float prev_source_end = 0.0, prev_target_end = 0.0; EST_Track part; int frame_start, frame_end; float stretch, t_last_time = 0, s_last_time = 0; EST_Relation match("Match"); EST_Item xx; EST_Track str; int i = 0; dp_match(target_seg, source_seg, match, local_cost, &xx); target_f0 = source_f0; frame_start = 0; frame_end = 0; str.resize(target_seg.length(), 1); cout << "tag: " << target_seg << endl; for (t = target_seg.head(); t; t = t->next()) { s = daughter1(t,"Match"); if (s == 0) // ie extra phone in target specification continue; frame_end = source_f0.index(s->f("end")); if ((frame_end - frame_start) < 1) { cout << "Warning no frames for: " << *t << endl; continue; } target_f0.sub_track(part, frame_start, (frame_end - frame_start + 1), 0, EST_ALL); stretch = (t->F("end") - prev_target_end) / (s->F("end") - prev_source_end); str.a(i) = stretch; str.t(i++) = t->F("end"); cout << "\nstretch: " << stretch << endl; cout << "source: " << *s << endl; cout << "target: " << *t << endl; cout << "frames: " << frame_start << " " << frame_end << endl; stretch_f0_time(part, stretch, s_last_time, t_last_time); prev_target_end = t->f("end"); prev_source_end = s->f("end"); frame_start = frame_end + 1; t_last_time = part.end(); s_last_time = source_f0.t(frame_end); cout << "last time = " << s_last_time << " " << t_last_time << endl; } target_f0.resize(frame_end, 1); target_f0.a(target_f0.num_frames() - 1) = 100; str.save("zz_stretch"); } void warp_pitchmarks(EST_Utterance &utt, EST_Track *source_pm, EST_Relation &source_seg, EST_Relation &target_seg) { EST_Track source_f0, target_f0, *target_pm; target_pm = new EST_Track; cout << "tag: "<< target_seg << endl; add_end_silences(target_seg); cout << "tag 2: "<< target_seg << endl; pitchmarks_to_f0(*source_pm, source_f0, 0.01); cout << "tag 3: "<< target_seg << endl; warp_f0(source_f0, source_seg, target_f0, target_seg); f0_to_pitchmarks(target_f0, *target_pm); utt.create_relation("TargetCoef"); utt.create_relation("SourceSegments"); *utt.relation("SourceSegments") = source_seg; EST_Item *item = utt.relation("TargetCoef")->append(); target_f0.save("tt_tar.f0", "est"); target_seg.save("tt_tar.lab"); source_seg.save("tt_sou.lab"); source_f0.save("tt_sou.f0", "est"); target_pm->save("target_coef_a.pm","est"); item->set("name", "coefs"); item->set_val("coefs", est_val(target_pm)); } float local_cost(const EST_Item *s1, const EST_Item *s2) { float insertion_cost = get_c_int(siod_get_lval("met_insertion", NULL)); float deletion_cost = get_c_int(siod_get_lval("met_deletion", NULL)); float substitution_cost = get_c_int(siod_get_lval("met_substitution", NULL)); EST_String null_sym = "nil"; // otherwise cost is either insertion cost, or cost_matrix value if (s1->name() == s2->name()) return 0; else { if (s1->name() == null_sym) return insertion_cost; else if (s2->name() == null_sym) return deletion_cost; else return substitution_cost; } } typedef float (*local_cost_function)(const EST_Item *item1, const EST_Item *item2); bool dp_match(const EST_Relation &lexical, const EST_Relation &surface, EST_Relation &match, local_cost_function lcf, EST_Item *null_syl); */ /*static void add_end_silences(EST_Relation &segment) { EST_Item *t, *n; t = segment.head(); if (!ph_is_silence(t->f("name"))) { n = t->insert_before(); n->set("name", ph_silence()); } t = segment.tail(); if (!ph_is_silence(t->f("name"))) { n = t->insert_after(); n->set("name", ph_silence()); } } */ festival/src/modules/UniSyn/us_unit.cc0000644000175000017500000005031211465010356016462 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Paul Taylor */ /* Date: 6 Jan 1998 */ /* --------------------------------------------------------------------- */ /* Acoustic Unit Concatenation */ /* */ /*************************************************************************/ #include "siod.h" #include "EST_sigpr.h" #include "EST_wave_aux.h" #include "EST_track_aux.h" #include "EST_ling_class.h" #include "us_synthesis.h" #include #include "Phone.h" void merge_features(EST_Item *from, EST_Item *to, int keep_id); void dp_time_align(EST_Utterance &utt, const EST_String &source_name, const EST_String &target_name, const EST_String &time_name, bool do_start); void concatenate_unit_coefs(EST_Relation &unit_stream, EST_Track &source_lpc); void us_unit_raw_concat(EST_Utterance &utt); void window_units(EST_Relation &unit_stream, EST_TVector &frames, float window_factor, EST_String window_name, bool window_symmetric, EST_IVector *pm_indices=0); bool dp_match(const EST_Relation &lexical, const EST_Relation &surface, EST_Relation &match, float ins, float del, float sub); void map_match_times(EST_Relation &target, const EST_String &match_name, const EST_String &time_name, bool do_start); static void window_frame(EST_Wave &frame, EST_Wave &whole, float scale, int start, int end, EST_WindowFunc *window_function, int centre_index=-1) { int i, j, send; EST_TBuffer window; int window_length = (end-start)+1; if (frame.num_samples() != (window_length)) frame.resize(window_length); frame.set_sample_rate(whole.sample_rate()); // Ensure we have a safe end if (end < whole.num_samples()) send = end; else send = whole.num_samples(); int print_centre; if ( centre_index < 0 ){ window_function( window_length, window, -1 ); print_centre = (window_length-1)/2+start; } else{ window_function( window_length, window, (centre_index-start)); print_centre = centre_index; } #if defined(EST_DEBUGGING) cerr << "(start centre end window_length wholewavelen) " << start << " " << print_centre << " " << end << " " << window_length << " " << whole.num_samples() << endl; #endif // To allow a_no_check access we do this in three stages for (i = 0, j = start; j < 0; ++i, ++j) frame.a_no_check(i) = 0; for ( ; j < send; ++i, ++j) frame.a_no_check(i) = (int)((float)whole.a_no_check(j) * window(i) * scale); for ( ; j < end; ++j,++i) frame.a_no_check(i) = 0; #if defined(EST_DEBUGGING) // It's not always very nice to resynthesise speech from // inserted zeros! These checks should alert the user (me ;) if( start<0 ) EST_warning( "padded start of pitch period with zeros (index %d)", i ); if( end>whole.num_samples() ) EST_warning( "padded end of pitch period with zeros (frame %d)", i ); #endif } // The window_signal function has been changed in several ways: // // *) The function now has an asymmetric window mode. // // In this mode, asymmetric windows are used from pitchmark at t-1 // to pitchmark at time t+1, with the maximum value of 1.0 at // pitchmark at time t. // // *) In the original symmetric mode: // // The first change is to ensure the window frames always have an // odd number of samples (a convention for how to handle rounding // problems when converting from times (float) to sample numbers // (int)). The centre sample corresponds to the pitch mark time. // // The second change is that the estimate of local pitch period is // always based in current and *previous* pitchmark. In the case // of the first pitch mark in track pm, the previous pitchmark is // assumed to be at zero time. Hopefully, this won't break much. // However, if this convention is not used everywhere else that // it's needed and some things break, then arguably those // things need to be fixed to adhere to this same convention... void window_signal(EST_Wave &sig, EST_Track &pm, EST_WaveVector &frames, int &i, float scale, float window_factor, EST_WindowFunc *window_function, bool window_symmetric, EST_IVector *pm_indices=0) { float first_pos, period=0.0; float prev_pm, current_pm; int first_sample, centre_sample, last_sample; int sample_rate = sig.sample_rate(); int pm_num_frames = pm.num_frames(); // estimate first period as pitchmark time itself (i.e. assume a previous // pitchmark at 0.0 time, waveform sample 0) prev_pm = 0.0; if( window_symmetric ) { if (pm_num_frames < 1 ) EST_error( "Attempted to Window around less than 1 pitchmark" ); for( int j=0; jn() << endl; #endif ++i; } } } void window_units( EST_Relation &unit_stream, EST_TVector &frames, float window_factor, EST_String window_name, bool window_symmetric, EST_IVector *pm_indices ) { int i; EST_Wave *sig; EST_Item *u; EST_Track *coefs; int num = 0; float scale; EST_WindowFunc *window_function; for (u = unit_stream.head(); u; u = u->next()) num += track(u->f("coefs"))->num_frames(); frames.resize(num); if( pm_indices != 0 ) pm_indices->resize(num); if (window_name == "") window_name = "hanning"; window_function = EST_Window::creator(window_name); for (i = 0, u = unit_stream.head(); u; u = u->next()) { sig = wave(u->f("sig")); coefs = track(u->f("coefs")); scale = (u->f_present("scale") ? u->F("scale") : 1.0); window_signal(*sig, *coefs, frames, i, scale, window_factor, window_function, window_symmetric, pm_indices); } } void us_unit_concat(EST_Utterance &utt, float window_factor, const EST_String &window_name, bool no_waveform=false, bool window_symmetric=true) { EST_Relation *unit_stream; EST_Track *source_coef = new EST_Track; EST_WaveVector *frames = new EST_WaveVector; EST_IVector *pm_indices = 0; unit_stream = utt.relation("Unit", 1); concatenate_unit_coefs(*unit_stream, *source_coef); utt.create_relation("SourceCoef"); EST_Item *item = utt.relation("SourceCoef")->append(); item->set("name", "coef"); item->set_val("coefs", est_val(source_coef)); if (!no_waveform){ if( !window_symmetric ) pm_indices = new EST_IVector; window_units(*unit_stream, *frames, window_factor, window_name, window_symmetric, pm_indices); item->set_val("frame", est_val(frames)); if( !window_symmetric ) item->set_val("pm_indices", est_val(pm_indices)); } } void us_get_copy_wave(EST_Utterance &utt, EST_Wave &source_sig, EST_Track &source_coefs, EST_Relation &source_seg) { EST_Item *s, *n; if (!utt.relation_present("Segment")) EST_error("utterance must have \"Segment\" relation\n"); utt.create_relation("TmpSegment"); for (s = source_seg.head(); s; s = s->next()) { n = utt.relation("TmpSegment")->append(); merge_features(n, s, 0); } utt.relation("Segment")->remove_item_feature("source_end"); dp_time_align(utt, "TmpSegment", "Segment", "source_", 0); utt.create_relation("Unit"); EST_Item *d = utt.relation("Unit")->append(); EST_Wave *ss = new EST_Wave; *ss = source_sig; EST_Track *c = new EST_Track; *c = source_coefs; d->set_val("sig", est_val(ss)); d->set_val("coefs", est_val(c)); utt.remove_relation("TmpSegment"); } void us_energy_normalise(EST_Relation &unit) { EST_Wave *sig; for (EST_Item *s = unit.head(); s; s = s->next()) { sig = wave(s->f("sig")); if (s->f_present("energy_factor")) sig->rescale(s->F("energy_factor")); } } void us_unit_raw_concat(EST_Utterance &utt) { EST_Wave *sig, *unit_sig; EST_Track *unit_coefs=0; float window_factor; int i, j, k; int first_pm, last_pm, last_length; float first_pos, last_pos; window_factor = get_c_float(siod_get_lval("window_factor", "UniSyn: no window_factor")); sig = new EST_Wave; sig->resize(1000000); sig->fill(0); j = 0; for (EST_Item *s = utt.relation("Unit", 1)->head(); s; s = s->next()) { unit_sig = wave(s->f("sig")); unit_coefs = track(s->f("coefs")); first_pos = unit_coefs->t(1); first_pm = (int)(first_pos * (float)unit_sig->sample_rate()); last_pos = unit_coefs->t(unit_coefs->num_frames()-2); last_pm = (int)(last_pos * (float)unit_sig->sample_rate()); last_length = unit_sig->num_samples() - last_pm; // cout << "first pm: " << first_pm << endl; // cout << "last pm: " << last_pm << endl; // cout << "last length: " << last_length << endl; j -= first_pm; for (i = 0; i < first_pm; ++i, ++j) sig->a_safe(j) += (short)((((float) i)/ (float)first_pm) *(float)unit_sig->a_safe(i)+0.5); for (; i < last_pm; ++i, ++j) sig->a(j) = unit_sig->a(i); for (k = 0; i < unit_sig->num_samples(); ++i, ++j, ++k) sig->a_safe(j) += (short)((1.0 - (((float) k) / (float) last_length)) * (float)unit_sig->a_safe(i) + 0.5); // j -= last_length; // j += 2000; } sig->resize(j); sig->set_sample_rate(16000); add_wave_to_utterance(utt, *sig, "Wave"); } void concatenate_unit_coefs(EST_Relation &unit_stream, EST_Track &source_lpc) { int num_source_frames = 0; int num_source_channels = 0;; float prev_time, abs_offset, rel_offset, period, offset; int i, j, k, l; EST_Track *coefs; EST_Item *u = unit_stream.head(); if( u == 0 ){ //sometimes we are just asked to synthesise empty utterances, and //code elsewhere wants us to continue... source_lpc.resize(0,0); } else{ EST_Track *t = 0; for ( ; u; u = u->next()) { t = track(u->f("coefs")); num_source_frames += t->num_frames(); } num_source_channels = t->num_channels(); source_lpc.resize(num_source_frames, num_source_channels); source_lpc.copy_setup(*t); prev_time = 0.0; // copy basic information for (i = 0, l = 0, u = unit_stream.head(); u; u = u->next()) { coefs = track(u->f("coefs")); for (j = 0; j < coefs->num_frames(); ++j, ++i) { for (k = 0; k < coefs->num_channels(); ++k) source_lpc.a_no_check(i, k) = coefs->a_no_check(j, k); source_lpc.t(i) = coefs->t(j) + prev_time; } prev_time = source_lpc.t(i - 1); u->set("end", prev_time); u->set("num_frames", coefs->num_frames()); } } // adjust pitchmarks abs_offset = 0.0; rel_offset = 0.0; // absolute offset in seconds abs_offset = get_c_float(siod_get_lval("us_abs_offset", "zz")); // relative offset as a function of local pitch period rel_offset = get_c_float(siod_get_lval("us_rel_offset", "zz")); if( abs_offset!=0.0 || rel_offset!=0.0 ){ cerr << "Adjusting pitchmarks" << endl; for (i = 0; i < source_lpc.num_frames(); ++i){ period = get_time_frame_size(source_lpc, (i)); offset = abs_offset + (rel_offset * period); source_lpc.t(i) = source_lpc.t(i) + offset; } } } // jointimes specifies centre of last pitch period in each // concatenated unit // void us_linear_smooth_amplitude( EST_Wave *w, // const EST_Track &pm, // const EST_FVector &jointimes) // { // int num_joins = jointimes.length(); // EST_Track *factor_contour = new EST_Track( num_joins ); // for( int i=0; iresize( pp_length, 1 ); for( int i=0; ia_no_check(i,0) = 0.0, j=0; ja_no_check( i, 0 ) += pow( float(frame.a_no_check( j )), float(2.0) ); contour->a_no_check(i,0) = sqrt( contour->a_no_check(i,0) / (float)j ); contour->t(i) = pm.t(i); } return contour; } EST_Val ffeature(EST_Item *item,const EST_String &fname); void us_linear_smooth_amplitude( EST_Utterance *utt ) { EST_WaveVector *pp = wavevector(utt->relation("SourceCoef")->first()->f("frame")); EST_Track *pm = track(utt->relation("SourceCoef")->first()->f("coefs")); EST_Track *energy = us_pitch_period_energy_contour( *pp, *pm ); energy->save( "./energy_track.est", "est" ); FILE *ofile = fopen( "./join_times.est", "w" ); EST_Relation *units = utt->relation("Unit"); for( EST_Item *u=units->head(); u; u=u->next() ){ EST_Item *diphone_left = u; // EST_Item *diphone_right = u->next(); fprintf( ofile, "%s\t%f\n", diphone_left->S("name").str(), diphone_left->F("end")); EST_Item *join_phone_left = item(diphone_left->f("ph1"))->next(); EST_String phone_name = join_phone_left->S("name"); if( ph_is_sonorant( phone_name ) && !ph_is_silence( phone_name )){ //if( (ffeature(join_phone_left, "ph_vc")).S() == "+"){ // ideally for sonorants cerr << "smoothing phone " << join_phone_left->S("name") << "\n"; // EST_Item *join_phone_right = item(diphone_right->f("ph1")); int left_end_index = energy->index(diphone_left->F("end")); int right_start_index = left_end_index + 1; float left_power = energy->a(left_end_index,0); float right_power = energy->a(right_start_index,0); float mean_power = (left_power+right_power)/2.0; float left_factor = left_power/mean_power; float right_factor = right_power/mean_power; int smooth_start_index = left_end_index-5; int smooth_end_index = right_start_index+5; // rescale left pitch periods float factor = 1.0; float factor_incr = (left_factor-1.0)/(float)(left_end_index - smooth_start_index); for( int i=smooth_start_index; i<=left_end_index; ++i, factor+=factor_incr ){ (*pp)[i].rescale( factor, 0 ); cerr << "rescaled frame " << i << "(factor " << factor << ")\n"; } // rescale right pitch periods factor = right_factor; factor_incr = (1.0-right_factor)/(float)(smooth_end_index-right_start_index); for( int i=right_start_index; i<=smooth_end_index; ++i, factor+=factor_incr){ (*pp)[i].rescale( factor, 0 ); cerr << "rescaled frame " << i << "(factor " << factor << ")\n"; } } else cerr << "no smoothing for " << join_phone_left->S("name") << "\n"; cerr < #include "EST_inline_utils.h" void us_generate_wave(EST_Utterance &utt, const EST_String &filter_method, const EST_String &ola_method) { EST_IVector *map, *frame_pm_indices; EST_WaveVector *frames; EST_Track *source_coef, *target_coef; EST_Wave *sig, *res; EST_FVector gain; frames = wavevector(utt.relation("SourceCoef", 1)->head()->f("frame")); source_coef = track(utt.relation("SourceCoef", 1)->head()->f("coefs")); target_coef = track(utt.relation("TargetCoef", 1)->head()->f("coefs")); map = ivector(utt.relation("US_map", 1)->head()->f("map")); sig = new EST_Wave; if ( ola_method == "asymmetric_window" ){ frame_pm_indices = ivector(utt.relation("SourceCoef", 1)->head()->f("pm_indices")); asymmetric_window_td_synthesis( *frames, *target_coef, *sig, *map, *frame_pm_indices ); } else if ( ola_method == "synth_period" ) td_synthesis2(*frames, *target_coef, *sig, *map); else td_synthesis(*frames, *target_coef, *sig, *map); if (filter_method == "lpc") { map_coefs(*source_coef, *target_coef, *map); // Save residual for external prcessing res = new EST_Wave; res->copy(*sig); utt.relation("TargetCoef", 1)->head()->set_val("residual",est_val(res)); // fast version lpc_filter_fast(*target_coef, *sig, *sig); // slower version (but cleaner) //lpc_filter_1(*target_coef, *sig, *sig); } add_wave_to_utterance(utt, *sig, "Wave"); } void map_coefs(EST_Track &source_coef, EST_Track &target_coef, EST_IVector &map) { int i, j; int m; if (source_coef.num_channels() != target_coef.num_channels()) EST_error("Different numbers of channels in LPC resynthesis: " "source %d, target %d\n", source_coef.num_channels(), target_coef.num_channels()); if (map.n() > target_coef.num_frames()) m = target_coef.num_frames(); else m = map.n(); for (i = 0; i < m; ++i) for (j = 0; j < target_coef.num_channels(); ++j) target_coef.a_no_check(i, j) = source_coef.a_no_check(map.a_no_check(i), j); // There can be one or two frames at the end of target_coed without // a map. Here we zero them, blindly assuming they are in silence for ( ; i < target_coef.num_frames(); i++) for (j = 0; j < target_coef.num_channels(); ++j) target_coef.a_no_check(i, j) = 0; } void td_synthesis(EST_WaveVector &frames, EST_Track &target_pm, EST_Wave &target_sig, EST_IVector &map) { int t_start; int i, j; float sr; int last_sample=0; int map_n = map.n(); if( (frames.length()>0) && (map_n>0) ){ sr = (float)frames(0).sample_rate(); last_sample = (int)(rint(target_pm.end()*sr)) + ((frames(frames.length()-1).num_samples() - 1)/2);//window_signal guarantees odd target_sig.resize(last_sample+1); target_sig.fill(0); target_sig.set_sample_rate((int)sr); for( i=0; i=0 ) target_sig.a_no_check(j + t_start) += frame.a_no_check(j); } } } void asymmetric_window_td_synthesis(EST_WaveVector &frames, EST_Track &target_pm, EST_Wave &target_sig, EST_IVector &map, EST_IVector &frame_pm_indices) { int t_start; int i, j; float sr; int last_sample=0; int map_n = map.n(); #if defined(EST_DEBUGGING) cerr << "(maplength framelength pm_indiceslength) " << map_n << " " << frames.n() << " " << frame_pm_indices.n() << endl; #endif if( (frames.length()>0) && (map_n>0) ){ sr = (float)frames(0).sample_rate(); last_sample = (int)(rint(target_pm.end()*sr)) + (frames(map(map_n-1)).num_samples() - frame_pm_indices(map(map_n-1)) - 1); target_sig.resize(last_sample+1, EST_ALL, 0); //0 -> don't set values target_sig.fill(0); target_sig.set_sample_rate((int)sr); for( i=0; i=0 // (it might be less than one where prosodic modification // has moved the first pitch mark relative to wave start) for( j=-min(0,t_start); j window; EST_FVector f; float s_window_factor= Param().F("unisyn.window_factor", 1.0); if (frames.length()> 0) sr = (float)frames(0).sample_rate(); else sr = 16000; // sort of, who cares, its going to be a 0 sample waveform if (map.n() > 0) last_sample = (int)(target_pm.end() * sr) + (frames(map(map.n()-1)).num_samples() / 2); target_sig.resize(last_sample); target_sig.fill(0); target_sig.set_sample_rate((int)sr); for (i = 0; i < map.n(); ++i) { const EST_Wave &frame = frames(map(i)); s_period = (int)(get_frame_size(target_pm, i, (int)sr) * s_window_factor); // cout << "period: " << s_period << endl; // start of window is mid point of analysis window // minus local synth period window_start = (frame.num_samples() / 2) - s_period; EST_Window::window_signal(frame, "hanning", window_start, s_period * 2, f, 1); t_start = ((int)(target_pm.t(i) * sr) - (f.n() / 2)); for (j = 0; j < f.n(); ++j) if (j + t_start>=0) target_sig.a_no_check(j + t_start) += (short)f.a_no_check(j); } } /*static void debug_options(EST_Relation &source_lab, EST_Relation &unit, EST_Track &source_coef, EST_Track &target_coef, EST_TVector &frames) { EST_IVector map; EST_Wave sig; EST_Relation pm_lab; if (siod_get_lval("us_debug_save_source", NULL) != NIL) { make_segment_double_mapping(source_lab, source_coef, source_lab, source_coef, map); td_synthesis(source_coef, frames, source_coef, sig, map); sig.save("source_sig.wav", "nist"); } if (siod_get_lval("us_debug_save_source_coefs", NULL) != NIL) source_coef.save("source_coefs.pm", "est"); if (siod_get_lval("us_debug_save_target_pm", NULL) != NIL) target_coef.save("target_coefs.pm"); if (siod_get_lval("us_debug_save_source_lab", NULL) != NIL) source_lab.save("source.lab"); if (siod_get_lval("us_debug_save_target_unit_lab", NULL) != NIL) unit.save("target_unit.lab"); } */ festival/src/modules/UniSyn/us_mapping.cc0000644000175000017500000006656312444573463017170 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Paul Taylor */ /* Date: 6 Jan 1998 */ /* --------------------------------------------------------------------- */ /* LPC residual synthesis alternative version */ /* */ /*************************************************************************/ #include "EST_error.h" #include "EST_inline_utils.h" #include "us_synthesis.h" #include "Phone.h" #include // void make_segment_single_mapping(EST_Relation &source_lab, // EST_Track &source_pm, // EST_Track &target_pm, EST_IVector &map) // { // int i = 0; // int s_i_start, s_i_end, t_i_start, t_i_end; // EST_Item *s; // float s_end, s_start, t_end, t_start, f, m; // map.resize(target_pm.num_frames()); // s_start = t_start = 0.0; // if (target_pm.t(target_pm.num_frames() - 1) < // source_lab.tail()->F("end",0)) // { // EST_warning("Target pitchmarks end before end of target segment " // "timings (%f vs %f). Expect a truncated utterance\n", // target_pm.t(target_pm.num_frames() - 1), // source_lab.tail()->F("end",0.0)); // } // //cout << "Source_pm" << source_pm.equal_space() << endl << endl; // //cout << "Target_pm" << target_pm.equal_space() << endl << endl; // for (s = source_lab.head(); s; s = s->next()) // { // s_end = s->F("source_end"); // t_end = s->F("end"); // s_i_start = source_pm.index_below(s_start); // s_i_end = source_pm.index_below(s_end); // t_i_start = target_pm.index_below(t_start); // t_i_end = target_pm.index_below(t_end); // // fudge to make sure that at least one frame is available // if (s_i_end <= s_i_start) // s_i_end += 1; // //printf("%d %d %d %d\n", s_i_start, s_i_end, t_i_start, t_i_end); // //printf("%f %f %f %f\n\n", s_start, s_end, t_start, t_end); // m = float (s_i_end - s_i_start)/ float(t_i_end - t_i_start); // for (i = t_i_start, f = 0.0; i < t_i_end; ++i, ++f) // map[i] = EST_NINT(f * m) + s_i_start; // s_start = s->F("source_end"); // t_start = s->F("end"); // } // if (i == 0) // map.resize(0); // nothing to synthesize // else // map.resize(i - 1); // } void make_segment_single_mapping(EST_Relation &source_lab, EST_Track &source_pm, EST_Track &target_pm, EST_IVector &map) { int i = 0; int s_i_start, s_i_end, t_i_start, t_i_end; EST_Item *s; float s_end, s_start, t_end, t_start, m; map.resize(target_pm.num_frames()); s_start = t_start = 0.0; s_i_start = t_i_start = 0; if (target_pm.t(target_pm.num_frames() - 1) < source_lab.tail()->F("end",0)) { EST_warning("Target pitchmarks end before end of target segment " "timings (%f vs %f). Expect a truncated utterance\n", target_pm.t(target_pm.num_frames() - 1), source_lab.tail()->F("end",0.0)); } for (s = source_lab.head(); s; s = s->next()) { // printf( "*********************************************\nphone %s\n", s->S("name").str()); s_end = s->F("source_end"); t_end = s->F("end"); s_i_end = source_pm.index_below(s_end); t_i_end = target_pm.index_below(t_end); // make sure that at least one frame is available if (s_i_end <= s_i_start) s_i_end += 1; // printf("(s_i_start s_i_end t_i_start t_i_end) %d %d %d %d\n", // s_i_start, s_i_end, t_i_start, t_i_end ); // printf("(s_i_start-s_i_end t_i_start-t_i_end) %d %d\n", // s_i_end - s_i_start, t_i_end - t_i_start); // OK for time alignment mapping function here to be single // linear across subcomponents?... // m = float(t_i_end-t_i_start+1)/float (s_i_end-s_i_start+1); m = (t_end-t_start)/(s_end-s_start); //m =1.0; // printf( "m=%f\n", m ); // time offsets for relative times float apm_t_off = (s_i_start==0) ? 0.0 : source_pm.t(s_i_start-1); float tpm_t_off = (t_i_start==0) ? 0.0 : target_pm.t(t_i_start-1); // printf("apm_t_off = %f\ntpm_t_off = %f\n", apm_t_off, tpm_t_off); int apm_i = s_i_start; // analysis pitch mark index float apm_t = source_pm.t(apm_i)-apm_t_off;// analysis pitch mark time float next_apm_t = source_pm.t(apm_i+1)-apm_t_off; for( i=t_i_start; i<=t_i_end; ++i ){ float tpm_t = target_pm.t(i)-tpm_t_off; // target pitch mark time // find closest pitchmark (assume only need forward search from current // point, since pitchmarks should always be increasing) while( (apm_i<=s_i_end) && (fabs((next_apm_t*m)-tpm_t) <= fabs((apm_t*m)-tpm_t)) ){ // printf("(next_apm_t apm_t) %f %f\n", // fabs((next_apm_t*m)-tpm_t), fabs((apm_t*m)-tpm_t) ); apm_t = next_apm_t; ++apm_i; next_apm_t = source_pm.t(apm_i+1)-apm_t_off; } // // printf( "tpm %d = apm %d\n", i, apm_i ); // int slow_index = source_pm.index( target_pm(i) ); // printf( "(my slow) %d %d\n", apm_i, slow_index ); map[i] = apm_i; } // for next loop if (s->next()) { s_i_start = s_i_end+1; t_i_start = t_i_end+1; s_start = source_pm.t(s_i_start); t_start = target_pm.t(t_i_start); } } if (i == 0) map.resize(0); // nothing to synthesize else map.resize(i); } void make_linear_mapping(EST_Track &pm, EST_IVector &map) { int pm_num_frames = pm.num_frames(); map.resize(pm_num_frames); for (int i = 0; i < pm_num_frames; ++i) map[i] = i; } static bool contiguous( const EST_Item*left, const EST_Item* right ) { if( (item(left->f("source_ph1")))->next() == item(right->f("source_ph1")) ) return true; return false; } // note this function overlooks the spacing of the 1st pitchmark (could // assume previous pitchmark at 0.0 time...) static void pitchmarksToSpaces( const EST_Track &pm, EST_IVector *spaces, int start_pm, int end_pm, int wav_srate ) { int left_pm, right_pm; int num_frames = end_pm-start_pm; spaces->resize( num_frames, 0 ); left_pm = (int) rint( pm.t(start_pm)*wav_srate ); //should always be > 0 for( int i=0; if("sig"))->sample_rate(); // currently, the pitchmarks are just moved, there's no deletion or // insertion int target_pm_length = source_pm.length(); target_pm.resize(target_pm_length, source_pm.num_channels()); // places to keep temporary debugging information EST_IVector source_spacing(target_pm_length); EST_IVector target_spacing(target_pm_length); EST_IVector voicing(target_pm_length); // handle special case of first half of first diphone unit EST_Item *diphone_left = units.head(); int left_start_index = diphone_left->I("middle_frame"); int left_end_index = source_pm.index(diphone_left->F("end")); for( int i=0; inext(); diphone_right; diphone_right=diphone_left->next() ){ printf( "%s\t%f\n", diphone_left->S("name").str(), diphone_left->F("end")); int right_start_index = left_end_index + 1; int right_end_index = right_start_index + diphone_right->I("middle_frame"); printf( "%d %d %d %d (l_start, l_end, r_start, r_end\n", left_start_index, left_end_index, right_start_index, right_end_index ); EST_String join_phone_name = item(diphone_left->f("ph1"))->next()->S("name"); cerr << "phone contigous " << contiguous(diphone_left,diphone_right) << endl; ///////////DEBUG////////////////////////////////////////////// int voicing_val; if( ph_is_sonorant( join_phone_name ) && ! ph_is_silence( join_phone_name ) ){ voicing_val = 1; } else voicing_val = 0; for( int i=left_start_index; i=0; --i ){ // int ns = (spaces[i]+spaces[i+1])/2; // spaces[i+1] += spaces[i] - ns; // spaces[i] = ns; // } // for( int i=right_start_index-left_start_index; iF("end") ); diphone_left = diphone_right; } // copy the remaining pitchmarks for( int i=left_start_index; if("sig"))->sample_rate(); // currently, the pitchmarks are just moved, there's no deletion or // insertion int target_pm_length = source_pm.length(); target_pm.resize(target_pm_length, source_pm.num_channels()); // places to keep temporary debugging information EST_IVector source_spacing(target_pm_length); EST_IVector target_spacing(target_pm_length); EST_IVector voicing(target_pm_length); // handle special case of first half of first diphone unit EST_Item *diphone_left = units.head(); int left_start_index = diphone_left->I("middle_frame"); int left_end_index = source_pm.index(diphone_left->F("end")); for( int i=0; inext(); diphone_right; diphone_right=diphone_left->next() ){ printf( "%s\t%f\n", diphone_left->S("name").str(), diphone_left->F("end")); int right_start_index = left_end_index + 1; int right_end_index = right_start_index + diphone_right->I("middle_frame"); printf( "%d %d %d %d (l_start, l_end, r_start, r_end\n", left_start_index, left_end_index, right_start_index, right_end_index ); EST_String join_phone_name = item(diphone_left->f("ph1"))->next()->S("name"); cerr << "phone contigous " << contiguous(diphone_left,diphone_right) << endl; ///////////DEBUG////////////////////////////////////////////// int voicing_val; if( ph_is_sonorant( join_phone_name ) && ! ph_is_silence( join_phone_name ) ){ voicing_val = 1; } else voicing_val = 0; for( int i=left_start_index; i 0 ){ // //////////////////////////////////////////////////////////////////////// // // filter them to modify // //////////////////////////////////////////////////////////////////////// // // copy modified pitchmark positions back into correct region of // // target_pm track // printf( "** using modified spaces ** \n" ); // for( int i=left_start_index; iF("end") ); diphone_left = diphone_right; } // copy the remaining pitchmarks for( int i=left_start_index; i(source_pm).save( "/home/korin/projects/smoothing_temp/f0/sourceCoef.est" ) != write_ok ) EST_warning( "couldn't write sourceCoef.est file" ); } void us_mapping(EST_Utterance &utt, const EST_String &method) { EST_Relation *source_lab, *target_lab; EST_IVector *map; EST_Track *source_coef=0, *target_coef=0; source_coef = track(utt.relation("SourceCoef")->head()->f("coefs")); target_coef = track(utt.relation("TargetCoef")->head()->f("coefs")); map = new EST_IVector; // cout << "mapping method: " << method << endl; if (method != "segment_single") source_lab = utt.relation("SourceSegments"); target_lab = utt.relation("Segment", 1); /* if (method == "segment") make_segment_double_mapping(*source_lab, *source_coef, *target_lab, *target_coef, *map); else if (method == "dp_segment") make_dp_mapping(*source_lab, *source_coef, *target_lab, *target_coef, "Match", *map); */ if (method == "linear") make_linear_mapping(*source_coef, *map); else if (method == "segment_single") make_segment_single_mapping(*target_lab, *source_coef, *target_coef, *map); else if (method == "interpolate_joins"){ cerr << "Doing interpolate_joins\n"; EST_Relation *units = utt.relation("Unit"); make_join_interpolate_mapping(*source_coef, *target_coef, *units,*map); } else if (method == "interpolate_joins2"){ cerr << "Doing interpolate_joins2\n"; EST_Relation *units = utt.relation("Unit"); make_join_interpolate_mapping2(*source_coef, *target_coef, *units,*map); } else EST_error("Mapping method \"%s\" not found\n", (const char *)method); utt.create_relation("US_map"); EST_Item *item = utt.relation("US_map")->append(); item->set_val("map", est_val(map)); } void add_wave_to_utterance(EST_Utterance &u, EST_Wave &sig, const EST_String &name) { u.create_relation(name); EST_Item *item = u.relation(name)->append(); item->set_val("wave", est_val(&sig)); } void map_to_relation(EST_IVector &map, EST_Relation &r, const EST_Track &source_pm, const EST_Track &target_pm) { EST_Item *s, *t, *a=NULL; EST_Utterance *u = r.utt(); int i; // cout << "source: " << source_pm; // cout << "target: " << target_pm; u->create_relation("smap"); u->create_relation("tmap"); for (i = 0; i < source_pm.num_frames(); ++i) { s = u->relation("smap")->append(); s->set("index", i); s->set("end", source_pm.t(i)); } for (i = 0; i < target_pm.num_frames(); ++i) { s = u->relation("tmap")->append(); s->set("index", i); s->set("end", target_pm.t(i)); } EST_Item *last_s = 0; for (s = u->relation("smap")->head(); s; s = s->next()) { int n = s->I("index"); for (t = u->relation("tmap")->head(); t; t = t->next()) { if (map(t->I("index")) == n) { if (last_s != s) a = u->relation("lmap")->append(s); last_s = s; a->append_daughter(t); t->set("map", n); } } } } /* void make_segment_double_mapping(EST_Relation &source_lab, EST_Track &source_pm, EST_Relation &target_lab, EST_Track &target_pm, EST_IVector &map) { int i = 0; int s_i_start, s_i_end, t_i_start, t_i_end; EST_Item *s, *t; float s_end, s_start, t_end, t_start, f, m; map.resize(target_pm.num_frames()); s_start = t_start = 0.0; if (target_pm.t(target_pm.num_frames() - 1) < target_lab.tail()->F("end")) EST_warning("Target pitchmarks end before end of target segment " "timings. Expect a truncated utterance.\n"); for (s = source_lab.head(), t = target_lab.head(); s && t; s = s->next(), t = t->next()) { if (s->S("name") != t->S("name")) cerr << "Warning: Source and Target segment names do not match: " << s->S("name") << " " << t->S("name") << endl; s_end = s->F("end"); t_end = t->F("end"); s_i_start = source_pm.index_below(s_start); s_i_end = source_pm.index_below(s->F("end")); t_i_start = target_pm.index_below(t_start); t_i_end = target_pm.index_below(t->F("end")); // fudge to make sure that at least one frame is available if (s_i_end <= s_i_start) s_i_end += 1; // printf("%d %d %d %d\n", s_i_start, s_i_end, t_i_start, t_i_end); // printf("%f %f %f %f\n", s_start, s_end, t_start, t_end); m = float (s_i_end - s_i_start)/ float(t_i_end - t_i_start); for (i = t_i_start, f = 0.0; i < t_i_end; ++i, ++f) map[i] = EST_NINT(f * m) + s_i_start; s_start = s->F("end"); t_start = t->F("end"); } if (i == 0) map.resize(0); // nothing to synthesize else map.resize(i - 1); } void make_dp_mapping(EST_Relation &source_lab, EST_Track &source_pm, EST_Relation &target_lab, EST_Track &target_pm, const EST_String &match_name, EST_IVector &map) { int i = 0, j; int s_i_start, s_i_end, t_i_start, t_i_end; EST_Item *s, *t; float s_end, s_start, t_end, t_start, f, m, prev_end; map.resize(target_pm.num_frames()); map.fill(-1); s_start = t_start = 0.0; // should really be replaced by feature functions. for (prev_end = 0.0, s = source_lab.head(); s; s = s->next()) { s->set("start", prev_end); prev_end = s->F("end"); } // should really be replaced by feature functions. for (prev_end = 0.0, s = target_lab.head(); s; s = s->next()) { s->set("start", prev_end); prev_end = s->F("end"); } if (target_pm.t(target_pm.num_frames() - 1) < target_lab.tail()->F("end", 1)) EST_warning("Target pitchmarks end before end of target segment " "timings. Expect a truncated utterance.\n"); for (s = source_lab.head(); s; s = s->next()) { s_start = s->F("start"); cout << "source: " << *s << endl; while (s && (!s->in_relation(match_name))) s = s->next(); cout << "active source: " << *s << endl; s_end = s->F("end"); cout << "daughter: " << daughter1(s->as_relation(match_name)) << endl; cout << "parent: " << parent(s->as_relation(match_name)) << endl; t = parent(s->as_relation(match_name)); cout << "active target: " << *t << endl; t_end = t->F("end"); t_start = t->F("start"); s_i_start = source_pm.index_below(s_start); s_i_end = source_pm.index_below(s->F("end")); t_i_start = target_pm.index_below(t_start); t_i_end = target_pm.index_below(t->F("end")); // fudge to make sure that at least one frame is available if (s_i_end <= s_i_start) s_i_end += 1; //printf("%d %d %d %d\n", s_i_start, s_i_end, t_i_start, t_i_end); //printf("%f %f %f %f\n", s_start, s_end, t_start, t_end); m = float (s_i_end - s_i_start)/ float(t_i_end - t_i_start); for (i = t_i_start, f = 0.0; i < t_i_end; ++i, ++f) map[i] = EST_NINT(f * m) + s_i_start; cout << endl; } for (i = 0, j = 0; i < target_pm.num_frames(); ++i) { cout << map(i) << " "; if (map(i) != -1) { map[j] = map(i); cout << map(j) << " "; target_pm.t(j++) = target_pm.t(i); } } if (j == 0) map.resize(0); // nothing to synthesize else map.resize(j); } */ festival/src/modules/UniSyn/us_features.cc0000644000175000017500000002177211465010356017331 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black and Paul Taylor */ /* Date : February 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* An implementation of Metrical Tree Phonology */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "EST_error.h" #include "us_features.h" void add_feature_function(EST_Relation &r, const EST_String &fname, const EST_String &funcname) { for (EST_Item *p = r.head(); p; p = p->next()) p->set_function(fname, funcname); } void add_non_terminal_features(EST_Item *s, EST_Features &f) { EST_Features::Entries a; for (EST_Item *p = s; p; p = p->next()) { if (daughter1(p) != 0) { add_non_terminal_features(daughter1(p), f); for (a.begin(f); a; ++a) p->set_val(a->k, a->v); } } } void add_non_terminal_features(EST_Relation &r, EST_Features &f) { add_non_terminal_features(r.head(), f); } EST_Item *named_daughter(EST_Item *syl, const EST_String &fname, const EST_String &fval) { if ((daughter1(syl) != 0) && (daughter1(syl)->f(fname) == fval)) return daughter1(syl); if ((daughter2(syl) != 0) && (daughter2(syl)->f(fname) == fval)) return daughter2(syl); return 0; } EST_Item *syl_nucleus(EST_Item *syl_struct) { EST_Item *t; if (syl_struct == 0) return 0; if ((t = named_daughter(syl_struct, "sylval", "Rhyme")) != 0) { // cout << "rhyme: " << *t << endl; t = named_daughter(t, "sylval", "Nucleus"); // cout << "nucleus: " << *t << endl; return daughter1(t); } return 0; } EST_Item *nth(EST_Relation &r, int n) { int i = 1; for (EST_Item *s = r.head(); s; s = s->next(), ++i) if (n == i) return s; cerr << "Couldn't find item " << n << " in relation " << r.name() << " of length " << r.length() << endl; festival_error(); return 0; } EST_Item *nth_leaf(EST_Item *r, int n) { int i = 1; EST_Item *p; for (p = first_leaf_in_tree(r); p != next_leaf(last_leaf_in_tree(r)); p = next_leaf(p), ++i) if (n == i) return p; cerr << "Couldn't find leaf " << n << " in relation " << r->relation()->name() <F("end") - s->F("start"); } EST_Val usf_start(EST_Item *s) { s = s->as_relation("Segment"); //cout << "in usf_start\n"; //cout << *s << endl; /* EST_Relation *r = s->relation(); if (r->f.S("timing_style") != "segment") EST_warning("Attempted to use start() feature function " "in non segment relation\n"); */ return (prev(s) == 0) ? 0.0 : prev(s)->F("end"); } EST_Val usf_tilt_phrase_position(EST_Item *s) { EST_String rel_name = s->S("time_path"); EST_Item *t, *a; if ((t = s->as_relation(rel_name)) == 0) { cerr << "item: " << *s << endl; EST_error("No relation %s for item\n", (const char *) rel_name); } a = parent(t); cout << "us features phrase pos\n"; //cout << "dereferencing syllable: " << *a << endl; cout << "start: " << a->F("start") << endl; cout << "end: " << a->F("end") << endl; if (s->S("name") == "phrase_start") return a->F("start"); else return a->F("end"); } EST_Val usf_tilt_event_position(EST_Item *s) { EST_String rel_name = s->S("time_path"); EST_Item *t, *a; if ((t = s->as_relation(rel_name)) == 0) EST_error("No relation %s for item\n", (const char *) rel_name); a = parent(t); cout << "us features tilt pos\n"; cout << "dereferencing syllable: " << *a << endl; cout << "vowel_start: " << a->F("vowel_start") << endl; cout << "start: " << a->F("start") << endl; cout << "end: " << a->F("end") << endl; return a->F("vowel_start") + s->F("rel_pos",0.0); } EST_Val usf_leaf_end(EST_Item *s) { if (!s->f_present("time_path")) EST_error("Attempted to use leaf end() feature function on " "item with no time_path feature set: %s\n", (const char *)s->relation()->name()); EST_String rel_name = s->S("time_path"); EST_Item *t, *a; if ((t = s->as_relation(rel_name)) == 0) EST_error("No relation %s for item\n", (const char *) rel_name); a = last_leaf_in_tree(t); return a->F("end"); } EST_Val usf_leaf_start(EST_Item *s) { if (!s->f_present("time_path")) EST_error("Attempted to use leaf start() feature function on " "item with no time_path feature set: %s\n", (const char *)s->relation()->name()); EST_String rel_name = s->S("time_path"); EST_Item *t, *a; if ((t = s->as_relation(rel_name)) == 0) EST_error("No relation %s for item\n", (const char *) rel_name); a = first_leaf_in_tree(t); // cout << "this is the first node of the tree\n"; //cout << *a << endl; return a->F("start"); } EST_Val usf_int_start(EST_Item *s) { EST_String rel_name = "IntonationPhrase"; EST_Item *t, *a; if ((t = s->as_relation(rel_name)) == 0) EST_error("No relation %s for item\n", (const char *) rel_name); a = first_leaf_in_tree(parent(t)->as_relation("MetricalTree")); return a->F("start"); } EST_Val usf_int_end(EST_Item *s) { EST_String rel_name = "IntonationPhrase"; EST_Item *t, *a; if ((t = s->as_relation(rel_name)) == 0) EST_error("No relation %s for item\n", (const char *) rel_name); a = last_leaf_in_tree(parent(t)->as_relation("MetricalTree")); return a->F("end"); } #endif EST_Val usf_vowel_start(EST_Item *s) { if (!s->f_present("time_path")) EST_error("Attempted to use vowel_time() feature function " "in relation with no time_relation feature defined\n"); EST_String rel_name = s->S("time_path"); EST_Item *n = syl_nucleus(s->as_relation(rel_name)); n = n->as_relation("Segment"); return n->F("start"); } void register_unisyn_features(void) { // register_featfunc("unisyn_duration", usf_duration); // register_featfunc("unisyn_start", usf_start); register_featfunc("unisyn_vowel_start", usf_vowel_start); // register_featfunc("unisyn_leaf_end", usf_leaf_end); // register_featfunc("unisyn_leaf_start", usf_leaf_start); // register_featfunc("unisyn_int_end", usf_int_end); // register_featfunc("unisyn_int_start", usf_int_start); // register_featfunc("unisyn_tilt_event_position", usf_tilt_event_position); // register_featfunc("unisyn_tilt_phrase_position", usf_tilt_phrase_position); } festival/src/modules/UniSyn/UniSyn.h0000644000175000017500000003503610067004272016066 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Paul Taylor */ /* Date: March 1998 */ /* --------------------------------------------------------------------- */ /* */ /*************************************************************************/ #ifndef __UNISYN_H__ #define __UNISYN_H__ #include "festival.h" typedef EST_TVector EST_WaveVector; #ifdef HAVE_US_TDPSOLA_TM void us_init_tdpsola(); void us_tdpsola_synthesis(EST_Utterance &utt, const EST_String &ola_method); #endif void us_linear_smooth_amplitude( EST_Utterance *utt ); /**@name Functions for Concatenating Units */ //@{ /** Iterate through the Unit relation and create the SourceCeof relation, which contains a series of windowed frames of speech and a track of pitch-synchronous coefficients. SourceCoef contains a single item with two features, coefs and frame coefs'value is a track with all the concatenated pitchmarks and coefficients from the units. us_unit_concat is where the pitch synchronous windowing of the frames in each Unit is performed and the result of this is stored as the value of frame Require:Unit Provide:SourceCoef @param utt: utterance @param window_factor: This specifies how large the analysis window is in relation to the local pitch period. A value of 1.0 is often used as this means each frame approximately extends from the previous pitch mark to the next. @param window_name: This specifies the type of window used. "hanning" is standard but any window type available from the signal processing library can be used. @param window_symmetric: if this is set to true, then symmetric analysis windows are used centred at each pitch mark, with size determined by the time difference between current and previous pitchmarks. @param no_waveform: if this is set to true, only the coefficients are copied into SourceCoef - no waveform analysis is performed. */ void us_unit_concat(EST_Utterance &utt, float window_factor, const EST_String &window_name, bool no_waveform=false, bool window_symmetric=true); /** This function provides the setup for copy resynthesis. In copy resynthesis, a natural waveform is used as the source speech for synthesis rather than diphones or other concatenated units. This is often useful for testing a prosody module or for altering the pitch or duration of a natural waveform for an experiment. (As such, this function should really be thought of as a very simple unit selection module) In addition to the speech waveform itself, the function requires a set of pitchmarks in the standard form, and a set of labels which mark segment boundaries. The Segment relation must already exist in the utterance prior to calling this function. First, the function creates aUnit relation with a single item containing the waveform and the pitchmarks. Next it adds a set of source_end features to each item in the Segment relation. It does this by calculating a mapping between the Segment relation and the input labels. This mapping is performed by dynamic programming, as often the two sets of labels don't match exactly. The final result, therefore is a Unit relation and Segment relation with source_end features. As this is exactly the same output of the standard concantenative synthesis modules, from here on the utterance can be processed as if the units were from a genuine synthesizer. Copy synthesis itself can be performed by .... Require:Segment Provide:Unit @param utt: utterance @param source_sig: waveform @param source_pm: pitchmarks belonging to waveform @param source_seg: set of items with end times referring to points in the waveform */ void us_get_copy_wave(EST_Utterance &utt, EST_Wave &source_sig, EST_Track &source_pm, EST_Relation &source_seg); /** This function produces a waveform from the Unit relation without prosodic modification. In effect, this function simply concatenates the waveform parts of the units in the unit relation. An overlap add operation is performed at unit boundaries so that waveform discontinuities don't occur. */ void us_unit_raw_concat(EST_Utterance &utt); /** Items in the Unit relation can take an optional flagenergy_factor, which scales the amplitude of the unit waveform. This is useful because units often have different energy levels due to different recording circumstances. An energy_factor of 1.0 leaves the waveform unchanged. */ void us_energy_normalise(EST_Relation &unit); //@} /**@name Functions for Producing Mappings */ //@{ /** This function produces the mapping between the SourceCoef track and TargetCoef track. The mapping is controlled by two types of modification, duration and pitch. Duration is specified by the Segment relation. Each item in this relation has two features source_end and target_end.source_end is marks the end point of that segment in the concatenated set of source coefficients, while target_end marks the desired end of that segment. Pitch modification is specified by the patterns of pitchmarks in the SourceCoef track and TargetCoef track. While these tracks actually represent periods, their reciprocal represents the source and target F0 contours. The mapping is an integer array with one element for every pitchmark in the TargetCoef track. Therefore, every target pitchmark has a mapping element, and the value of that element is the frame number in the SourceCoef track which should be used to generate the frame of speech for that target pitchmark. Depending on the mapping, source frames can be duplicated or skipped. If the duration is constant, a higher target pitch will mean source frames are duplicated. If the pitch is constant, a longer target duration will also mean source frames are duplicated. The duration and pitch modifications are calculated at the same time, leading to a single mapping. Require:SourceCoef, TargetCoef, Segment Provide:US_Map */ void us_mapping(EST_Utterance &utt, const EST_String &method); // for graphical display only: void map_to_relation(EST_IVector &map, EST_Relation &r, const EST_Track &source_pm, const EST_Track &target_pm); //@} /**@name Functions for Generating Waveforms */ //@{ /** Standard waveform generation function. This function genrates the actual synthetic speech waveform, using information in the SourceCoef, TargetCoef and US_map relations. The first stage involves time domain processing, whereby a speech waveform or residual waveform is generated. The second (optional) stage passes this waveform through the set of filter coefficients specified in the TargetCoef track. The output synthetic waveform is put in the Wave relation. LPC resynthesis is performed by the lpc_filter_1 function. Require:SourceCoef, TargetCoef, US_map Provide:Wave @param utt: utterance @param filter_method: type of filter used - normally "lpc" or none ("") @param td_method: type of time domain synthesis. */ void us_generate_wave(EST_Utterance &utt, const EST_String &filter_method, const EST_String &ola_method); /** This copies coefficients from source_coef into target_coef according to the frame mapping specified by map. target_coef should already have been allocated, and the pitchmarks in the time array set to appropriate values. (this can be done by the f0_to_pitchmarks function). */ void map_coefs(EST_Track &source_coef, EST_Track &target_coef, EST_IVector &map); /** Time domain resynthesis. Generate a speech waveform by copying frames into a set of time positions given by target_pm. The frame used for each time position is given by map, and the frames themselves are stored individually as waveforms in frames. @param target_sig: output waveform @param target_pm: new pitchmark positions @param frames: array containing waveforms, each representing a single analysis frame @param map: mapping between target_pm and frames. */ void td_synthesis(EST_WaveVector &frames, EST_Track &target_pm, EST_Wave &target_sig, EST_IVector &map); /** Variant of td_synthesis, where each frame is re-windowed according to the size of the local synthesis pitch period. @param target_sig: output waveform @param target_pm: new pitchmark positions @param frames: array containing waveforms, each representing a single analysis frame @param map: mapping between target_pm and frames. */ void td_synthesis2(EST_WaveVector &frames, EST_Track &target_pm, EST_Wave &target_sig, EST_IVector &map); //@} void asymmetric_window_td_synthesis(EST_WaveVector &frames, EST_Track &target_pm, EST_Wave &target_sig, EST_IVector &map, EST_IVector &frame_pm_indices); /**@name Pitchmark Functions */ //@{ /** This function generates the target pitchmarks from the target F0 contour. The pitchmarks are generated by reading a value, \(f_{0}\) off the f0 contour at time \(t\), calculating the local pitch period \(\tau = 1/f_{0}\), and placing a pitchmark at time \(T + t\). The process is then repeated by reading the F0 value at this new point and so on. The F0 contour must be continuous in all regions, that is unvoiced regions must have pseudo f0 values also. Although artificial contours are best generated in this way to begin with, the function \ref{**} can be used to interpolate through unvoiced regions for non-continuous contours. As the last F0 value in the contour may not be the end of the utterance (for example if the last phone is unvoiced), the pitchmarks may be extended past the end of the contour. After processing, the generated track only contains the target pitchmarks, but later functions may fill the amplitude array of the track with target coefficients, and hence the space for these can be allocated at this stage. @param fz: input F0 contour. @param pm: set of pitchmarks to be generated. These are set to the correct size in the function. @param num_channels: (optional) number of coefficients used in further processing. @param default_f0: (optional) f0 value for interpolated end values @param target_end: (optional) fill from the end of the contour to this point with default f0 values. */ void f0_to_pitchmarks(EST_Track &fz, EST_Track &pm, int num_channels=0, float default_f0=100.0, float target_end=-1); /** This is a utility function for converting a set of pitchmarks back to an F0 contour and is usually used in system development etc. The generated F0 is evenly spaced. @param pm: input set of pitchmarks to be generated @param fz: otuput F0 contour. @param shift: frame shift of generated contour in seconds. */ void pitchmarks_to_f0(EST_Track &pm, EST_Track &fz, float shift); //@} void register_unisyn_features(void); #endif // __UNISYN_H__ festival/src/modules/UniSyn/us_features.h0000644000175000017500000000724207262605650017176 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black and Paul Taylor */ /* Date : February 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* An implementation of Metrical Tree Phonology */ /* */ /*=======================================================================*/ #ifndef __US_FEATURES_H__ #define __US_FEATURES_H__ #include "festival.h" EST_Val usf_vowel_start(EST_Item *s); void add_feature_function(EST_Relation &r, const EST_String &fname, const EST_String &f); EST_Item *syl_nucleus(EST_Item *syl_struct); EST_Item *named_daughter(EST_Item *syl, const EST_String &fname, const EST_String &fval); EST_Val vowel_start_time(EST_Item *s); bool verify_utterance_relations(EST_Utterance &u, const EST_String &names, int err); bool verify_relation_features(EST_Relation &r, const EST_String &names, int err); bool verify_relation_features(EST_Utterance &u, const EST_String rname, const EST_String &names, int err); EST_Item *nth_leaf(EST_Item *r, int n); EST_Item *nth(EST_Relation &r, int n); #endif festival/src/modules/UniSyn/us_synthesis.h0000644000175000017500000001310610067004272017373 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Paul Taylor */ /* Date: March 1998 */ /* --------------------------------------------------------------------- */ /* */ /*************************************************************************/ #ifndef __US_SYNTHESIS_H__ #define __US_SYNTHESIS_H__ #include "EST.h" typedef EST_TVector EST_WaveVector; VAL_REGISTER_TYPE_DCLS(wavevector,EST_WaveVector) VAL_REGISTER_TYPE_DCLS(ivector,EST_IVector) SIOD_REGISTER_TYPE_DCLS( wavevector, EST_WaveVector) void add_wave_to_utterance(EST_Utterance &u, EST_Wave &sig, const EST_String &name); /*void us_lpc_synthesis(EST_Utterance &utt); void us_mapping(EST_Utterance &utt, const EST_String &method); void us_td_synthesis(EST_Utterance &utt, const EST_String &filter_method, const EST_String &ola_method); void lpc_synthesis(EST_Track &source_coef, EST_Track &target_coef, EST_TVector &frames, EST_Wave &sig, EST_IVector &map); void window_signal(EST_Wave &sig, EST_Track &pm, EST_TVector &frames, int &i, float scale, float window_factor, const EST_String &window_name); void window_signal(EST_Relation &unit_stream, EST_TVector &frames, float window_factor, EST_String window_name); void concatenate_coefs(EST_Relation &unit_stream, EST_Track &source_lpc); void us_unit_copy_wave(EST_Utterance &utt, EST_Wave &source_sig, EST_Track *source_pm); void us_unit_concat(EST_Utterance &utt); void us_full_cut(EST_Relation &unit); void us_energy_normalise(EST_Relation &unit); void us_unit_raw_concat(EST_Utterance &utt); void pitchmarks_to_f0(EST_Track &pm, EST_Track &fz, float shift); void f0_to_pitchmarks(EST_Track &fz, EST_Track &pm, float target_end = -1.0); void targets_to_pitchmarks(EST_Relation &targ, EST_Track &pitchmarks, int num_channels,float end); void targets_to_f0(EST_Relation &targ, EST_Track &f0, const float shift); void stretch_f0_time(EST_Track &f0, float stretch, float s_last_time, float t_last_time); void warp_f0(EST_Track &source_f0, EST_Relation &source_seg, EST_Track &target_f0, EST_Relation &target_seg); void warp_pitchmarks(EST_Utterance &utt, EST_Track *source_pm, EST_Relation &source_seg, EST_Relation &target_seg); void us_F0targets_to_pitchmarks(EST_Utterance &utt, const EST_String &seg_relation); void add_wave_to_utterance(EST_Utterance &u, EST_Wave &sig, const EST_String &name); void make_linear_mapping(EST_Track &pm, EST_IVector &map); void make_dp_mapping(EST_Relation &source_lab, EST_Track &source_pm, EST_Relation &target_lab, EST_Track &target_pm, const EST_String &match_name, EST_IVector &map); void make_segment_double_mapping(EST_Relation &source_lab, EST_Track &source_pm, EST_Relation &target_lab, EST_Track &target_pm, EST_IVector &map); void make_segment_single_mapping(EST_Relation &source_lab, EST_Track &source_pm, EST_Track &target_pm, EST_IVector &map); */ #endif // __US_SYNTHESIS_H__ festival/src/modules/donovan/Makefile0000644000175000017500000000574407262605650016364 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## The code in this directory was developed by ## ## Steve Isard and Alistair Conkie ## ########################################################################### TOP=../../.. DIRNAME=src/modules/donovan H = t2s.h donovan.h SRCSC = makewav.c load_diphs.c coeffs.c excitation.c pitch.c durations.c SRCSCXX = donovan.cc SRCS = $(SRCSC) $(SRCSCXX) OBJS = $(SRCSC:.c=.o) $(SRCSCXX:.cc=.o) FILES = Makefile $(SRCS) $(H) LOCAL_INCLUDES = -I../include INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/donovan/makewav.c0000644000175000017500000001247207262605650016517 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Steve Isard and Alistair Conkie */ /* 1984 and 1996 */ /* This version was gifted by Steve and Alistair for this new */ /* copyright, the original retains their original copyright */ /* */ /*************************************************************************/ #include #include #include "t2s.h" void makewave(CONFIG *config, ACOUSTIC *as) { short *ptptr; short ptbuf[OUT_BUF+NCOEFFS]; float rbuf[NCOEFFS],amp; int i,j,k,l; int voiced; int zz = 0; short *mcebuf; short u_cache, s_cache = 0; short wkspace[2] = {0,0}; for(zz=0;zzf_sz;i++) { mcebuf = &(as->mcebuf[i]->frame[0]); voiced = mcebuf[1]/2; if(voiced == 0) { amp = 2*(float)sqrt((double)mcebuf[0]); } else { amp = (float) sqrt((double) (mcebuf[0] * voiced) ); } for(k=FR_DATA-NCOEFFS;k < FR_DATA;k++) rbuf[k-(FR_DATA-NCOEFFS)] = (float)mcebuf[k]/32767; rfctolpc(rbuf); /* convert to lpc coeffs */ for(j=0;j < as->duration[i];j++,ptptr++) { register float v,*cptr; register short *backptr; short *endptr; float exc; exc = iexc(voiced,as,wkspace); v = (float)(exc == 0 ? 0 : amp*10*exc); /* 10 is NEW */ /* following loop depends on an initial NCOEFFS zeros preceding * ptbuf to make *(--backptr) zero the first time through. * We are depending on the ILS header before ptbuf to supply * these zeros. If waves are ever to be synthesized without * an ILS header on the front, special provision will have to be * made. */ endptr = ptptr - NCOEFFS; backptr = ptptr; cptr = rbuf + NCOEFFS; /* point at (empty) last cell of rbuf, * because we are going to use coeffs in * reverse order */ for(; backptr > endptr;) v += (float)( *(--backptr) * *(--cptr) ); /* this is where we need to concern ourselves with */ /* flushing the buffer from time to time */ ptbuf[zz++] = (short)(v); if(zz>=(OUT_BUF+NCOEFFS)) { for(l=zz-NCOEFFS;l #include #include "EST_cutils.h" #include "t2s.h" static int load_index(CONFIG *config); static int load_diphs(CONFIG *config); /* awb added */ int nindex = 0; static ENTRY *indx = 0; static FRAME *dico = 0; int load_speech(CONFIG *config) { if (load_index(config) != 0) /* in alphabetical order */ return -1; if (load_diphs(config) != 0) return -1; return 0; } static int load_index(CONFIG *config) { char s[100]; int i; if (indx == 0) indx = walloc(ENTRY,NDIPHS); if((config->xfd=fopen(config->index_file,"rb")) == NULL) { (void)fprintf(stderr,"Can't open file %s\n",config->index_file); return -1; } for(i=0;(fgets(s,100,config->xfd) != NULL) && (i < NDIPHS);i++) { sscanf(s,"%s %d %d %d",indx[i].diph,&indx[i].beg,&indx[i].mid,&indx[i].end); } nindex = i; fclose(config->xfd); return 0; } static int load_diphs(CONFIG *config) { int i,j; if (dico == 0) dico = walloc(FRAME,NFRAMES); if((config->dfd=fopen(config->diphone_file,"rb")) == NULL) { fprintf(stderr,"Can't open file %s\n",config->diphone_file); return -1; } /* zero the first one... */ for(i=0;idfd) != 0) && (i < NFRAMES);i++) { ; } /* check the first little bit is as we expect... */ if ((dico[1].frame[0] != 181) || (dico[1].frame[1] != 176)) { if ((SWAPSHORT(dico[1].frame[0]) == 181) && (SWAPSHORT(dico[1].frame[1]) == 176)) { /* Its bytes swapped */ for (j=1;jdiphone_file); fclose(config->dfd); return -1; } } fclose(config->dfd); return 0; } int lookup(char *diph) { int low, high, mid; low = 0; high = nindex-1; while(low <= high) { mid = (low+high) / 2; if(strcmp(diph,indx[mid].diph)<0) high = mid-1; else if(strcmp(diph,indx[mid].diph)>0) low = mid+1; else return(mid); } return(-1); } void phonstoframes(SPN *ps, ACOUSTIC *as) { int i,j; int ref; as->f_sz = 0; for(i=0;ip_sz-1;i++) sprintf(ps->diphs[i],"%s-%s",ps->phons[i],ps->phons[i+1]); ps->pb[0] = 0; /* Gets treated a little bit specially */ /* insert zero frame */ as->mcebuf[as->f_sz++] = &dico[0]; for(i=0;ip_sz-1;i++) { ref = lookup(ps->diphs[i]); /* gives back the reference no. */ if(ref == -1) { (void)fprintf(stderr,"Diphone not found - %s\n",ps->diphs[i]); ref = 0; } if(as->f_sz+50 > as->f_max) { as_realloc(as->f_max*2,as->p_max,as); } for(j=indx[ref].beg;j<=indx[ref].end;j++) { if(j==indx[ref].mid) ps->pb[i+1] = as->f_sz; as->mcebuf[as->f_sz++] = &dico[j]; } } as->mcebuf[as->f_sz++] = &dico[0]; as->mcebuf[as->f_sz++] = &dico[0]; as->mcebuf[as->f_sz++] = &dico[0]; ps->pb[ps->p_sz] = as->f_sz-1; } festival/src/modules/donovan/coeffs.c0000644000175000017500000000732107262605650016326 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Steve Isard */ /* Date : 1984 */ /* This version was gifted by Steve for this new */ /* copyright, the original retains their original copyright */ /* */ /*************************************************************************/ #include #include #include "t2s.h" /* transform a set of reflection coefficients to linear prediction * coefficients using algorithm from HCL-H */ void rfctolpc(float *buf) { float a,b; register float *cptr; register int n,k; /* HCL-H algorithm goes through coeffs in reverse of order in which they * appear here, working on buffer numbered from 1 upwards. To get same * effect, we make cptr point at space just after last coeff and let n and * k go down from -1 instead of up from 1. */ cptr = buf + NCOEFFS; /* There should be NCOEFFS coeffs. Point just * after last one. */ for(n = -1; n >= -NCOEFFS; n--) { *(cptr+n) = -(*(cptr+n)); for(k = -1; 2*k >= n; k--) { a = *(cptr+k); b = *(cptr+n-k); *(cptr+k) = a - b * *(cptr+n); *(cptr+n-k) = b - a * *(cptr+n); } } } festival/src/modules/donovan/excitation.c0000644000175000017500000000734507262605650017236 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Steve Isard and Alistair Conkie */ /* Date : 1984 and 1996 */ /* This version was gifted by Steve for this new */ /* copyright, the original retains their original copyright */ /* */ /*************************************************************************/ #include #include #include "t2s.h" #define SQRTSIX 0.408248 int don_random_seed = 1; static short ial(void); /* wkspace[0] pitch period index wkspace[1] where in pitch period */ float iexc(short voiced, ACOUSTIC *as, short *wkspace) { switch(wkspace[1]) { case 0: wkspace[1] = as->pitch[wkspace[0]++] - 1; if(voiced) return((float)SQRTSIX); break; case 1: wkspace[1]--; if(voiced) return((float)2*SQRTSIX); break; case 2: wkspace[1]--; if(voiced) return((float)SQRTSIX); break; default: wkspace[1]--; if(voiced) return((float)0); break; } return( (float)(ial()) - 0.5); } static short ial() /* random number generator */ { int seed = don_random_seed; short i1,i2,i3; i1 = 1&(seed); i2 = (((seed)&4) >> 2); i3 = i1^i2; seed = (((seed) >> 1) + (i3 << 10)); don_random_seed = seed; return(i3); } festival/src/modules/donovan/pitch.c0000644000175000017500000001032607262605650016167 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alistair Conkie */ /* Date : 1986 */ /* This version was gifted by Alistair for this new */ /* copyright, the original retains their original copyright */ /* */ /*************************************************************************/ #include #include "t2s.h" static int interpolated_freq(int k, SPN *ps); static int interpolate(int a,int b,int c,int d,int e); void calc_pitch(SPN *ps, ACOUSTIC *as) { int j,k; int y; int l = 0; int k_old = 0; int k_fine = 0; int x = 0; for(j=0;jt_sz;j++) ps->abs_targ[j] = ps->cum_dur[ps->targ_phon[j]] + ps->pc_targs[j]*ps->duration[ps->targ_phon[j]]/100.0; for(k=0;kcum_dur[ps->p_sz];k+=100) { y = interpolated_freq(k,ps); x += 100*y; while(x>SR) { k_fine = k + interpolate(x-100*y,0,x,100,10000); x -= SR; as->pitch[l++] = k_fine-k_old; if(l == as->p_max) { as_realloc(as->f_max,as->p_max*2,as); } k_old = k_fine; } } as->p_sz = l; as->pitch[0] += FR_SZ/2; /* to compensate for mismatch */ } static int interpolated_freq(int k, SPN *ps) { int i; int freq; if(!ps->t_sz) return(DEF_F0); else if(kabs_targ[0]) return(ps->targ_freq[0]); else if(k>=ps->abs_targ[ps->t_sz-1]) return(ps->targ_freq[ps->t_sz-1]); for(i=1;it_sz;i++) { if((kabs_targ[i]) && (k>=ps->abs_targ[i-1])) { freq = interpolate(ps->abs_targ[i-1], ps->targ_freq[i-1], ps->abs_targ[i], ps->targ_freq[i],k); return(freq); } } return(-1); /* should never arrive here */ } static int interpolate(int a,int b,int c,int d,int e) { int f; f = (c*b + d*e - e*b -a*d)/(c-a); return(f); } festival/src/modules/donovan/durations.c0000644000175000017500000000756112442333465017075 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alistair Conkie */ /* Date : 1996 */ /* This version was gifted by Alistair for this new */ /* copyright, the original retains their original copyright */ /* */ /*************************************************************************/ #include #include "t2s.h" static int min(int a, int b); static float local_fmax(float a, float b); void durations(SPN *ps, ACOUSTIC *as) { int durdist; int interdist; float multiplier_i; float proportion; int i; int j; for(i=0;ip_sz;i++) ps->scale[i] = (float)ps->duration[i] / (float)((ps->pb[i+1]-ps->pb[i])*FR_SZ); ps->cum_dur[0] = 0; /* do cumulative at same time */ for(i=0,j=0;if_sz;i++) { if(i == ps->pb[j]) { if(j != 0) { ps->cum_dur[j] = ps->duration[j-1] + ps->cum_dur[j-1]; } as->duration[i] = FR_SZ; ps->duration[j] = FR_SZ; /* saves adding later */ j++; } else { durdist = min(i-ps->pb[j-1],ps->pb[j]-i); interdist = ps->pb[j] - ps->pb[j-1]; proportion = (float)durdist/(float)interdist; multiplier_i = local_fmax(0.01,4.0*proportion*(ps->scale[j-1]-1.0)+1.0); as->duration[i] = FR_SZ*multiplier_i; ps->duration[j-1] += as->duration[i]; } } } static int min(int a, int b) { return((ab)?a:b); } festival/src/modules/donovan/donovan.cc0000644000175000017500000002556211465010357016671 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black (Steve Isard and Alistair Conkie) */ /* Date : July 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Interface to Donovan (Isard) LPC diphone code */ /* */ /* Uses the FreeSpeech version of the code. Those .c files are */ /* with both systems and hopefully will continue to be so. Only the */ /* necessary parts of FreeSpeech are included (the waveform synthesizer) */ /* The rest is plugged in from this file */ /* */ /* Note the FreeSpeech code is GNU Copyright not like the above */ /* */ /*=======================================================================*/ #include #include #include "festival.h" #include "donovan.h" static CONFIG *make_config(void); static void delete_config(CONFIG *config); static SPN *make_spn(EST_Utterance &u); static void delete_spn(SPN *ps); static ACOUSTIC *make_as(SPN *ps); static void delete_as(ACOUSTIC *ss); static CONFIG *don_config=0; static short *outbuff = 0; static int outpos = 0; static int outmax = 0; static void don_make_silence(int duration); LISP FT_Donovan_Load_Diphones(LISP params) { if (don_config != 0) delete_config(don_config); don_config = make_config(); don_config->index_file=wstrdup(get_param_str("index_file",params,"index")); don_config->diphone_file = wstrdup(get_param_str("diphone_file",params,"diphs")); if (load_speech(don_config) != 0) // load in the diphones festival_error(); return NIL; } static void delete_config(CONFIG *config) { (void) config; } static CONFIG *make_config(void) { // Allocate a new config structure CONFIG *config = walloc(CONFIG,1); config->input_file = 0; config->output_file = 0; config->index_file = 0; config->diphone_file = 0; config->hash_file = 0; config->format = 0; config->ifd = 0; config->ofd = 0; config->xfd = 0; config->dfd = 0; return config; } LISP FT_Donovan_Synthesize_Utt(LISP utt) { EST_Utterance *u = get_c_utt(utt); EST_Item *item = 0; SPN *ps; ACOUSTIC *as; if (nindex == 0) { cerr << "Festival: no donovan diphones loaded\n"; festival_error(); } don_random_seed = 1; // so resynthesizing an utterance is always the same /* Build structure */ ps = make_spn(*u); // get info from utterance structure /* This doesn't work if there are less than two phones so */ /* deal with those cases explicitly */ if (ps->p_sz < 1) outpos = 0; else if (ps->p_sz < 2) don_make_silence(ps->duration[0]); else { as = make_as(ps); /* do the actual synthesis */ phonstoframes(ps,as); durations(ps,as); calc_pitch(ps,as); makewave(don_config,as); delete_as(as); } delete_spn(ps); // Add wave as stream into utterance EST_Wave *w = new EST_Wave; w->resize(outpos,1); for (int i=0; ilength(); i++) w->a_no_check(i) = outbuff[i]; w->set_sample_rate(SR); item = u->create_relation("Wave")->append(); item->set_val("wave",est_val(w)); return utt; } static void don_make_silence(int duration) { // Make silence for size of the first (and only) phone // duation is in sample points int i; short *buff = walloc(short,duration); for (i=0; i < duration; i++) buff[i] = 0; audio_play(buff,sizeof(short),duration,NULL); wfree(buff); } static SPN *make_spn(EST_Utterance &u) { // Build initial structure for donovan code SPN *ps = walloc(SPN,1); EST_Relation *seg = u.relation("Segment"); EST_Relation *targ = u.relation("Target"); EST_Item *s; LISP cps; const char *ph_name; EST_Item *rt; int i,j; float pos,seg_start,seg_dur; ps->p_sz = seg->length(); ps->p_max = ps->p_sz+1; ps->t_sz = num_leaves(targ->head()); ps->t_max = ps->t_sz+1; ps->phons = walloc(char *,ps->p_max); ps->duration = walloc(int,ps->p_max); ps->cum_dur = walloc(int,ps->p_max); ps->pb = walloc(int,ps->p_max); ps->scale = walloc(float,ps->p_max); ps->diphs = walloc(char *,ps->p_max); for (i=0; ip_sz; i++) ps->diphs[i] = walloc(char,8); ps->pc_targs = walloc(int,ps->t_max); ps->targ_phon = walloc(int,ps->t_max); ps->targ_freq = walloc(int,ps->t_max); ps->abs_targ = walloc(int,ps->t_max); for (j=i=0,s=seg->first(); s != 0; s=s->next(),i++) { if (((cps=ft_get_param("PhoneSet")) == NIL) || ((streq(get_c_string(cps),"holmes")))) ph_name = s->name(); else ph_name = map_phone(s->name(),get_c_string(cps),"holmes"); ps->phons[i] = wstrdup(ph_name); seg_start = ffeature(s,"segment_start"); seg_dur = ffeature(s,"segment_duration"); ps->duration[i] = (int)(seg_dur * 10000); /* in frames */ if (i>0) ps->cum_dur[i] = ps->cum_dur[i-1]; else ps->cum_dur[i] = 0; ps->cum_dur[i] += ps->duration[i]; for (rt = daughter1(s,"Target"); rt != 0; rt = rt->next(),j++) { ps->targ_phon[j] = i; ps->targ_freq[j] = rt->I("f0"); pos = ((rt->F("pos") - seg_start) / seg_dur) * 99.9; ps->pc_targs[j] = (int)pos; } } // makewave thinks its writing to a file but its actually writting // to an incore buffer. We allocate it here and make it bigger if // necessary if (outbuff != NULL) wfree(outbuff); if (i == 0) outmax = 10; else outmax = (int)(ps->cum_dur[i-1]*1.1); outbuff = walloc(short,outmax); outpos = 0; return ps; } static void delete_spn(SPN *ps) { // claim back the space from ps int i; if (ps == NULL) return; for (i=0; ip_sz; i++) { wfree(ps->diphs[i]); wfree(ps->phons[i]); } wfree(ps->phons); wfree(ps->duration); wfree(ps->cum_dur); wfree(ps->pb); wfree(ps->scale); wfree(ps->diphs); wfree(ps->pc_targs); wfree(ps->targ_phon); wfree(ps->targ_freq); wfree(ps->abs_targ); wfree(ps); return; } static ACOUSTIC *make_as(SPN *ps) { ACOUSTIC *as = walloc(ACOUSTIC,1); int nframes = ps->cum_dur[ps->p_sz-1]; int npp = nframes * 2; as->p_sz = 0; as->f_sz = 0; as->p_max = npp; as->f_max = nframes; as->mcebuf = walloc(FRAME*,nframes); as->duration = walloc(short,nframes); as->pitch = walloc(short,npp); return as; } static void delete_as(ACOUSTIC *as) { if (as == NULL) return; wfree(as->mcebuf); wfree(as->duration); wfree(as->pitch); wfree(as); return; } void as_realloc(int nframes, int npp, ACOUSTIC *as) { // I don't think this will ever be called in Festival (void)nframes; (void)npp; (void)as; cerr << "Donovan diphones: as_realloc called unexpectedly\n"; festival_error(); } void audio_play(short *start,int sz,int number,CONFIG *config) { // The lower level system thinks its calling an original function // but here we intercept the output and put in a EST_Wave class // This function will be called a number of times, but we know // (roughly) what the maximum size will be, but just in case // we make it bigger if necessary (void)config; if (outpos+number > outmax) { int noutmax = (int)((float)(outpos+number)*1.1); short *noutbuf = walloc(short,noutmax); memmove(noutbuf,outbuff,sizeof(short)*outpos); wfree(outbuff); outbuff = noutbuf; outmax = noutmax; } memmove(&outbuff[outpos],start,number*sz); outpos += number; } void festival_donovan_init(void) { // Donovan (Isard) LPC diphone set proclaim_module("donovan"); init_subr_1("Donovan_Init",FT_Donovan_Load_Diphones, "(Donovan_Init PARAMS)\n\ Initialize the Donovan LPC diphone database. PARAMS are an assoc list\n\ of parameter name and value. The two parameters are index_file (value is\n\ a pathname for \"diphlocs.txt\") and diphone_file (value is a pathname\n\ for \"lpcdiphs.bin\"). [see LPC diphone synthesizer]"); festival_def_utt_module("Donovan_Synthesize",FT_Donovan_Synthesize_Utt, "(Donovan_Synthesize UTT)\n\ Synthesize a waveform using the Donovan LPC diphone synthesizer.\n\ This is called from Synthesize when the Synth_Method Parameter has the\n\ value Donovan. [see LPC diphone synthesizer]"); } festival/src/modules/donovan/t2s.h0000644000175000017500000001743712376701537015612 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alistair Conkie and Steve Isard */ /*-----------------------------------------------------------------------*/ #ifndef _T2S_H_ #define _T2S_H_ #define NDIPHS 3000 #define NFRAMES 23000 #define FR_DATA 16 /* shorts per frame, coeffs + assorted */ #define FW 1 #define CW 2 #define PUNCT 3 #define DEF_F0 125 #define SR 10000 /* sample rate */ #define FR_SZ 132 /* standard frame size */ /* malloc defaults */ #define DEF_BUFFER 1024 #define DEF_LING_LIST 100 #define DEF_SPL 100 #define DEF_PHONS 100 #define DEF_TARGS 100 #define DEF_FRAMES 100 #define DEF_PM 100 #define PHON_SZ 5 #define DIPH_SZ 10 #define OUT_BUF 2048 #define NCOEFFS 12 /* non-rhotic vowel classification for assim.c */ #define V_DEL_R 0 #define V_AIR 1 #define V_EER 2 #define V_OOR 3 #define V_R2SCHWA 4 /* various typedefs */ typedef struct { char *input_file; char *output_file; char *index_file; char *diphone_file; char *hash_file; char *format; int type; /* format by any other name */ FILE *ifd; FILE *ofd; FILE *xfd; FILE *dfd; void *db; int fw_num; int sonority_num; int dur0_num; } CONFIG; typedef struct { int max; int sz; char *ptr; } BUFFER; typedef struct { char *word; int type; char *transcription; } LING; typedef struct { int max; int sz; LING **text; } LING_LIST; typedef struct key { char *keyword; int keycount; } KEY; typedef struct { char phoneme[5]; int syll; int dur; char *sprosod1; char *sprosod2; float strength1; float strength2; /* for combined elements */ } SPROSOD; typedef struct { int max; int sz; SPROSOD **phoneme; } SPROSOD_LIST; typedef struct { char diph[10]; int beg; int mid; int end; } ENTRY; typedef struct { short frame[FR_DATA]; } FRAME; typedef struct { int p_sz; int p_max; int t_sz; int t_max; char **phons; int *duration; int *cum_dur; int *pc_targs; int *targ_phon; int *targ_freq; int *abs_targ; /* maybe in samples */ int *pb; float *scale; char **diphs; } SPN; typedef struct { int f_sz; int p_sz; int f_max; int p_max; FRAME **mcebuf; short *duration; /* since variants may be required */ short *pitch; } ACOUSTIC; extern KEY fw[]; extern KEY son[]; extern KEY dur0[]; /* now definitions of global data */ /* awb -- deleted */ /* extern ENTRY indx[NDIPHS]; */ /* extern FRAME dico[NFRAMES]; */ extern int nindex; extern char *dbName; /* program prototypes */ /* audio.c */ void audio_open(CONFIG *config); void audio_play(short *start,int sz,int number,CONFIG *config); void audio_close(CONFIG *config); void audio_flush(CONFIG *config); /* makewave.c */ void makewave(CONFIG *config, ACOUSTIC *as); /* coeffs.c */ void rfctolpc(float *buf); /* conv.c */ void conv(CONFIG *config, LING_LIST *ling_list, SPROSOD_LIST *spl); void spl_cpy(int index,int syll, char *phon, int dur, char *type, float strength, SPROSOD_LIST *spl); void spl_cat(int index,char *type, float strength, SPROSOD_LIST *spl); int vowel(char *ph) ; /* durations.c */ void durations(SPN *ps, ACOUSTIC *as); /* excitation.c */ float iexc(short voiced, ACOUSTIC *as, short *wkspace); /* go.c */ void go(CONFIG *config, BUFFER *buffer, LING_LIST *ling_list, SPROSOD_LIST *spl, SPN *ps, ACOUSTIC *as); /* grammar.c */ void grammar(LING_LIST *ling_list); /* interface.c */ char *nrl_rules(char *in); /* load_diphs.c */ int load_speech(CONFIG *config); int lookup(char *diph); void phonstoframes(SPN *ps, ACOUSTIC *as); /* nrl_edin.c */ void nrl_edin_conv(char *str, char *str2); /* pitch.c */ void calc_pitch(SPN *ps, ACOUSTIC *as); /* prosody.c */ void prosody(SPROSOD_LIST *spl, SPN *ps); /* space.c */ void init(CONFIG *config, BUFFER *buffer, LING_LIST *ling_list, SPROSOD_LIST *spl, SPN *ps, ACOUSTIC *as); void terminate(CONFIG *config, BUFFER *buffer, LING_LIST *ling_list, SPROSOD_LIST *spl, SPN *ps, ACOUSTIC *as); void buffer_malloc(int num,BUFFER *buffer); void buffer_realloc(int num, BUFFER *buffer); void buffer_free(BUFFER *buffer); void ling_list_malloc(int num, LING_LIST *ling_list); void ling_list_realloc(int num, LING_LIST *ling_list); void ling_list_free(LING_LIST *ling_list); void spl_malloc(int num, SPROSOD_LIST *spl); void spl_realloc(int num, SPROSOD_LIST *spl); void spl_free(SPROSOD_LIST *spl); void ps_malloc(int nphons, int ntargs, SPN *ps); void ps_realloc(int nphons, int ntargs, SPN *ps); void ps_free(SPN *ps); void as_malloc(int nframes, int npp, ACOUSTIC *as); void as_realloc(int nframes, int npp, ACOUSTIC *as); void as_free(ACOUSTIC *as); /* syllab.c */ char *syllabify(char *string, CONFIG *config); char *stress(char *input); /* t2s.c */ void process_sentence(CONFIG *config, BUFFER *buffer, LING_LIST *ling_list, SPROSOD_LIST *spl, SPN *ps, ACOUSTIC *as); /* tags.c */ void tags(CONFIG *config, BUFFER *buffer, LING_LIST *ling_list); /* transcribe.c */ void transcribe(CONFIG *config, LING_LIST *ling_list); /* ulaw.c */ unsigned char linear2ulaw(int sample); int ulaw2linear(unsigned char ulawbyte); /* utils.c */ char **split(char *in); void tidy_split(char **root); KEY *binary(char *word, KEY tab[], int n); /* library prototypes int fprintf(FILE *stream, char *format, ... ); int printf(const char *format, ... ); int getopt(int argc,char **argv, char *optstring); int sscanf(char *s,char * format, ... ); int fread (char *ptr, int size, int nitems, FILE *stream); int fwrite (char *ptr, int size, int nitems, FILE *stream); int fclose(FILE *stream); */ #endif /* _T2S_H_ */ festival/src/modules/donovan/donovan.h0000644000175000017500000000613407262605650016533 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : July 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Interface to Donovan (Isard) LPC diphone code */ /* */ /*=======================================================================*/ #ifndef __DONOVAN_H__ #define __DONOVAN_H__ #ifdef __cplusplus extern "C" { #endif /* Maybe cut this down a bit */ #include "t2s.h" extern int don_random_seed; #ifdef __cplusplus } #endif #endif festival/src/modules/parser/Makefile0000644000175000017500000000551507262605650016210 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## Probabilistic parser and Metrical Trees ## ########################################################################### TOP=../../.. DIRNAME=src/modules/parser H = parser.h CPPSRCS = pparser.cc #CPPSRCS = pparser.cc $(TSRCS) SRCS = $(CPPSRCS) OBJS = $(CPPSRCS:.cc=.o) FILES=Makefile $(SRCS) $(H) LOCAL_INCLUDES = -I../include INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/parser/pparser.cc0000644000175000017500000001707312131543131016517 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : June 1997 */ /*-----------------------------------------------------------------------*/ /* */ /* Probabilistic parser for (S)CFG */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "parser.h" #include "EST_SCFG_Chart.h" LISP FT_PParse_Generalized_Utt(LISP args, LISP env) { LISP utt; // Utterance to parse LISP gram; // Grammar Name LISP prel; // Relation Name to parse over LISP pfeat; // Feature Name to parse over LISP crel; // Relation to create parse in args = leval(car(args), env); utt = car(args); gram = car(cdr(args)); prel = car(cdr(cdr(args))); pfeat = car(cdr(cdr(cdr(args)))); crel = car(cdr(cdr(cdr(cdr(args))))); // Parse utt using items in prel, using features in pfeat, and store // the parser output into crel EST_Utterance *u; LISP rules; const char *gram_name, *prel_name, *pfeat_name, *crel_name; u = get_c_utt(utt); gram_name = get_c_string(gram); prel_name = get_c_string(prel); pfeat_name = get_c_string(pfeat); crel_name = get_c_string(crel); rules = siod_get_lval(gram_name, NULL); if (rules == NULL) return utt; EST_SCFG grammar(rules); scfg_parse(u->relation(prel_name), pfeat_name, u->create_relation(crel_name), grammar); return utt; } LISP FT_PParse_Utt(LISP utt) { // Parse Words (using part of speech tags) using given // probabilistic grammar EST_Utterance *u = get_c_utt(utt); LISP rules; rules = siod_get_lval("scfg_grammar", NULL); if (rules == NULL) return utt; EST_SCFG grammar(rules); scfg_parse(u->relation("Word"),"phr_pos", u->create_relation("Syntax"),grammar); return utt; } LISP FT_MultiParse_Utt(LISP utt) { // You give them a parser and they just want more ... // Because in some modes utterance may contain multiple sentences // and the grammars we have only have only deal in more // traditional sentences this tries to split the utterance into // sentences and parse them individualls and add them to // a single Syntax relation as a list of trees. EST_Utterance *u = get_c_utt(utt); LISP rules, eos_tree; EST_Item *s,*e,*st,*et; rules = siod_get_lval("scfg_grammar", NULL); if (rules == NULL) return utt; eos_tree = siod_get_lval("scfg_eos_tree",NULL); u->create_relation("Syntax"); EST_SCFG_Chart chart; chart.set_grammar_rules(rules); for (st=u->relation("Token")->head(); st; st = st->next()) { for (et=st->next(); et; et=et->next()) if (wagon_predict(et,eos_tree) != 0) break; // Now find related words s = first_leaf(st)->as_relation("Word"); e = first_leaf(et->next())->as_relation("Word"); chart.setup_wfst(s,e,"phr_pos"); chart.parse(); chart.extract_parse(u->relation("Syntax"),s,e,TRUE); st = et; } return utt; } void MultiParse(EST_Utterance &u) { // You give them a parser and they just want more ... // Because in some modes utterance may contain multiple sentences // and the grammars we have only have only deal in more // traditional sentences this tries to split the utterance into // sentences and parse them individualls and add them to // a single Syntax release as a list of trees. LISP rules, eos_tree; EST_Item *s, *w; rules = siod_get_lval("scfg_grammar", NULL); if (rules == NULL) EST_error("Couldn't find grammar rules\n"); eos_tree = siod_get_lval("scfg_eos_tree",NULL); u.create_relation("Syntax"); EST_SCFG_Chart chart; chart.set_grammar_rules(rules); // produce a parse wherever there is a sentence end marker or // the end of utterance. for (w = s = u.relation("Word")->head(); w; w = w->next()) if (w->f_present("sentence_end") || (w->next() == 0)) { chart.setup_wfst(s, w->next(), "phr_pos"); chart.parse(); chart.extract_parse(u.relation("Syntax"), s, w->next(), TRUE); s = w->next(); } } void festival_parser_init(void) { proclaim_module("parser"); festival_def_utt_module("ProbParse",FT_PParse_Utt, "(ProbParse UTT)\n\ Parse part of speech tags in Word relation. Loads the grammar \n\ from scfg_grammar_filename and saves the best parse\n\ in the Syntax Relation."); init_fsubr("ProbParseGeneralized", FT_PParse_Generalized_Utt, "(ProbParseGeneralized (list utt gram prel pfeat crel))\n" "Parse utt over the prel relation using its pfeat feature\n" "Load grammar from gram, and save parse in relation crel"); festival_def_utt_module("MultiProbParse",FT_MultiParse_Utt, "(MultiProbParse UTT)\n\ Parse part of speech tags in Word relation. Unlike ProbParse this \n\ allows multiple sentences to appear in the one utterance. The CART \n\ tree in eos_tree is used to define end of sentence. Loads the \n\ grammar from scfg_grammar_filename and saves the best parse\n\ in the Syntax Relation."); } festival/src/modules/parser/parser.h0000644000175000017500000000614012131543131016172 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : March 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* Various parsers/metrical tree modules */ /* */ /*=======================================================================*/ #ifndef __PARSER_H__ #define __PARSER_H__ LISP FT_MetricalTree_Utt(LISP utt); LISP FT_PParse_Utt(LISP utt); LISP FT_PParse_Generalized_Utt(LISP args, LISP env); LISP FT_MultiParse_Utt(LISP utt); #endif festival/src/modules/UniSyn_diphone/Makefile0000644000175000017500000000530407263321403017633 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../../.. DIRNAME=src/modules/UniSyn_diphone H = us_diphone.h TSRCS = CPPSRCS = UniSyn_diphone.cc us_diphone_unit.cc us_diphone_index.cc $(TSRCS) SRCS = $(CPPSRCS) OBJS = $(CPPSRCS:.cc=.o) FILES = Makefile $(SRCS) $(H) INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/UniSyn_diphone/UniSyn_diphone.cc0000644000175000017500000001632607763351505021456 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Paul Taylor */ /* Date: February 1998 */ /* --------------------------------------------------------------------- */ /* Waveform Generation Scheme Interface File */ /* */ /*************************************************************************/ #include "EST.h" #include "us_diphone.h" #include "festival.h" extern USDiphIndex *diph_index; LISP us_dbs = NIL; LISP us_make_group_file(LISP lname, LISP params); int read_diphone_index(const EST_String &filename, USDiphIndex &di); void us_check_db(); void us_add_diphonedb(USDiphIndex *db); LISP FT_us_get_diphones(LISP lutt) { EST_Utterance *utt = get_c_utt(lutt); us_get_diphones(*utt); return lutt; } LISP us_select_db(LISP name) { // Select diphone set LISP lpair; lpair = siod_assoc_str(get_c_string(name),us_dbs); if (lpair == NIL) EST_error("US DB: no diphone database named \"%s\" defined\n", (const char *) get_c_string(name)); else diph_index = us_db(car(cdr(lpair))); return name; } LISP us_list_dbs(void) { // List names of currently loaded dbs LISP names,n; for (names=NIL,n=us_dbs; n != NIL; n=cdr(n)) names = cons(car(car(n)),names); return reverse(names); } LISP us_db_params(void) { // Return parameters of current db us_check_db(); return diph_index->params; } LISP us_diphone_init(LISP args) { EST_String x; USDiphIndex *d_index = new USDiphIndex; d_index->grouped = false; d_index->params = args; d_index->name = get_param_str("name",args,"name"); d_index->index_file = get_param_str("index_file",args,""); read_diphone_index(d_index->index_file, *d_index); // This is needed because there is no get_param_EST_String function x = get_param_str("grouped",args,""); if (x == "true") { d_index->grouped = true; if (d_index->ts.open(d_index->index_file) != 0) { cerr << "US DB: can't open grouped diphone file " << d_index->index_file << endl; festival_error(); } // set up the character constant values for this stream d_index->ts.set_SingleCharSymbols(";"); } else { *cdebug << ":" << get_param_str("grouped",args,"") << ":" << endl; *cdebug << "index grouped:" << d_index->grouped << endl; *cdebug << "true:" << true << endl; *cdebug << "false:" << false << endl; d_index->coef_dir = get_param_str("coef_dir",args,""); d_index->sig_dir = get_param_str("sig_dir",args,""); d_index->coef_ext = get_param_str("coef_ext",args,""); d_index->sig_ext = get_param_str("sig_ext",args,""); } us_add_diphonedb(d_index); return rintern(d_index->name); } LISP FT_us_full_cut(LISP lutt, LISP lrname) { EST_Utterance *utt = get_c_utt(lutt); EST_String rname = get_c_string(lrname); us_full_cut(*utt->relation(rname)); // parse_diphone_times(*(utt->relation(rname)), // *(utt->relation("SourceSegments"))); return lutt; } void festival_UniSyn_diphone_init(void) { proclaim_module("UniSyn_diphone"); init_subr_0("us_list_dbs", us_list_dbs, "(us_list_dbs)\n\ List names of UniSyn databases currently loaded."); init_subr_0("us_db_params", us_db_params, "(us_db_params)\n\ Return parameters of current UniSyn database."); init_subr_1("us_db_select", us_select_db, "(us_db_select NAME)\n\ Select named UniSyn database."); init_subr_1("us_get_diphones", FT_us_get_diphones, "(us_get_synthesis UTT)\n\ Construct a unit stream in UTT comprising suitable diphones. The unit \n\ stream produced is suitable for immediate use in us_ps_synthesis."); init_subr_2("us_make_group_file", us_make_group_file, "(us_make_group_file FILENAME PARAMS)\n\ Make a group file from the currently specified diphone set. PARAMS \n\ is an optional assoc list and allows specification of the \n\ track_file_format (default est_binary), sig_file_format (default \n\ snd) and sig_sample_format (default mulaw). This is recommended \n\ for LPC databases. For PSOLA based databases the sig_sample_format \n\ should probably be set to short."); init_subr_2("us_full_cut", FT_us_full_cut, "(us_ps_synthesis UTT SIGPR)\n\ Synthesize utterance UTT using signal processing technique SIGPR \n\ for the UniSyn pitch-synchronous synthesizer."); init_subr_1("us_diphone_init", us_diphone_init, "(us_diphone_init DIPHONE_NAME)\n\ Initialise UniSyn diphone synthesizer with database DIPHONE_NAME."); init_subr_1("diphone_present", us_check_diphone_presence, "(diphone_present? STR)\n\ Checks whether the given STRing corresponds to any diphone in the\n\ current database."); } festival/src/modules/UniSyn_diphone/us_diphone_unit.cc0000644000175000017500000002272111465010356021701 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Paul Taylor */ /* Date: 1998, 1999 */ /* --------------------------------------------------------------------- */ /* LPC residual synthesis alternative version */ /* */ /*************************************************************************/ #include "siod.h" #include "EST.h" #include "us_diphone.h" #include "Phone.h" extern USDiphIndex *diph_index; void dur_to_end(EST_Relation &r) { float prev_end = 0; for (EST_Item *p = r.head(); p ; p = p->next()) { p->set("end", p->F("dur") + prev_end); prev_end = p->F("end"); } } void add_end_silences(EST_Relation &segment, EST_Relation &target) { EST_Item *t, *n; float shift = 0.0; const float pause_duration = 0.1; t = segment.head(); if (!ph_is_silence(t->f("name"))) { n = t->insert_before(); n->set("name", ph_silence()); n->set("dur", pause_duration); shift += pause_duration; } t = segment.tail(); if (!ph_is_silence(t->S("name"))) { n = t->insert_after(); n->set("name", ph_silence()); n->set("dur", pause_duration); shift += pause_duration; } dur_to_end(segment); target.tail()->set("pos", (target.tail()->F("pos") + shift)); } void add_end_silences(EST_Relation &segment) { EST_Item *t, *n; t = segment.head(); if (!ph_is_silence(t->S("name"))) { n = t->insert_before(); n->set("name", ph_silence()); } t = segment.tail(); if (!ph_is_silence(t->S("name"))) { n = t->insert_after(); n->set("name", ph_silence()); } } void parse_diphone_times(EST_Relation &diphone_stream, EST_Relation &source_lab) { EST_Item *s, *u; EST_Track *pm; int e_frame, m_frame = 0; float dur_1 = 0.0, dur_2 = 0.0, p_time; float t_time = 0.0, end; p_time = 0.0; for (s = source_lab.head(), u = diphone_stream.head(); u; u = u->next(), s = s->next()) { pm = track(u->f("coefs")); e_frame = pm->num_frames() - 1; m_frame = u->I("middle_frame"); if (m_frame < 0) m_frame=0; dur_1 = pm->t(m_frame); if (e_frame < m_frame) e_frame=m_frame; dur_2 = pm->t(e_frame) - dur_1; s->set("source_end", (dur_1 + p_time)); p_time = s->F("source_end") + dur_2; end = dur_1 + dur_2 + t_time; t_time = end; u->set("end", t_time); } if (s) s->set("source_end", (dur_2 + p_time)); } void load_separate_diphone(int unit, bool keep_full, const EST_String &cut_type) { // Load in the coefficients and signame for this diphone // It caches the results in the diphone index entry, though // someone else may clear them. Note the full file is loaded // each time which isn't optimal if there are multiple diphones // is the same file int samp_start, samp_end; int pm_start, pm_end, pm_middle; EST_Track full_coefs, dcoefs, *coefs; // float q_start, q_middle, q_end; if (full_coefs.load(diph_index->coef_dir + "/" + diph_index->diphone[unit].S("filename") + diph_index->coef_ext) != format_ok) { cerr << "US DB: failed to read coefs file from " << diph_index->coef_dir + "/" + diph_index->diphone[unit].S("filename") + diph_index->coef_ext << endl; EST_error(""); } pm_start = full_coefs.index(diph_index->diphone[unit].f("start")); pm_middle = full_coefs.index(diph_index->diphone[unit].f("middle")); pm_end = full_coefs.index(diph_index->diphone[unit].f("end")); // option for taking half a diphone only if (cut_type == "first_half") pm_end = pm_middle; else if (cut_type == "second_half") pm_start = pm_middle; // find time of mid-point, i.e. boundary between phones full_coefs.sub_track(dcoefs, pm_start, pm_end - pm_start + 1, 0, EST_ALL); // Copy coefficients so the full coeffs can be safely deleted coefs = new EST_Track(dcoefs); for (int j = 0; j < dcoefs.num_frames(); ++j) coefs->t(j) = dcoefs.t(j) - full_coefs.t(Gof((pm_start - 1), 0)); diph_index->diphone[unit].set("first_dur", full_coefs.t(pm_middle) - full_coefs.t(pm_start)); diph_index->diphone[unit].set("second_dur", full_coefs.t(pm_end) - full_coefs.t(pm_middle)); if (keep_full) { EST_Track *f = new EST_Track; *f = full_coefs; diph_index->diphone[unit].set_val("full_coefs",est_val(f)); } diph_index->diphone[unit].set_val("coefs", est_val(coefs)); diph_index->diphone[unit].set("middle_frame", pm_middle - pm_start -1); EST_Wave full_sig, sub_sig; if (diph_index->sig_dir == "none") return; if (full_sig.load(diph_index->sig_dir + "/" + diph_index->diphone[unit].f("filename") + diph_index->sig_ext) != format_ok) { cerr << "US DB: failed to read signal file from " << diph_index->sig_dir + "/" + diph_index->diphone[unit].f("filename") + diph_index->sig_ext << endl; EST_error(""); } // go to the periods before and after samp_start = (int)(full_coefs.t(Gof((pm_start - 1), 0)) * (float)full_sig.sample_rate()); if (pm_end+1 < full_coefs.num_frames()) pm_end++; samp_end = (int)(full_coefs.t(pm_end) * (float)full_sig.sample_rate()); full_sig.sub_wave(sub_sig, samp_start, samp_end - samp_start + 1); EST_Wave *sig = new EST_Wave(sub_sig); diph_index->diphone[unit].set_val("sig", est_val(sig)); if (keep_full) { EST_Wave *s = new EST_Wave; *s = full_sig; diph_index->diphone[unit].set_val("full_sig", est_val(s)); } } void load_full_diphone(int unit) { // Load in the coefficients and signame for this diphone // It caches the results in the diphone index entry, though // someone else may clear them. Note the full file is loaded // each time which isn't optimal if there are multiple diphones // is the same file int pm_start, pm_end, pm_middle; EST_Track *full_coefs; full_coefs = new EST_Track; if (full_coefs->load(diph_index->coef_dir + "/" + diph_index->diphone[unit].f("filename") + diph_index->coef_ext) != format_ok) { cerr << "US DB: failed to read coefs file from " << diph_index->coef_dir + "/" + diph_index->diphone[unit].f("filename") + diph_index->coef_ext << endl; EST_error(""); } pm_start = full_coefs->index(diph_index->diphone[unit].f("start")); pm_middle = full_coefs->index(diph_index->diphone[unit].f("middle")); pm_end = full_coefs->index(diph_index->diphone[unit].f("end")); diph_index->diphone[unit].set_val("full_coefs", est_val(full_coefs)); EST_Wave *full_sig = new EST_Wave; if (full_sig->load(diph_index->sig_dir + "/" + diph_index->diphone[unit].f("filename") + diph_index->sig_ext) != format_ok) { cerr << "US DB: failed to read signal file from " << diph_index->sig_dir + "/" + diph_index->diphone[unit].f("filename") + diph_index->sig_ext << endl; EST_error(""); } diph_index->diphone[unit].set_val("full_sig", est_val(full_sig)); } festival/src/modules/UniSyn_diphone/us_diphone_index.cc0000644000175000017500000004423111465010356022031 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Paul Taylor */ /* Date: 1998, 1999 */ /* --------------------------------------------------------------------- */ /* LPC residual synthesis alternative version */ /* */ /*************************************************************************/ #include "siod.h" #include "EST.h" #include "us_diphone.h" #include "Phone.h" static bool US_full_coefs = false; USDiphIndex *diph_index = 0; extern LISP us_dbs; static void us_get_all_diphones(EST_Relation &diphone); static void load_grouped_diphone(int unit); void get_diphone(EST_Item &d); static int find_diphone_index_simple(const EST_String &d,USDiphIndex &di); void us_check_db(); void us_add_diphonedb(USDiphIndex *db); void load_full_diphone(int unit); VAL_REGISTER_CLASS(us_db,USDiphIndex) SIOD_REGISTER_CLASS(us_db,USDiphIndex) USDiphIndex::USDiphIndex() : dihash(1500) { gc_protect(¶ms); } USDiphIndex::~USDiphIndex() { gc_unprotect(¶ms); } void us_check_db() { if (diph_index == 0) EST_error("US DB: no diphone database loaded\n"); diph_index->ts.restart(); } void awb_free_diph_index() { /* chaising memory leaks */ if (diph_index != 0) { delete diph_index; diph_index = 0; } } /*static EST_String get_diphone_name(EST_Item *item,const EST_String dir) { // Get diphone name which may differ from phone name // Looks for us_diphone_, us_diphone, or name in that order EST_String d1; static EST_String dname = "us_diphone"; if (!item) return ""; else if ((d1 = item->S(dname+"_"+dir)) != "0") return d1; else if ((d1 = item->S(dname)) != "0") return d1; else return item->S("name"); } */ static EST_String get_diphone_name(EST_Item *item,const EST_String dir) { // Get diphone name which may differ from phone name // Looks for us_diphone_, us_diphone, or name in that order EST_String d1; static EST_String dname = "us_diphone"; static EST_String def = "0"; if (!item) return ""; else if ((d1 = (EST_String)item->f(dname+"_"+dir,def)) != "0") return d1; else if ((d1 = (EST_String)item->f(dname,def)) != "0") return d1; else return item->f("name","0").string(); } void us_get_diphones(EST_Utterance &utt) { // Create unit stream with coefficients and signals for each // diphone EST_Item *p, *d; EST_String name1, name2, file; us_check_db(); if (!utt.relation_present("Unit")) utt.create_relation("Unit"); US_full_coefs = (siod_get_lval("us_full_coefs", NULL) == NIL) ? false : true; p = utt.relation("Segment")->head(); name1 = get_diphone_name(p,"left"); // left part of first diphone utt.relation("Unit")->f.set("grouped", ((diph_index->grouped) ? 1 : 0)); if (!diph_index->grouped) { utt.relation("Unit")->f.set("coef_dir", diph_index->coef_dir); utt.relation("Unit")->f.set("sig_dir", diph_index->sig_dir); utt.relation("Unit")->f.set("coef_ext", diph_index->coef_ext); utt.relation("Unit")->f.set("sig_ext", diph_index->sig_ext); } for (p = p->next(); p; p = p->next()) { d = utt.relation("Unit")->append(); name2 = get_diphone_name(p,"right"); d->set("name", (name1 + "-" + name2)); get_diphone(*d); name1 = get_diphone_name(p,"left"); } // utt.create_relation("SourceSegments"); // for (p = utt.relation("Segment", 1)->head(); p; p = p->next()) // { // d = utt.relation("SourceSegments")->append(); // d->set_name(p->name()); // } if (!US_full_coefs) parse_diphone_times(*(utt.relation("Unit", 1)), *(utt.relation("Segment", 1))); } LISP us_check_diphone_presence(LISP name) { /* requested by "Nicholas Volk" */ int x = find_diphone_index_simple(get_c_string(name),*diph_index); if ( x < 0 ) return NIL; else return name; } LISP us_make_group_file(LISP lname, LISP params) { EST_String group_file, index_file; EST_String track_file_format, sig_file_format, sig_sample_format; EST_Relation diphone; EST_TokenStream ts; EST_Item *d; EST_Wave *sig; EST_Track *tr; FILE *fp, *fp_group; const int block_size = 1024; int pos; us_check_db(); track_file_format = get_param_str("track_file_format",params,"est_binary"); sig_file_format = get_param_str("sig_file_format",params,"snd"); sig_sample_format = get_param_str("sig_sample_format",params,"mulaw"); group_file = make_tmp_filename(); group_file += ".group"; index_file = get_c_string(lname); us_get_all_diphones(diphone); if ((fp = fopen(group_file, "wb")) == NULL) EST_error("US DB: failed to open group file as temporary file\n"); for (d = diphone.head(); d; d = d->next()) { sig = wave(d->f("sig")); tr = track(d->f("coefs")); pos = ftell(fp); d->set("track_start", pos); tr->save(fp, track_file_format); pos = ftell(fp); d->set("wave_start", pos); sig->save_file(fp, sig_file_format, sig_sample_format, EST_NATIVE_BO); } fclose(fp); if ((fp = fopen(index_file, "wb")) == NULL) EST_error("US DB: failed to open group file \"%s\" for writing\n", (const char *) index_file); fprintf(fp, "EST_File index\n"); fprintf(fp, "DataType ascii\n"); fprintf(fp, "NumEntries %d\n", diphone.length()); fprintf(fp, "IndexName %s\n", (const char *)diph_index->name); fprintf(fp, "DataFormat grouped\n"); fprintf(fp, "Version 2\n"); fprintf(fp, "track_file_format %s\n",(const char *)track_file_format); fprintf(fp, "sig_file_format %s\n",(const char *)sig_file_format); fprintf(fp, "EST_Header_End\n"); for (d = diphone.head(); d; d = d->next()) fprintf(fp, "%s %d %d %d\n", (const char *)d->S("name"), d->I("track_start"), d->I("wave_start"), d->I("middle_frame")); // Copy binary data from temporary group file to end of // real group file char buf[block_size]; int r; if ((fp_group = fopen(group_file, "rb")) == NULL) { fprintf(stderr,"Unexpected lost temporary group file from \"%s\"\n", (const char *)group_file); return NIL; } while ((r = fread(buf, sizeof(char), block_size, fp_group)) != 0) fwrite(buf, sizeof(char), r, fp); fclose(fp); fclose(fp_group); unlink(group_file); return NIL; } static void us_get_all_diphones(EST_Relation &diphone) { EST_Item *d; EST_String name1; for (int i = 0; i < diph_index->diphone.n(); ++i) { d = diphone.append(); d->set("name", diph_index->diphone[i].S("name")); get_diphone(*d); } } int read_diphone_index(const EST_String &filename, USDiphIndex &di) { EST_TokenStream ts; int i, ref; int num_entries; EST_Option hinfo; EST_EstFileType t; EST_String v, pointer, n; bool ascii; EST_read_status r; EST_Item d; di.index_offset = 0; if (ts.open(filename) != 0) { cerr << "US DB: can't open index file " << filename << endl; return misc_read_error; } // set up the character constant values for this stream ts.set_SingleCharSymbols(";"); if ((r = read_est_header(ts, hinfo, ascii, t)) != format_ok) return r; if (t != est_file_index) return misc_read_error; num_entries = hinfo.ival("NumEntries"); di.grouped = (hinfo.val("DataFormat") == "grouped") ? true : false; di.diphone.resize(num_entries); if (di.grouped) { di.track_file_format = hinfo.val_def("track_file_format","est"); di.sig_file_format = hinfo.val_def("sig_file_format","est"); for (i = 0; i < num_entries; ++i) { di.diphone[i].set("name", ts.get().string()); di.diphone[i].set("count", 0); di.diphone[i].set("track_start", atoi(ts.get().string())); di.diphone[i].set("wave_start", atoi(ts.get().string())); di.diphone[i].set("middle_frame", atoi(ts.get().string())); di.dihash.add_item(di.diphone[i].f("name"),i); } di.index_offset = ts.tell(); // cout << "index offset = " << di.index_offset << endl; } else { int n_num_entries = num_entries; for (i = 0; i < n_num_entries; ++i) { if (ts.eof()) EST_error("US DB: unexpected EOF in \"%s\": %d entries " "instead of %s\n", (const char *)filename, i, num_entries); n = ts.get().string(); di.diphone[i].set("name", n); di.diphone[i].set("filename", ts.get().string()); di.diphone[i].set("count", 0); if (!di.diphone[i].S("filename").contains("&", 0)) { di.diphone[i].set("start", atof(ts.get().string())); di.diphone[i].set("middle", atof(ts.get().string())); di.diphone[i].set("end", ts.get().string()); if ((di.diphone[i].F("start")>=di.diphone[i].F("middle"))|| (di.diphone[i].F("middle") >= di.diphone[i].F("end"))) { cerr << "US DB: diphone index for " << n << " start middle end not in order, ignored " << endl; i--; n_num_entries--; } } di.dihash.add_item(n,i); } // now copy reference entries for (i = 0; i < n_num_entries; ++i) if (di.diphone[i].S("filename").contains("&", 0)) { pointer = di.diphone[i].S("filename").after("&", 0); // cout << "pointer: = " << pointer << endl; if ((ref = find_diphone_index_simple(pointer,di)) == -1) { cerr << "US DB: Illegal diphone pointer in index file: " << i << " " << di.diphone[i].S("name") << " -> " << di.diphone[i].S("filename") << endl; EST_error(""); } di.diphone[i].set("filename", di.diphone[ref].S("filename")); di.diphone[i].set("start", di.diphone[ref].S("start")); di.diphone[i].set("middle", di.diphone[ref].S("middle")); di.diphone[i].set("end", di.diphone[ref].S("end")); // accurate values of these depend on where the // pitchmarks are placed. // di.diphone[i].fset("first_dur",di.diphone[ref].fS("first_dur")); // di.diphone[i].fset("second_dur", //di.diphone[ref].fS("second_dur")); } } return format_ok; } void get_diphone(EST_Item &d) { int unit; unit = find_diphone_index(d); if (diph_index->diphone[unit].f("count") == 0) { if (diph_index->grouped) load_grouped_diphone(unit); else { if (US_full_coefs) load_full_diphone(unit); else load_separate_diphone(unit, false, "all"); } diph_index->diphone[unit].set("count", d.I("count", 0) + 1); } if (US_full_coefs) { d.set_val("full_sig", diph_index->diphone[unit].f("full_sig")); d.set_val("full_coefs", diph_index->diphone[unit].f("full_coefs")); } else { d.set_val("sig", diph_index->diphone[unit].f("sig")); d.set_val("coefs", diph_index->diphone[unit].f("coefs")); d.set_val("middle_frame", diph_index->diphone[unit].f("middle_frame")); } if (!diph_index->grouped) { d.set_val("filename", diph_index->diphone[unit].f("filename")); d.set_val("diphone_start", diph_index->diphone[unit].F("start")); d.set_val("diphone_middle", diph_index->diphone[unit].F("middle")); d.set_val("diphone_end", diph_index->diphone[unit].F("end")); } } static void load_grouped_diphone(int unit) { int middle_frame; EST_Track *coefs; EST_Wave *sig; int wave_start, track_start; coefs = new EST_Track; sig = new EST_Wave; track_start = diph_index->diphone[unit].I("track_start"); wave_start = diph_index->diphone[unit].I("wave_start"); middle_frame = diph_index->diphone[unit].I("middle_frame"); diph_index->ts.seek(track_start + diph_index->index_offset); coefs->load(diph_index->ts); // type is self determined at present diph_index->ts.seek(wave_start + diph_index->index_offset); sig->load(diph_index->ts,diph_index->sig_file_format); diph_index->diphone[unit].set_val("coefs", est_val(coefs)); diph_index->diphone[unit].set("middle_frame", middle_frame); diph_index->diphone[unit].set_val("sig", est_val(sig)); } static int find_diphone_index_simple(const EST_String &d,USDiphIndex &di) { int found,r; r = di.dihash.val(d,found); if (found) return r; else return -1; } int find_diphone_index(const EST_Item &d) { // Find the approrpiate entry in the diphone index table // mapping to alternates if required. int index; EST_String diname = d.f("name"); // If all goes well the diphone will be found directly in the index index=find_diphone_index_simple(diname,*diph_index); if ((index=find_diphone_index_simple(diname,*diph_index)) != -1) return index; // But for various reasons it might not be there so allow // a run time specification of alternates. This isn't optimal // but is better than falling immediately back on just a default // diphone LISP alt_left = get_param_lisp("alternates_left",diph_index->params,NIL); LISP alt_right = get_param_lisp("alternates_right",diph_index->params,NIL); EST_String di_left = diname.before("-"); EST_String di_right = diname.after("-"); EST_String di_left_alt = get_param_str(di_left,alt_left,di_left); EST_String di_right_alt = get_param_str(di_right,alt_right,di_right); EST_String di_alt = di_left_alt+"-"+di_right_alt; if ((index=find_diphone_index_simple(di_alt,*diph_index)) != -1) { // cout << "UniSyn: using alternate diphone " << di_alt << " for " << // diname << endl; return index; } // It really isn't there so return the default one and print and error // now EST_String default_diphone = get_param_str("default_diphone",diph_index->params,""); if (default_diphone != "") { index = find_diphone_index_simple(default_diphone,*diph_index); if (index == -1) { cerr << "US DB: can't find diphone " << d.f("name") << " and even default diphone (" << default_diphone << ") doesn't exist" << endl; EST_error(""); } else cerr << "UniSyn: using default diphone " << default_diphone << " for " << diname << endl; return index; } else { cerr << "US DB: can't find diphone " << d.f("name") << " nor alternatives" << endl; EST_error(""); } return -1; } void us_full_cut(EST_Relation &unit) { EST_Track *full_coefs, *sub_coefs; EST_Wave *full_sig, sub_sig; EST_Item *s; int pm_start, pm_end, pm_middle; int samp_start, samp_end; float start_time; for (s = unit.head(); s; s = s->next()) { sub_coefs = new EST_Track; full_coefs = track(s->f("full_coefs")); full_sig = wave(s->f("full_sig")); pm_start = full_coefs->index(s->F("diphone_start")); pm_middle = full_coefs->index(s->F("diphone_middle")); pm_end = full_coefs->index(s->F("diphone_end")); full_coefs->copy_sub_track(*sub_coefs, pm_start, pm_end - pm_start + 1); start_time = full_coefs->t(Gof((pm_start - 1), 0)); for (int j = 0; j < sub_coefs->num_frames(); ++j) sub_coefs->t(j) = sub_coefs->t(j) - start_time; s->set("middle_frame", pm_middle - pm_start -1); s->set_val("coefs", est_val(sub_coefs)); // go to the periods before and after samp_start = (int)(full_coefs->t(Gof((pm_start - 1), 0)) * (float)full_sig->sample_rate()); if (pm_end+1 < full_coefs->num_frames()) pm_end++; samp_end = (int)(full_coefs->t(pm_end) * (float)full_sig->sample_rate()); full_sig->sub_wave(sub_sig, samp_start, samp_end - samp_start + 1); EST_Wave *sig = new EST_Wave(sub_sig); s->set_val("sig", est_val(sig)); } } void us_add_diphonedb(USDiphIndex *db) { // Add this to list of loaded diphone dbs and select it LISP lpair; if (us_dbs == NIL) gc_protect(&us_dbs); lpair = siod_assoc_str(db->name,us_dbs); if (lpair == NIL) { // new diphone db of this name us_dbs = cons(cons(rintern(db->name), cons(siod(db),NIL)), us_dbs); } else { // already one of this name cerr << "US_db: warning redefining diphone database " << db->name << endl; setcar(cdr(lpair),siod(db)); } diph_index = db; } festival/src/modules/UniSyn_diphone/us_diphone.h0000644000175000017500000001015507763351505020514 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Paul Taylor */ /* Date: 6 Jan 1998 */ /* --------------------------------------------------------------------- */ /* LPC residual synthesis alternative version */ /* */ /*************************************************************************/ #ifndef __US_DIPHONE_H__ #define __US_DIPHONE_H__ #include "EST_FileType.h" #include "EST_TVector.h" #include "EST_THash.h" #include "EST_Token.h" #include "ling_class/EST_Relation.h" SIOD_REGISTER_CLASS_DCLS(us_db,USDiphIndex) VAL_REGISTER_CLASS_DCLS(us_db,USDiphIndex) class USDiphIndex { public: USDiphIndex(); ~USDiphIndex(); EST_String name; EST_String index_file; EST_String group_file; EST_String track_file_format; EST_String sig_file_format; bool grouped; int index_offset; EST_TokenStream ts; // for grouped diphones EST_String coef_dir; EST_String sig_dir; EST_String coef_ext; EST_String sig_ext; LISP params; EST_TVector diphone; EST_TStringHash dihash; }; int read_diphone_database(const EST_String &filename, USDiphIndex &index); void check_us_db(); void load_separate_diphone(int unit, bool keep_full, const EST_String &cut_type="all"); int find_diphone_index(const EST_Item &d); void parse_diphone_times(EST_Relation &diphone_stream, EST_Relation &source_lab); void us_get_diphones(EST_Utterance &utt); void us_full_cut(EST_Relation &unit); LISP us_check_diphone_presence(LISP name); #endif // __US_DIPHONE_H__ festival/src/modules/diphone/Makefile0000644000175000017500000000611407262605650016336 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## A new clean implementation of diphone selection and concatenation ## ## code supporting the CSTR standard diphone db format ## ## Mostly written by Alistair Conkie ## ########################################################################### TOP=../../.. DIRNAME=src/modules/diphone H = diphone.h #EXTRASRCS = di_psolaTM.cc SRCS = diphone.cc di_reslpc.cc di_psola.cc di_select.cc \ di_pitch.cc di_io.cc oc.cc OBJS = $(SRCS:.cc=.o) FILES=Makefile $(SRCS) $(H) $(EXTRASRCS) LOCAL_INCLUDES = -I../include LOCAL_DEFINES = $(MODULE_DIPHONE_DEFINES) INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/diphone/diphone.cc0000644000175000017500000006100711465010357016627 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alistair Conkie */ /* Date : August 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* A free version of diphone selection and concatenation supported the */ /* standard CSTR diphone dbs */ /* */ /*=======================================================================*/ #include #include #include "festival.h" #include "ModuleDescription.h" #include "diphone.h" static ModuleDescription di_description = { "diphone", 1.0, "CSTR", "Alistair Conkie, Alan Black ", { "A free version of diphone selection and concatenation supported the", "standard CSTR diphone dbs ", NULL }, { { "Segment", "Segments fo the utterance." }, { "Target", "Intonation targets" }, { NULL, NULL } }, { { NULL,NULL} }, { { "Wave", "Synthesized waveform" }, { NULL,NULL} }, { {NULL,NULL,NULL,NULL} } }; static int lookupd(DIPHONE_DATABASE *database,char *p1, char *p2); static void di_diphones(DIPHONE_SPN *ps, DIPHONE_DATABASE *database); static void make_silence(DIPHONE_DATABASE *db,DIPHONE_SPN *ps,DIPHONE_OUTPUT *o); static void merge_silences(EST_Utterance &u); static void delete_diphone_db(DIPHONE_DATABASE *config); static DIPHONE_SPN *make_spn(EST_Utterance &u); static void delete_spn(DIPHONE_SPN *ps); static DIPHONE_ACOUSTIC *make_as(DIPHONE_SPN *ps); static void delete_as(DIPHONE_ACOUSTIC *ss); static DIPHONE_OUTPUT *make_output(DIPHONE_DATABASE *database,int size); static void delete_output(DIPHONE_OUTPUT *o); static LISP diphone_dbs = NIL; static DIPHONE_DATABASE *di_db=0; static EST_Regex RXdi_ud("[$_]"); VAL_REGISTER_TYPE(diphone_db,DD_STRUCT) SIOD_REGISTER_CLASS(diphone_db,DD_STRUCT) LISP FT_Diphone_Load_Diphones(LISP params) { DIPHONE_DATABASE *db; EST_Pathname groupfile; db = make_diphone_db(); groupfile = get_param_str("group_file",params,""); if (!streq(groupfile,"")) { db->gtype = di_grouped; di_load_grouped_db(groupfile,db,params); } else { db->gtype = di_ungrouped; di_general_parameters(db,params); di_fixed_parameters(db,params); di_load_database(db); // load in the diphones/index } di_add_diphonedb(db); // add to db list, selects it too return NIL; } void di_general_parameters(DIPHONE_DATABASE *db,LISP params) { // These may be reset from those set in group file db->name = wstrdup(get_param_str("name",params,"none")); db->default_diphone = wstrdup(get_param_str("default_diphone",params,"")); db->alternates_before = get_param_lisp("alternates_before",params,NIL); gc_protect(&db->alternates_before); db->alternates_after = get_param_lisp("alternates_after",params,NIL); gc_protect(&db->alternates_after); db->sig_access_type_str = wstrdup(get_param_str("access_type",params,"direct")); if (streq(db->sig_access_type_str,"direct")) db->sig_access_type = di_direct; else if (streq(db->sig_access_type_str,"dynamic")) db->sig_access_type = di_dynamic; else if (streq(db->sig_access_type_str,"ondemand")) db->sig_access_type = di_ondemand; else { cerr << "Diphone: unknown access method " << db->sig_access_type_str << endl; festival_error(); } } void di_fixed_parameters(DIPHONE_DATABASE *db,LISP params) { // These cannot be changed in a compiled group file int nphones; db->index_file = wstrdup(get_param_str("index_file",params,"index")); db->type_str = wstrdup(get_param_str("type",params,"pcm")); if (streq(db->type_str,"pcm")) db->type = di_pcm; else if (streq(db->type_str,"lpc")) db->type = di_lpc; else { cerr << "Diphone: unknown database type \"" << db->type_str << "\"" << endl; festival_error(); } db->signal_dir = wstrdup(get_param_str("signal_dir",params,"signal/")); db->signal_ext = wstrdup(get_param_str("signal_ext",params,".vox")); db->signal_type = wstrdup(get_param_str("signal_type",params,"vox")); db->lpc_dir = wstrdup(get_param_str("lpc_dir",params,"lpc/")); db->lpc_ext = wstrdup(get_param_str("lpc_ext",params,".lpc")); db->lpc_res_ext = wstrdup(get_param_str("lpc_res_ext",params,".res")); db->lpc_res_type = wstrdup(get_param_str("lpc_res_type",params,"esps")); db->lpc_type = wstrdup(get_param_str("lpc_type",params,"htk")); db->lpc_order = get_param_int("lpc_order",params,18); db->lpc_frame_shift = get_param_float("lpc_frame_shift",params,0.010); db->lpc_res_offset = get_param_float("lpc_res_offset",params, db->lpc_frame_shift); db->lpc_frame_offset = get_param_int("lpc_frame_offset",params,1); if (streq("t",get_param_str("lpc_pitch_synch",params,"f"))) db->lpc_pitch_synch = TRUE; db->pitch_dir = wstrdup(get_param_str("pitch_dir",params,"pitch/")); db->pitch_ext = wstrdup(get_param_str("pitch_ext",params,".pm")); db->samp_freq = get_param_int("samp_freq",params,16000); // old name was "group_type", new name "group_encoding" db->group_encoding_str=wstrdup(get_param_str("group_type",params,"raw")); db->group_encoding_str = wstrdup(get_param_str("group_encoding",params, db->group_encoding_str)); db->def_f0 = get_param_int("def_f0",params,125); if (streq(db->group_encoding_str,"raw")) db->group_encoding = di_raw; else if (streq(db->group_encoding_str,"ulaw")) db->group_encoding = di_ulaw; else { cerr << "Diphone: unknown group encoding" << endl; festival_error(); } db->phoneset = wstrdup(get_param_str("phoneset",params,"none")); nphones = phoneset_name_to_set(db->phoneset)->num_phones(); db->ndiphs = nphones*nphones; // max limit of diphones db->ndiphs = get_param_int("num_diphones",params,db->ndiphs); db->nindex = db->ndiphs; db->sig_band = get_param_int("sig_band",params,400); } static void delete_diphone_db(DIPHONE_DATABASE *db) { int i,j; wfree(db->type_str); wfree(db->name); wfree(db->index_file); wfree(db->signal_dir); wfree(db->signal_ext); wfree(db->signal_type); wfree(db->lpc_dir); wfree(db->lpc_ext); wfree(db->lpc_res_ext); wfree(db->pitch_dir); wfree(db->pitch_ext); wfree(db->phoneset); wfree(db->group_encoding_str); wfree(db->sig_access_type_str); gc_unprotect(&db->alternates_before); gc_unprotect(&db->alternates_after); if (db->gtype == di_grouped) { wfree(db->indx[0]->diph); // ptr to the whole diphname table wfree(db->allsignal); wfree(db->allulawsignal); wfree(db->allframes); } for (i=0; i < db->nindex; i++) { if (db->gtype == di_ungrouped) { wfree(db->indx[i]->diph); wfree(db->indx[i]->file); } wfree(db->indx[i]); if ((db->gtype == di_ungrouped) || (db->group_encoding != di_raw)) wfree(db->vox[i]->signal); wfree(db->vox[i]); wfree(db->pm[i]->mark); wfree(db->pm[i]); if (db->lpc[i]->f != 0) { if ((db->gtype == di_ungrouped) || // when they are not ptrs (db->group_encoding != di_raw)) // into db->allframes for (j=0; j < db->lpc[i]->nframes; j++) wfree(db->lpc[i]->f[j]); wfree(db->lpc[i]->f); } wfree(db->lpc[i]); } wfree(db->indx); wfree(db->vox); wfree(db->pm); wfree(db->lpc); wfree(db->offsets); if (db->gfd != NULL) fclose(db->gfd); } DIPHONE_DATABASE *make_diphone_db(void) { // Allocate a new database structure DIPHONE_DATABASE *db = walloc(DIPHONE_DATABASE,1); memset(db,0,sizeof(DIPHONE_DATABASE)); // all zeroed db->type = di_pcm; db->gtype = di_ungrouped; db->swap = FALSE; db->index_file = 0; db->signal_dir = 0; db->pitch_dir = 0; db->xfd = 0; db->samp_freq = 0; db->sig_band = 0; db->phoneset = 0; db->alternates_before = NIL; db->alternates_after = NIL; db->allsignal = 0; db->allulawsignal = 0; db->offsets = 0; db->gfd = 0; db->default_diphone = 0; db->lpc_pitch_synch = FALSE; db->nindex = 0; return db; } LISP FT_Diphone_Synthesize_Utt(LISP utt) { EST_Utterance *u = get_c_utt(utt); DIPHONE_SPN *ps; DIPHONE_ACOUSTIC *as; DIPHONE_OUTPUT *output; EST_Item *item=0; *cdebug << "Diphone module" << endl; if (di_db == 0) { cerr << "Diphone: no diphone database loaded" << endl; festival_error(); } // Find diphone names (_ and $ etc) if specified for this db apply_hooks(siod_get_lval("diphone_module_hooks",NULL),utt); /* Build structure */ merge_silences(*u); ps = make_spn(*u); // get info from utterance structure output = make_output(di_db,(ps->p_sz != 0 ? ps->cum_dur[ps->p_sz-1] : 0)); if (ps->p_sz < 2) { // just build a silence of the specified time make_silence(di_db,ps,output); } else { as = make_as(ps); di_diphones(ps,di_db); di_calc_pitch(di_db,ps,as); di_frame_select(di_db,ps,as); if (di_db->type == di_lpc) di_reslpc(di_db,as,output); // residual excited LPC else if (di_db->type == di_pcm) di_psola_tm(di_db,as,output); // can't be distributed else { cerr << "Diphone: unsupported database form\n"; festival_error(); } delete_as(as); } delete_spn(ps); // Add wave into utterance EST_Wave *w = new EST_Wave; w->resize(output->o_sz,1,0); w->set_sample_rate(di_db->samp_freq); for (int i=0; ilength(); i++) w->a_no_check(i) = output->track[i]; item = u->create_relation("Wave")->append(); item->set_val("wave",est_val(w)); delete_output(output); return utt; } LISP FT_reslpc_resynth(LISP file) { EST_Utterance *u = new EST_Utterance; EST_Item *item = 0; DIPHONE_ACOUSTIC *as; DIPHONE_OUTPUT *output; if (di_db == 0) { cerr << "Diphone: no diphone database loaded" << endl; festival_error(); } as = walloc(DIPHONE_ACOUSTIC,1); memset(as,0,sizeof(DIPHONE_ACOUSTIC)); output = make_output(di_db,0); reslpc_resynth(get_c_string(file),di_db,as,output); wfree(as); // Add wave into utterance EST_Wave *w = new EST_Wave; w->resize(output->o_sz-(2*di_db->sig_band),1); for (int i=0; ilength(); i++) w->a_no_check(i) = output->track[di_db->sig_band+i]; w->set_sample_rate(di_db->samp_freq); delete_output(output); item = u->create_relation("Wave")->append(); item->set_val("wave",est_val(w)); return siod(u); } static void make_silence(DIPHONE_DATABASE *db,DIPHONE_SPN *ps,DIPHONE_OUTPUT *o) { // Make the output directly as silence int sil_size = (int)((float)db->samp_freq * 0.020); if ((ps->p_sz > 0) && (ps->cum_dur[ps->p_sz-1] > sil_size)) sil_size = ps->cum_dur[ps->p_sz-1]; if (o->o_max < (sil_size+2*db->sig_band)) { wfree(o->track); o->track = walloc(short,(sil_size+2*db->sig_band)); o->o_max = (sil_size+2*db->sig_band); } memset(o->track,0,o->o_max*sizeof(short)); o->o_sz = o->o_max; } static void merge_silences(EST_Utterance &u) { // Remove multiple silences EST_Item *s, *ns; for (s=u.relation("Segment")->first(); s != 0; s=ns) { ns = s->next(); if ((ns != 0) && (ph_is_silence(s->name())) && (s->name() == ns->name())) // same *type* of silence { // delete this segment remove_item(s,"Segment"); } } } static void di_diphones(DIPHONE_SPN *ps, DIPHONE_DATABASE *database) { int ph=0; ps->cum_dur[0] = 0; for(ph=0;php_sz;ph++) { ps->cum_dur[ph+1] = ps->duration[ph] + ps->cum_dur[ph]; } for(ph=0;php_sz-1;ph++) { sprintf(ps->diphs[ph],"%s-%s",ps->phons[ph],ps->phons[ph+1]); ps->ref[ph] = lookupd(database,ps->phons[ph],ps->phons[ph+1]); *cdebug << ps->diphs[ph] << " " << ((database->indx[ps->ref[ph]]->file == 0) ? "grouped" : database->indx[ps->ref[ph]]->file) << endl; load_pitch_file(database,ps->ref[ph],di_direct); // ensure its loaded } } static int lookupd(DIPHONE_DATABASE *database,char *p1, char *p2) { // This should be made much faster int i; char diphone[20]; LISP alt1, alt2; sprintf(diphone,"%s-%s",p1,p2); for(i=0;inindex;i++) { if(!strcmp(database->indx[i]->diph,diphone)) return i; } // Try alternates alt2 = siod_assoc_str(p2,database->alternates_after); if (alt2 != NIL) { sprintf(diphone,"%s-%s",p1,get_c_string(car(cdr(alt2)))); for(i=0;inindex;i++) if(!strcmp(database->indx[i]->diph,diphone)) { *cdebug << "Diphone alternate: " << diphone << " substituted for " << p1 << "-" << p2 << endl; return i; } } alt1 = siod_assoc_str(p1,database->alternates_before); if (alt1 != NIL) { sprintf(diphone,"%s-%s",get_c_string(car(cdr(alt1))),p2); for(i=0;inindex;i++) if(!strcmp(database->indx[i]->diph,diphone)) { *cdebug << "Diphone alternate: " << diphone << " substituted for " << p1 << "-" << p2 << endl; return i; } } if ((alt2 != NIL) && (alt1 != NIL)) { sprintf(diphone,"%s-%s", get_c_string(car(cdr(alt1))), get_c_string(car(cdr(alt2)))); for(i=0;inindex;i++) if(!strcmp(database->indx[i]->diph,diphone)) { *cdebug << "Diphone alternate: " << diphone << " substituted for " << p1 << "-" << p2 << endl; return i; } } // If underscores and dollars exist do it again without them if ((EST_String(p1).contains(RXdi_ud)) || (EST_String(p2).contains(RXdi_ud))) { EST_String dp1=p1, dp2=p2; // fprintf(stddebug,"Diphone: removing _ and $\n"); dp1.gsub(RXdi_ud,""); dp2.gsub(RXdi_ud,""); return lookupd(database,dp1,dp2); } if ((database->default_diphone != NULL) && (!streq(database->default_diphone,""))) { fprintf(stderr,"Diphone: using default diphone %s for %s-%s\n", database->default_diphone,p1,p2); for(i=0;inindex;i++) if(!strcmp(database->indx[i]->diph,database->default_diphone)) { *cdebug << "Diphone alternate: " << diphone << " substituted for " << p1 << "-" << p2 << endl; return i; } } fprintf(stderr,"Diphone: diphone (or alternates) not found: %s\n",diphone); festival_error(); return 0; } static DIPHONE_OUTPUT *make_output(DIPHONE_DATABASE *database,int samps) { // Alloc the output buffer int nos; DIPHONE_OUTPUT *o = walloc(DIPHONE_OUTPUT,1); // estimate the number of samples nos = (int)((float)samps*1.1)+(2*database->sig_band); o->o_sz = 0; o->o_max = nos; o->track = walloc(short,nos); return o; } static void delete_output(DIPHONE_OUTPUT *o) { wfree(o->track); wfree(o); } static DIPHONE_SPN *make_spn(EST_Utterance &u) { // Build initial structure for donovan code DIPHONE_SPN *ps = walloc(DIPHONE_SPN,1); EST_Relation *seg = u.relation("Segment"); EST_Relation *targ = u.relation("Target"); EST_Item *s; LISP cps; const char *ph_name; EST_Item *rt; int i; float pos,seg_start,seg_dur,seg_end; ps->p_sz = seg->length(); ps->p_max = ps->p_sz+1; ps->t_sz = 0; ps->phons = walloc(char *,ps->p_max); ps->duration = walloc(int,ps->p_max); ps->cum_dur = walloc(int,ps->p_max); ps->diphs = walloc(char *,ps->p_max); ps->ref = walloc(int,ps->p_max); ps->pm_req = walloc(int,ps->p_max); ps->pm_giv = walloc(int,ps->p_max); ps->pm_chg_diph = walloc(int,ps->p_max); for (i=0; ip_sz; i++) ps->diphs[i] = walloc(char,20); ps->t_max = num_leaves(targ->head())+4; ps->pc_targs = walloc(int,ps->t_max); ps->targ_phon = walloc(int,ps->t_max); ps->targ_freq = walloc(int,ps->t_max); ps->abs_targ = walloc(int,ps->t_max); // Ensure there is a target at the start if ((targ->length() == 0) || (targ->first_leaf()->F("pos") != 0)) { ps->targ_phon[0] = 0; if (targ->length() == 0) ps->targ_freq[0] = di_db->def_f0; else ps->targ_freq[0] = targ->first()->I("f0"); ps->pc_targs[0] = 0; ps->t_sz++; } seg_end = 0; for (i=0,s=seg->first(); s != 0; s=s->next(),i++) { seg_start = seg_end; seg_end = s->F("end"); seg_dur = seg_end - seg_start; EST_String pph_name; // possibly calculated phone name if ((pph_name=s->f("diphone_phone_name").string()) != "0") ph_name = pph_name; else if (((cps=ft_get_param("PhoneSet")) == NIL) || ((streq(get_c_string(cps),di_db->phoneset)))) ph_name = s->name(); else ph_name = map_phone(s->name(),get_c_string(cps),di_db->phoneset); ps->phons[i] = wstrdup(ph_name); ps->duration[i] = (int)(seg_dur * di_db->samp_freq); /* in frames */ if (i>0) ps->cum_dur[i] = ps->cum_dur[i-1]; else ps->cum_dur[i] = 0; ps->cum_dur[i] += ps->duration[i]; for (rt = daughter1(s,"Target"); rt != 0; rt = rt->next(),ps->t_sz++) { ps->targ_phon[ps->t_sz] = i; ps->targ_freq[ps->t_sz] = rt->I("f0"); pos = ((rt->F("pos") - seg_start) / seg_dur) * 99.9; ps->pc_targs[ps->t_sz] = (int)pos; } } // Ensure there is a target at the end if ((targ->length() == 0) || (targ->last_leaf()->F("pos") != seg->last_leaf()->F("end"))) { ps->targ_phon[ps->t_sz] = i-1; if (targ->length() == 0) ps->targ_freq[ps->t_sz] = di_db->def_f0; else ps->targ_freq[ps->t_sz] = targ->last_leaf()->I("f0"); ps->pc_targs[ps->t_sz] = 100; ps->t_sz++; } return ps; } static void delete_spn(DIPHONE_SPN *ps) { // claim back the space from ps int i; if (ps == NULL) return; for (i=0; ip_sz; i++) { wfree(ps->diphs[i]); wfree(ps->phons[i]); } wfree(ps->diphs); wfree(ps->phons); wfree(ps->duration); wfree(ps->cum_dur); wfree(ps->ref); wfree(ps->pm_req); wfree(ps->pm_giv); wfree(ps->pm_chg_diph); wfree(ps->pc_targs); wfree(ps->targ_phon); wfree(ps->targ_freq); wfree(ps->abs_targ); wfree(ps); return; } static DIPHONE_ACOUSTIC *make_as(DIPHONE_SPN *ps) { DIPHONE_ACOUSTIC *as = walloc(DIPHONE_ACOUSTIC,1); int npp = (int)((float)ps->cum_dur[ps->p_sz-1]/ ((float)di_db->samp_freq/1000.0)); as->p_sz = 0; as->p_max = npp; as->pitch = walloc(int,npp); as->cum_pitch = walloc(int,npp+1); as->inwidth = walloc(int,npp+1); as->outwidth = walloc(int,npp+1); as->phon_ref = walloc(int,npp); as->phon_mark = walloc(int,npp); as->diph_ref = walloc(int,npp); as->diph_mark = walloc(int,npp); return as; } static void delete_as(DIPHONE_ACOUSTIC *as) { if (as == NULL) return; wfree(as->pitch); wfree(as->cum_pitch); wfree(as->inwidth); wfree(as->outwidth); wfree(as->phon_ref); wfree(as->phon_mark); wfree(as->diph_ref); wfree(as->diph_mark); wfree(as); return; } void di_add_diphonedb(DIPHONE_DATABASE *db) { // Add this to list of loaded diphone dbs and select it LISP lpair; DIPHONE_DATABASE *ddb; if (diphone_dbs == NIL) gc_protect(&diphone_dbs); lpair = siod_assoc_str(db->name,diphone_dbs); if (lpair == NIL) { // new diphone db of this name diphone_dbs = cons(cons(rintern(db->name), cons(siod(db),NIL)), diphone_dbs); } else { // already one of this name, don't know howto free it cerr << "Diphone: warning redefining diphone database " << db->name << endl; ddb = diphone_db(car(cdr(lpair))); delete_diphone_db(ddb); setcar(cdr(lpair),siod(db)); } di_db = db; } LISP FT_Diphone_select(LISP name) { // Select diphone set LISP lpair; lpair = siod_assoc_str(get_c_string(name),diphone_dbs); if (lpair == NIL) { cerr << "Diphone: no diphone database named " << get_c_string(name) << " defined\n"; festival_error(); } else { DIPHONE_DATABASE *db = diphone_db(car(cdr(lpair))); di_db = db; } return name; } LISP FT_Diphone_group(LISP name,LISP lfilename) { FT_Diphone_select(name); di_save_grouped_db(get_c_string(lfilename),di_db); return NIL; } static LISP FT_Diphone_list_dbs(void) { // List names of currently loaded dbs LISP names,n; for (names=NIL,n=diphone_dbs; n != NIL; n=cdr(n)) names = cons(car(car(n)),names); return reverse(names); } void festival_diphone_init(void) { // New diphone synthesizer proclaim_module("diphone", &di_description); #ifdef SUPPORT_PSOLA_TM // Only available for research proclaim_module("di_psolaTM"); #endif init_subr_1("Diphone_Init",FT_Diphone_Load_Diphones, "(Diphone_Init PARAMS)\n\ Initialize a general diphone database. PARAMS are an assoc list\n\ of parameter name and value. [see Diphone_Init]"); festival_def_utt_module("Diphone_Synthesize",FT_Diphone_Synthesize_Utt, "(Diphone_Synthesize UTT)\n\ Synthesize a waveform using the currently selected diphone database.\n\ This is called from Synthesize when the Synth_Method Parameter has the\n\ value Diphone. [see Diphone synthesizer]"); // proclaim_module("Diphone_Synthesize",&di_description); // init_module_subr("Diphone_Synthesize",FT_Diphone_Synthesize_Utt, // &di_description); init_subr_1("Diphone.select",FT_Diphone_select, "(Diphone.select DB_NAME)\n\ Select a preloaded diphone set named DB_NAME, diphone sets are\n\ identified by their name parameter."); init_subr_2("Diphone.group",FT_Diphone_group, "(Diphone.group DB_NAME GROUPFILE)\n\ Create a group file for DB_NAME in GROUPFILE. A group file is a saved\n\ image of the database containing only the necessary information.\n\ It is an efficient way for loading and using diphone databases\n\ and is the recommended format for diphone databases for normal\n\ use. [see Group files]"); init_subr_1("reslpc_resynth",FT_reslpc_resynth, "(reslpc_resynth FILENAME)\n\ Resynthesize FILENAME using LPC plus residual. The current diphone\n\ database must have definitions for LPC for this to work."); init_subr_0("Diphone.list",FT_Diphone_list_dbs, "(Diphone.list)\n\ List the names of the currently loaded diphone databases."); init_subr_2("diphone.oc",find_optimal_coupling, "(diphone.oc TABLE WEIGHTS)\n\ For a table of units (diphones) find the optimal coupling points. WEIGHTS\n\ is a list of weights to apply to the vector coefficients used in measuring\n\ the goodness of the joins."); } festival/src/modules/diphone/di_reslpc.cc0000644000175000017500000003502010126474330017137 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black with all the magic stuff from */ /* Alistair Conkie, Steve Isard and Paul Taylor */ /* Date : October 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Residual excited LPC synthesis */ /* */ /* From "Issues in High Quality LPC Analysis and Synthesis" */ /* Melvyn Hunt, Dariusz Zweirzynski and Raymond Carr, Eurospeech 89 */ /* vol 2 pp 348-351, Paris 1989 */ /* */ /*=======================================================================*/ #include #include #include #include #include "festival.h" #include "diphone.h" static void add_lpc_coeff(int nth,float *p_coeffs,EST_Track &lpc_coeff); static void add_residual(int insize, int outsize, int position, short *p_residual, short *residual); static void lpc_resynth(EST_Track &lpc_coeffs, short *residual, DIPHONE_ACOUSTIC *as, DIPHONE_OUTPUT *output); static void lpc_resynth_fixp(EST_Track &lpc_coeffs, short *residual, DIPHONE_ACOUSTIC *as, DIPHONE_OUTPUT *output); static int min(int a, int b); static void ref2lpc_fixp(const int *rfc, int *lpc, int order); void reslpc_resynth(const EST_String &file, DIPHONE_DATABASE *db, DIPHONE_ACOUSTIC *as, DIPHONE_OUTPUT *output) { // Resynthesize a named file from the current // This is not fully general, just for my testing EST_Track lpc_coeffs; EST_Wave residual; int i,frame_len; if (lpc_coeffs.load(EST_String(db->lpc_dir)+file+EST_String(db->lpc_ext)) != format_ok) { cerr << "Diphone: lpc resynthesis, failed to read lpc file" << endl; festival_error(); } if (residual.load(EST_String(db->lpc_dir)+file+EST_String(db->lpc_res_ext)) != format_ok) { cerr << "Diphone: lpc resynthesis, failed to read residual file" << endl; festival_error(); } if (db->lpc_pitch_synch) { as->outwidth = walloc(int,lpc_coeffs.num_frames()); output->o_max = 0; for (i=0; ioutwidth[i] = frame_len; output->o_max += frame_len; // Get rid of extra first coeff (time position) // for (int j=0; jo_max *= 2; output->track = walloc(short,output->o_max); // lpc_coeffs.set_num_channels(lpc_coeffs.num_channels()-1); } else { as->outwidth = walloc(int,lpc_coeffs.num_frames()); frame_len = (int)(db->lpc_frame_shift*db->samp_freq); for (i=0; ioutwidth[i] = frame_len; output->o_max = i*frame_len*2; output->track = walloc(short,output->o_max); } short *shresidual = walloc(short,residual.num_samples()); for (i=0; i < residual.num_samples(); i++) shresidual[i] = residual.a_no_check(i); // Reconstruct the waveform if (siod_get_lval("debug_reslpc",NULL) != NIL) { residual.save("res.wav","nist"); lpc_coeffs.save("lpc.ascii","ascii"); } if (siod_get_lval("lpc_fixedpoint",NULL) != NIL) lpc_resynth_fixp(lpc_coeffs,shresidual,as,output); else lpc_resynth(lpc_coeffs,shresidual,as,output); wfree(shresidual); } void di_reslpc(DIPHONE_DATABASE *db, DIPHONE_ACOUSTIC *as, DIPHONE_OUTPUT *output) { int i; int curr_diph; int curr_mark; int max_marks; EST_Track lpc_coeffs(as->p_sz-1,db->lpc_order); short *residual; residual = walloc(short,as->cum_pitch[as->p_sz-1]+db->sig_band); memset(residual,0,sizeof(short)*as->cum_pitch[as->p_sz-1]+db->sig_band); /* original widths */ for(i=0;ip_sz-1;i++) { curr_diph = as->diph_ref[i]; curr_mark = as->diph_mark[i]; max_marks = db->pm[curr_diph]->nmark; if (curr_mark == 0) { as->inwidth[i] = db->pm[curr_diph]->mark[1]-db->pm[curr_diph]->mark[0]; } else if (curr_mark == max_marks-1) { as->inwidth[i] = db->pm[curr_diph]->mark[curr_mark] -db->pm[curr_diph]->mark[curr_mark-1]; } else { as->inwidth[i] = min(db->pm[curr_diph]->mark[curr_mark+1] -db->pm[curr_diph]->mark[curr_mark], db->pm[curr_diph]->mark[curr_mark] -db->pm[curr_diph]->mark[curr_mark-1]); } } /* new widths */ as->outwidth[0] = as->pitch[0]; for(i=1;ip_sz;i++) as->outwidth[i] = min(as->pitch[i],as->pitch[i-1]); int res_pos=0; /* For each pitch period, copy the lpc coeffs and construct a residual */ for(i=0;ip_sz-1;i++) { curr_diph = as->diph_ref[i]; curr_mark = as->diph_mark[i]; if (as->outwidth[i] <= 5) continue; // too wee to worry about /* get coeffs for this pitch period */ add_lpc_coeff(i, di_get_diph_lpc_mark(curr_diph,curr_mark,db), lpc_coeffs); /* get residual for this pitch period (and slice it) */ add_residual(as->inwidth[i], as->outwidth[i], res_pos, di_get_diph_res_mark(curr_diph,curr_mark,as->inwidth[i],db), residual); res_pos += as->outwidth[i]; } if (siod_get_lval("debug_reslpc",NULL) != NIL) { // residual.save("res.wav","nist"); lpc_coeffs.save("lpc.ascii","ascii"); } // Reconstruct the waveform if (siod_get_lval("lpc_fixedpoint",NULL) != NIL) lpc_resynth_fixp(lpc_coeffs,residual,as,output); else lpc_resynth(lpc_coeffs,residual,as,output); wfree(residual); } static void add_lpc_coeff(int nth,float *p_coeffs,EST_Track &lpc_coeff) { // Add coeffs frame to lpc_coeff at position lpc_coeff.copy_frame_in(nth, p_coeffs); } static void add_residual(int insize, int outsize, int position, short *p_residual, short *residual) { // Slice/pad this pitch's residual to outsize and add to full residual // at position. The mid points should align int i; int ibp = insize/2; int obp = outsize/2; if (insize < outsize) { // its too short so place it with 50% points aligned memmove(&residual[position+(obp-ibp)], p_residual, insize*sizeof(short)); } else // insize > outside { // Its too long so only take part of it memmove(&residual[position], &p_residual[ibp-obp], outsize*sizeof(short)); // Do power reduction float factor = (float)outsize/(float)insize; for (i=0; i < outsize; i++) residual[position+i] = (int)((float)residual[position+i]*factor); } #if 0 // Windowing (see if it makes a difference) #define DI_PI 3.14159265358979323846 float *window = walloc(float,outsize); int i; for (i=0; i < outsize/5; i++) window[i] = (0.5-(0.5*cos((double)(DI_PI*2*i)/(double)((2*outsize)/5-1)))); for (i=outsize/5; i < outsize; i++) window[i] = (0.5-(0.5*cos((double)(DI_PI*2*(i+(outsize/5)*3))/ (double)(8*outsize/5-1)))); #if 0 for (i=0,f=0; i < outsize/2+1; i++,f+=incr) window[i] = f; incr = 2.0/((outsize/2)); for (i=0,f=0; i < (outsize/2)+1; i++,f+=incr) window[outsize-1-i] = f; #endif for (i=0; i < outsize; i++) { residual.data()[position+i] = (short)(window[i]*(float)residual.data()[position+i]); } wfree(window); #endif } void ref2lpc(const float *rfc, float *lpc, int order) { // Here we use Christopher Longet Higgin's algorithm converted to // an equivalent by awb. It doesn't have hte reverse order or // negation requirement. float a,b; int n,k; for (n=0; n < order; n++) { lpc[n] = rfc[n]; for (k=0; 2*(k+1) <= n+1; k++) { a = lpc[k]; b = lpc[n-(k+1)]; lpc[k] = a-b*lpc[n]; lpc[n-(k+1)] = b-a*lpc[n]; } } } static void lpc_resynth(EST_Track &lpc_coeffs, short *residual, DIPHONE_ACOUSTIC *as, DIPHONE_OUTPUT *output) { // Reconstruct the waveform from the coefficients and residual // This is copied from Steve's donovan code with minor changes float *ptbuf; float *rbuf; int i,j,ci,cz; int zz,r; int NCOEFFS = lpc_coeffs.num_channels()-1; // lpc_power is 0 rbuf = walloc(float,NCOEFFS); ptbuf = walloc(float,output->o_max); memset(ptbuf,0,sizeof(float)*NCOEFFS); zz=NCOEFFS; r=0; float *tmm_coefs = new float[NCOEFFS+1]; for(i=0;ioutwidth[i] >= output->o_max) { cerr << "Diphone: output buffer overflow " << endl; break; } for(j=0 ;j < as->outwidth[i];j++,zz++,r++) { ptbuf[zz] = (float)residual[r]; // This loop is where all the time is spent for (ci=0,cz=zz-1; ci < NCOEFFS; ci++,cz--) ptbuf[zz] += rbuf[ci] * ptbuf[cz]; } } // Clipping is required because of some overflows which really points // to a lower level problem. Probably windowing the residual is what // I should really do. for (i=NCOEFFS,r=0; i < zz; i++,r++) if (ptbuf[i] > 32766) output->track[r] = (short)32766; else if (ptbuf[i] < -32766) output->track[r] = (short)-32766; else output->track[r] = (short)ptbuf[i]; output->o_sz = r; wfree(ptbuf); wfree(rbuf); delete [] tmm_coefs; } static void lpc_resynth_fixp(EST_Track &lpc_coeffs, short *residual, DIPHONE_ACOUSTIC *as, DIPHONE_OUTPUT *output) { // Reconstruct the waveform from the coefficients and residual // This is copied from Steve's donovan code with minor changes // This is the fixed point (sort of integer) version of the above // The coeffs are in the range -32768 to +32767 and where possible // integer calculations are done. This is primarily to allow // this to work on machine swithout a floating point processor // (i.e. awb's palmtop) int *ptbuf; int *rbuf; int i,j,ci,cz; int zz,r; int NCOEFFS = lpc_coeffs.num_channels()-1; // lpc_power is 0 rbuf = walloc(int,NCOEFFS); ptbuf = walloc(int,output->o_max); memset(ptbuf,0,sizeof(int)*NCOEFFS); zz=NCOEFFS; r=0; int *tmm_coefs = new int[NCOEFFS+1]; for(i=0;ioutwidth[i] >= output->o_max) { cerr << "Diphone: output buffer overflow " << endl; break; } for(j=0 ;j < as->outwidth[i];j++,zz++,r++) { ptbuf[zz] = ((int)residual[r]); // This loop is where all the time is spent for (ci=0,cz=zz-1; ci < NCOEFFS; ci++,cz--) ptbuf[zz] += (rbuf[ci] * ptbuf[cz]) / 32768; } } // Clipping is required because of some overflows which really points // to a lower level problem. Probably windowing the residual is what // I should really do. for (i=NCOEFFS,r=0; i < zz; i++,r++) { if (ptbuf[i] > 32766) output->track[r] = (short)32766; else if (ptbuf[i] < -32766) output->track[r] = (short)-32766; else output->track[r] = (short)ptbuf[i]; } output->o_sz = r; wfree(ptbuf); wfree(rbuf); delete [] tmm_coefs; } static int min(int a, int b) { return((a #include #include #include #include "festival.h" #include "diphone.h" #ifdef SUPPORT_PSOLA_TM #include "di_psolaTM.cc" #else void di_psola_tm(DIPHONE_DATABASE *db, DIPHONE_ACOUSTIC *as, DIPHONE_OUTPUT *output) { (void)db; (void)as; (void)output; cerr << "Diphone: di_psola is not available in this installation" << endl; festival_error(); } #endif festival/src/modules/diphone/di_select.cc0000644000175000017500000001522710126474330017135 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alistair Conkie */ /* Date : August 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* A free version of diphone selection and concatenation supported the */ /* standard CSTR diphone dbs */ /* */ /*=======================================================================*/ #include #include #include "festival.h" #include "diphone.h" static void bresenham_mod(DIPHONE_SPN *ps, DIPHONE_ACOUSTIC *as); static void nochange_mod(DIPHONE_SPN *ps, DIPHONE_ACOUSTIC *as); void di_frame_select(DIPHONE_DATABASE *database, DIPHONE_SPN *ps, DIPHONE_ACOUSTIC *as) { /* resist the temptation to call this select :-) */ int i, j; int bias = 0; for(i=0;ip_sz;i++) { ps->pm_req[i] = 0; ps->pm_giv[i] = 0; } /* count the needed pitch marks for each phoneme (or diph) -> ps */ j = 0; /* can j get too big ? */ for(i=0;ip_sz;i++) { if((as->cum_pitch[i]>=ps->cum_dur[j]) && (as->cum_pitch[i]cum_dur[j+1])) ps->pm_req[j] += 1; else if((as->cum_pitch[i]>=ps->cum_dur[j+1]) && (as->cum_pitch[i]cum_dur[j+2])) ps->pm_req[++j] += 1; else { *cdebug << "Diphone: some sort of pitch error" << endl; ps->pm_req[++j] += 1; } } /* count the given pitch marks for each phoneme (or diph) -> ps */ ps->pm_chg_diph[0] = 0; for(i=0;ip_sz-1;i++) { ps->pm_giv[i] += database->pm[ps->ref[i]]->lmark; ps->pm_chg_diph[i+1] = database->pm[ps->ref[i]]->rmark; ps->pm_giv[i+1] += database->pm[ps->ref[i]]->rmark; } if (siod_get_lval("diphone_no_mods",NULL) != NIL) nochange_mod(ps,as); else bresenham_mod(ps,as); /* there is something still to do in terms of selecting frames */ /* diph_mark */ bias = 0; for(i=0;ip_sz-1;i++) { if ((as->phon_mark[i] + bias) < 0) as->diph_mark[i] = as->phon_mark[i]; else if ((as->phon_mark[i] + bias) < database->pm[as->diph_ref[i]]->nmark) as->diph_mark[i] = as->phon_mark[i] + bias; else if (as->phon_mark[i] < database->pm[as->diph_ref[i]]->nmark) as->diph_mark[i] = as->phon_mark[i]; else // It needs less than one pitch mark; as->diph_mark[i] = database->pm[as->diph_ref[i]]->nmark-1; if ((as->diph_ref[i] != as->diph_ref[i+1]) && (i+1 != as->p_sz-1)) bias = - database->pm[as->diph_ref[i]]->rmark; if ((as->phon_ref[i] != as->phon_ref[i+1]) && (i+1 != as->p_sz-1)) bias = database->pm[as->diph_ref[i+1]]->lmark; } } static void nochange_mod(DIPHONE_SPN *ps, DIPHONE_ACOUSTIC *as) { // Does no duration or pitch modifications, useful in testing int ph,i,j; j=0; for(ph=0;php_sz;ph++) { for (i=0; i < ps->pm_giv[ph]; i++,j++) { as->phon_ref[j] = ph; as->phon_mark[j] = i; if (i >= ps->pm_chg_diph[ph]) as->diph_ref[j] = ps->ref[ph]; else as->diph_ref[j] = ps->ref[ph-1]; } } as->p_sz = j; } static void bresenham_mod(DIPHONE_SPN *ps, DIPHONE_ACOUSTIC *as) { int error = 0; int i,x,y,deltax,deltay, ph; as->p_tmp = 0; for(ph=0;php_sz;ph++) { error = 0; deltax = ps->pm_req[ph]; deltay = ps->pm_giv[ph]; x = 0; y = 0; if (deltax < deltay) { for (i=0;i=deltay) { as->phon_ref[as->p_tmp] = ph; as->phon_mark[as->p_tmp] = i-1; if ((i-1) >= ps->pm_chg_diph[ph]) as->diph_ref[as->p_tmp] = ps->ref[ph]; else as->diph_ref[as->p_tmp] = ps->ref[ph-1]; as->p_tmp++; /* printf("phon [%d], corr[%d] = %d\n",ph,x,i-1); */ x++; error -= deltay; } } } else { for(i=0;iphon_ref[as->p_tmp] = ph; as->phon_mark[as->p_tmp] = y; if(y >= ps->pm_chg_diph[ph]) as-> diph_ref[as->p_tmp] = ps->ref[ph]; else as-> diph_ref[as->p_tmp] = ps->ref[ph-1]; as->p_tmp++; /* printf("phon [%d], corr[%d] = %d\n",ph,i,y); */ error += deltay; if(error>deltax) { y++; error -= deltax; } } } } } festival/src/modules/diphone/di_pitch.cc0000644000175000017500000001126410126474330016762 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alistair Conkie */ /* Date : August 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* A free version of diphone selection and concatenation supported the */ /* standard CSTR diphone dbs */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "diphone.h" static int interpolated_freq(int k, DIPHONE_SPN *ps,int def_f0); static int interpolate(int a,int b,int c,int d,int e); void di_calc_pitch(DIPHONE_DATABASE *db, DIPHONE_SPN *ps, DIPHONE_ACOUSTIC *as) { int j,k; int y; int l = 0; int k_old = 0; int k_fine = 0; int x = 0; for(j=0;jt_sz;j++) ps->abs_targ[j] = (int)(ps->cum_dur[ps->targ_phon[j]] + ps->pc_targs[j]*ps->duration[ps->targ_phon[j]]/100.0); as->cum_pitch[0] = 0; for(k=0;kcum_dur[ps->p_sz];k+=100) { y = interpolated_freq(k,ps,db->def_f0); x += 100*y; while(x>db->samp_freq) { k_fine = k + interpolate(x-100*y,0,x,100,db->samp_freq); x -= db->samp_freq; as->pitch[l] = k_fine-k_old; as->cum_pitch[l+1] = as->pitch[l] + as->cum_pitch[l]; l++; if (l == as->p_max) { cerr << "Diphone: too many pitch marks\n"; festival_error(); } k_old = k_fine; } } as->p_sz = l; return; } static int interpolated_freq(int k, DIPHONE_SPN *ps,int def_f0) { int i; int freq; if(!ps->t_sz) return(def_f0); else if(kabs_targ[0]) return(ps->targ_freq[0]); else if(k>=ps->abs_targ[ps->t_sz-1]) return(ps->targ_freq[ps->t_sz-1]); for(i=1;it_sz;i++) { if((kabs_targ[i]) && (k>=ps->abs_targ[i-1])) { freq = interpolate(ps->abs_targ[i-1], ps->targ_freq[i-1], ps->abs_targ[i], ps->targ_freq[i],k); return(freq); } } return(-1); /* should never arrive here */ } static int interpolate(int a,int b,int c,int d,int e) { int f; f = (c*b + d*e - e*b -a*d)/(c-a); return(f); } festival/src/modules/diphone/di_io.cc0000644000175000017500000007720710126474330016273 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alistair Conkie */ /* Date : August 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* The groupfile stuff is all awb's fault. */ /* */ /*************************************************************************/ #include #include "EST_unix.h" #include #include "festival.h" #include "diphone.h" static unsigned int DIPHONE_MAGIC=0x46544449; /* FTDI */ static void load_index(DIPHONE_DATABASE *database); static void load_diphs(DIPHONE_DATABASE *database); static void load_lpc_file(DIPHONE_DATABASE *db,int diph,int mode); static void lpc2ref(const float *lpc, float *rfc, int order); static void extract_lpc_frames(DIPHONE_DATABASE *db, int diph, EST_Track &lpc); static void load_signal_file(DIPHONE_DATABASE *db, int i, int mode); static void database_malloc(int ndiphs, DIPHONE_DATABASE *database); static void di_group_load_signal(DIPHONE_DATABASE *db); static void di_group_load_lpc_params(DIPHONE_DATABASE *db); static void di_group_load_pm(DIPHONE_DATABASE *db); void di_load_database(DIPHONE_DATABASE *database) { // Load the ungrouped form database_malloc(database->ndiphs,database); load_index(database); load_diphs(database); } static void database_malloc(int ndiphs, DIPHONE_DATABASE *database) { // So why am I not using all those cute C++ classes ? // well I suppose I just don't know enough about binary loading // and saving to trust them, but that's a poor excuse. int i; database->nindex = 0; database->zone = 0; database->indx = walloc(DI_INDEX *,ndiphs); database->vox = walloc(DI_VOX *,ndiphs); database->pm = walloc(DI_PM *,ndiphs); database->lpc = walloc(DI_LPC *,ndiphs); for(i=0;iindx[i] = walloc(DI_INDEX,1); database->vox[i] = walloc(DI_VOX,1); database->vox[i]->signal = 0; database->pm[i] = walloc(DI_PM,1); database->pm[i]->mark = 0; database->lpc[i] = walloc(DI_LPC,1); database->lpc[i]->f = 0; } } static void load_index(DIPHONE_DATABASE *database) { EST_TokenStream ts; int i; EST_String line; if (ts.open(database->index_file) == -1) { cerr << "Diphone: Can't open file " << database->index_file << endl; festival_error(); } for (i=0; (!ts.eof()) && (indiphs);) { line = ts.get_upto_eoln(); if ((line.length() > 0) && (line[0] != ';')) { EST_TokenStream ls; ls.open_string(line); database->indx[i]->diph = wstrdup(ls.get().string()); database->indx[i]->file = wstrdup(ls.get().string()); database->indx[i]->beg = atof(ls.get().string()); database->indx[i]->mid = atof(ls.get().string()); database->indx[i]->end = atof(ls.get().string()); ls.close(); i++; } } if (i == database->ndiphs) { cerr << "Diphone: too many diphones in DB" << endl; festival_error(); } database->nindex = i; database->ndiphs = i; ts.close(); } static void load_diphs(DIPHONE_DATABASE *database) { int i; for(i=0;inindex;i++) { load_signal_file(database,i,database->sig_access_type); // if (database->type = di_lpc) // load_lpc_file(database,i,database->sig_access_type); load_pitch_file(database,i,database->sig_access_type); } } static void load_lpc_file(DIPHONE_DATABASE *db,int diph,int mode) { // Load LPC coefficients EST_String lpc_file; EST_Track lpc; if (db->lpc[diph]->f != 0) return; // already loaded if (mode == di_direct) { lpc_file = EST_String(db->lpc_dir) + db->indx[diph]->file + db->lpc_ext; if (lpc.load(lpc_file) != format_ok) { cerr << "Diphone: failed to read lpc file " << lpc_file << endl; festival_error(); } if (lpc.num_channels() != db->lpc_order) { cerr << "Diphone: lpc file " << lpc_file << " has order " << lpc.num_channels() << " while database has " << db->lpc_order << endl; festival_error(); } // Extract frames (pitch synchronously) extract_lpc_frames(db,diph,lpc); } return; } static void ref2lpc(const float *rfc, float *lpc, int order) { // Here we use Christopher Longet Higgin's algorithm converted to // an equivalent by awb. Its doesn't have hte reverse order or // negation requirement. float a,b; int n,k; for (n=0; n < order; n++) { lpc[n] = rfc[n]; for (k=0; 2*(k+1) <= n+1; k++) { a = lpc[k]; b = lpc[n-(k+1)]; lpc[k] = a-b*lpc[n]; lpc[n-(k+1)] = b-a*lpc[n]; } } } static void extract_lpc_frames(DIPHONE_DATABASE *db, int diph, EST_Track &lpc) { // Extract LPC frames from lpc, one for each pitch mark int frame_num; float pos,factor; float ps_pos; int i,j,k; db->lpc[diph]->f = walloc(float *,db->pm[diph]->nmark); float *lpcs = walloc(float,lpc.num_channels()); for (i=0; i < db->pm[diph]->nmark; i++) { if (db->lpc_pitch_synch) { db->lpc[diph]->f[i] = walloc(float,lpc.num_channels()); pos = (((float)db->pm[diph]->mark[i]-db->sig_band)/ (float)db->samp_freq) + (db->indx[diph]->beg/1000.0); for (j=1,ps_pos=0; jlpc[diph]->f[i][0] = lpcs[0]; lpc2ref(&lpcs[1],&db->lpc[diph]->f[i][1], lpc.num_channels()-1); break; } } if (j==lpc.num_frames()) { cerr << "Diphone: lpc access, failed to find lpc coeffs" << endl; festival_error(); } } else { // Not pitch synchronous so find closest frames and // interpolate between them db->lpc[diph]->f[i] = walloc(float,lpc.num_channels()); // position of current mark in seconds pos = (((float)db->pm[diph]->mark[i]-db->sig_band)/ (float)db->samp_freq) + (db->indx[diph]->beg/1000.0); // Convert to frames, rounding and subtracting start offset frame_num = (int)((pos/db->lpc_frame_shift)) - db->lpc_frame_offset; if (frame_num+1 < lpc.num_frames()) { // Interpolate between them factor = (pos - ((1+frame_num)*db->lpc_frame_shift))/ db->lpc_frame_shift; for (j=0; j < lpc.num_channels(); j++) { db->lpc[diph]->f[i][j] = lpc(frame_num,j) + (factor * (lpc(frame_num+1,j)- lpc(frame_num,j))); } } if (frame_num >= lpc.num_frames()) { cerr << "Diphone: LPC frame past end of file \"" << db->indx[diph]->file << "\"" << endl; memset(db->lpc[diph]->f[i],0,sizeof(float)*lpc.num_channels()); } else // Last one so just take it as is { lpc.copy_frame_out(frame_num, db->lpc[diph]->f[i], 0, lpc.num_channels()); } } } db->lpc[diph]->nframes = db->pm[diph]->nmark; wfree(lpcs); // db->lpc_order = lpc.num_channels(); } static void lpc2ref(const float *lpc, float *rfc, int order) { // LPC to reflection coefficients // from code from Borja Etxebarria int i,j; float f,ai; float *vo,*vx; float *vn = new float[order]; i = order-1; rfc[i] = ai = lpc[i]; f = 1-ai*ai; i--; for (j=0; j<=i; j++) rfc[j] = (lpc[j]+((ai*lpc[i-j])))/f; /* vn=vtmp in previous #define */ vo=rfc; for ( ;i>0; ) { ai=vo[i]; f = 1-ai*ai; i--; for (j=0; j<=i; j++) vn[j] = (vo[j]+((ai*vo[i-j])))/f; rfc[i]=vn[i]; vx = vn; vn = vo; vo = vx; } delete [] vn; } static void load_signal_file(DIPHONE_DATABASE *db, int i, int mode) { // Load signal (or lpc residual) file) int beg_samp,end_samp,zone,nsamples; EST_String signal_file; EST_String sig_type; int offset,error; beg_samp = 0; zone = 0; nsamples = 0; if (db->gtype == di_ungrouped) { beg_samp = (int)((db->indx[i]->beg)/1000.0*db->samp_freq); end_samp = (int)((db->indx[i]->end)/1000.0*db->samp_freq); nsamples = end_samp - beg_samp; zone = db->sig_band; db->zone = zone; db->vox[i]->nsamples = nsamples+(2*zone); db->vox[i]->signal = 0; } if (mode == di_direct) { if (db->gtype == di_ungrouped) { EST_Wave w; if (db->type == di_lpc) { signal_file = EST_String(db->lpc_dir) + EST_String(db->indx[i]->file) + EST_String(db->lpc_res_ext); sig_type = db->lpc_res_type; // Different LPC techniques will leave various offsets // in the residule, you have to specify this explicitly beg_samp -= (int)(db->lpc_res_offset * db->samp_freq); } else { signal_file = EST_String(db->signal_dir) + EST_String(db->indx[i]->file) + EST_String(db->signal_ext); sig_type = db->signal_type; } offset = beg_samp-zone; if (offset < 0) offset = 0; if (w.load_file(signal_file,sig_type, db->samp_freq, "short", EST_NATIVE_BO, 1, offset, nsamples+2*zone) != format_ok) { cerr << "Diphone: failed to read " << sig_type << " format signal file " << signal_file << endl; festival_error(); } db->vox[i]->signal = walloc(short,w.num_samples()); if (beg_samp-zone < 0) // wasn't enough space at beginning error = abs(beg_samp-zone); else error = 0; memset(db->vox[i]->signal,0,error*sizeof(short)); for (int j=0; j < w.num_samples()-error; j++) db->vox[i]->signal[error+j] = w(j); db->vox[i]->nsamples = w.num_samples()-error; } else // grouped so have to access the group file { if (db->gfd == NULL) { cerr << "Diphone: can no longer access the group file" << endl; festival_error(); } if (db->group_encoding == di_raw) { db->vox[i]->signal = walloc(short,db->vox[i]->nsamples); fseek(db->gfd,db->gsignalbase+(db->offsets[i]*2),SEEK_SET); fread(db->vox[i]->signal,sizeof(short), db->vox[i]->nsamples,db->gfd); if (db->swap) swap_bytes_short(db->vox[i]->signal,db->vox[i]->nsamples); } else if (db->group_encoding == di_ulaw) { unsigned char *ulaw = walloc(unsigned char,db->vox[i]->nsamples); db->vox[i]->signal = walloc(short,db->vox[i]->nsamples); fseek(db->gfd,db->gsignalbase+(db->offsets[i]),SEEK_SET); fread(ulaw,sizeof(unsigned char),db->vox[i]->nsamples,db->gfd); ulaw_to_short(ulaw,db->vox[i]->signal,db->vox[i]->nsamples); wfree(ulaw); } else { cerr << "Diphone: unknown group type" << endl; festival_error(); } } } } void load_pitch_file(DIPHONE_DATABASE *database, int i, int mode) { // load files from newer Track format int mark[5000]; EST_String pitch_file; EST_Track pms; float fnum; int k,k1,k2,m,zone,beg_samp,p; if ((database->pm[i]->mark != 0) || (mode != di_direct)) return; pitch_file = EST_String(database->pitch_dir)+database->indx[i]->file+ database->pitch_ext; if (pms.load(pitch_file) != format_ok) { cerr << "Diphone: Can't open pitch file " << pitch_file << endl; festival_error(); } /* assumptions.. only those within the limits of the diphone */ beg_samp = (int)((database->indx[i]->beg)/1000.0*database->samp_freq); zone = database->sig_band; k = 0; k1 = 0; k2 = 0; for (p=0; p < pms.num_frames(); p++) { fnum = pms.t(p)*1000.0; if((fnum>database->indx[i]->beg) && (fnumindx[i]->mid)) { mark[k] = (int)(fnum/1000.0*database->samp_freq - beg_samp + zone); if ((mark[k] >= database->vox[i]->nsamples+zone) || (mark[k] > 64534)) { fprintf(stderr,"Diphone: Mark out of range -- too large %s\n", (const char *)pitch_file); k--; k1--; } if(mark[k] < zone) { fprintf(stderr,"Diphone: Mark out of range -- too small %s\n", (const char *)pitch_file); k--; k1--; } k++; k1++; } else if((fnum>=database->indx[i]->mid) && (fnumindx[i]->end)) { mark[k] = (int)(fnum/1000.0*database->samp_freq - beg_samp + zone); if ((mark[k] >= database->vox[i]->nsamples+zone) || (mark[k] > 64534)) { fprintf(stderr,"Diphone: Mark out of range -- too large %s\n", (const char *)pitch_file); k--; k2--; } if(mark[k] < zone) { fprintf(stderr,"Diphone: Mark out of range -- too small %s\n", (const char *)pitch_file); k--; k2--; } k++; k2++; } } database->pm[i]->mark = walloc(unsigned short,k); for(m=0;mpm[i]->mark[m] = (unsigned short)mark[m]; database->pm[i]->nmark = (unsigned short)k; database->pm[i]->lmark = (unsigned short)k1; database->pm[i]->rmark = (unsigned short)k2; if (database->pm[i]->rmark == 0) { *cdebug << "Diphone: modifying edge pms for " << database->indx[i]->diph << endl; database->pm[i]->rmark = 1; database->pm[i]->lmark -= 1; } if (database->pm[i]->nmark <= 0) { cerr << "Diphone: diphone " << database->indx[i]->diph << " has 0 pitchmarks" << endl; festival_error(); } } #if 0 void load_pitch_file(DIPHONE_DATABASE *database, int i, int mode) { char s[100]; int mark[5000]; EST_String pitch_file; FILE *pfd; float fnum; int k,k1,k2,m,zone,beg_samp; if ((database->pm[i]->mark != 0) || (mode != di_direct)) return; pitch_file = EST_String(database->pitch_dir)+database->indx[i]->file+ database->pitch_ext; if((pfd=fopen(pitch_file,"rb")) == NULL) { cerr << "Diphone: Can't open pitch file " << pitch_file << endl; festival_error(); } /* assumptions.. only those within the limits of the diphone */ beg_samp = (int)((database->indx[i]->beg)/1000.0*database->samp_freq); zone = database->sig_band; k = 0; k1 = 0; k2 = 0; while(fgets(s,100,pfd) != NULL) { sscanf(s,"%f",&fnum); if((fnum>database->indx[i]->beg) && (fnumindx[i]->mid)) { mark[k] = (int)(fnum/1000.0*database->samp_freq - beg_samp + zone); if ((mark[k] >= database->vox[i]->nsamples+zone) || (mark[k] > 64534)) { fprintf(stderr,"Diphone: Mark out of range -- too large %s\n", (const char *)pitch_file); k--; k1--; } if(mark[k] < zone) { fprintf(stderr,"Diphone: Mark out of range -- too small %s\n", (const char *)pitch_file); k--; k1--; } k++; k1++; } else if((fnum>=database->indx[i]->mid) && (fnumindx[i]->end)) { mark[k] = (int)(fnum/1000.0*database->samp_freq - beg_samp + zone); if ((mark[k] >= database->vox[i]->nsamples+zone) || (mark[k] > 64534)) { fprintf(stderr,"Diphone: Mark out of range -- too large %s\n", (const char *)pitch_file); k--; k2--; } if(mark[k] < zone) { fprintf(stderr,"Diphone: Mark out of range -- too small %s\n", (const char *)pitch_file); k--; k2--; } k++; k2++; } } database->pm[i]->mark = walloc(unsigned short,k); for(m=0;mpm[i]->mark[m] = (unsigned short)mark[m]; database->pm[i]->nmark = (unsigned short)k; database->pm[i]->lmark = (unsigned short)k1; database->pm[i]->rmark = (unsigned short)k2; if (database->pm[i]->rmark == 0) { *cdebug << "Diphone: modifying edge pms for " << database->indx[i]->diph << endl; database->pm[i]->rmark = 1; database->pm[i]->lmark -= 1; } if (database->pm[i]->nmark <= 0) { cerr << "Diphone: diphone " << database->indx[i]->diph << " has 0 pitchmarks" << endl; festival_error(); } fclose(pfd); } #endif /* Buffer to hold current diphone signal when using ondemand access */ /* method. It remembers the last phone accessed as typical access is */ /* the same one for a few times */ static short *diph_buffer = 0; static int diph_max_size = 0; static int last_diph = -1; static DIPHONE_DATABASE *last_db = 0; short *di_get_diph_signal(int diph,DIPHONE_DATABASE *db) { // Get the diphone signal (or residual) from wherever if (db->sig_access_type == di_direct) // all pre-loaded return db->vox[diph]->signal; else if (db->sig_access_type == di_dynamic) // Load and keep { if (db->vox[diph]->signal == 0) load_signal_file(db,diph,di_direct); return db->vox[diph]->signal; } else if (db->sig_access_type == di_ondemand) // Load and free afterwards { // Loads it into a common buffer, over written each time if ((diph == last_diph) && (db == last_db)) // ensure db hasn't changed return diph_buffer; load_signal_file(db,diph,di_direct); if (diph_max_size < db->vox[diph]->nsamples) { wfree(diph_buffer); diph_buffer = walloc(short,db->vox[diph]->nsamples); diph_max_size = db->vox[diph]->nsamples; } memmove(diph_buffer,db->vox[diph]->signal, db->vox[diph]->nsamples*sizeof(short)); wfree(db->vox[diph]->signal); db->vox[diph]->signal = 0; last_db = db; last_diph = diph; return diph_buffer; } else { cerr << "Diphone: unknown diphone signal access strategy" << endl; festival_error(); } return NULL; } /* The buffer used to hold the requested frame */ static float frame_buff[128]; float *di_get_diph_lpc_mark(int diph,int mark,DIPHONE_DATABASE *db) { // Get the coeff frame for diph at mark load_lpc_file(db,diph,di_direct); memmove(frame_buff, db->lpc[diph]->f[mark], sizeof(float)*db->lpc_order); return frame_buff; } short *di_get_diph_res_mark(int diph,int mark,int size,DIPHONE_DATABASE *db) { // Get the residual for diph at mark, use the signal field // to hold it as they are so similar. short *residual; residual = di_get_diph_signal(diph,db); // Take residual around this midpoint int pos_samp = db->pm[diph]->mark[mark] - size/2; if (pos_samp < 0) { pos_samp = 0; *cdebug << "DIPHONE: sig_band too short to the left" << endl; } if (pos_samp+size >= db->vox[diph]->nsamples) { pos_samp = db->vox[diph]->nsamples - size; *cdebug << "DIPHONE: sig_band too short to the right" << endl; } return &residual[pos_samp]; } void di_load_grouped_db(const EST_Pathname &filename, DIPHONE_DATABASE *db, LISP global_params) { // Get index file and saved data from grouped file int i,j; unsigned int magic; int strsize; char *diphnames; LISP params; if ((db->gfd=fopen(filename,"rb")) == NULL) { cerr << "Diphone: cannot open group file " << filename << " for reading" << endl; festival_error(); } fread(&magic,sizeof(int),1,db->gfd); if (magic == SWAPINT(DIPHONE_MAGIC)) db->swap = TRUE; else if (magic != DIPHONE_MAGIC) { cerr << "Diphone: " << filename << " not a group file" << endl; festival_error(); } params = lreadf(db->gfd); // read the parameters in LISP form di_general_parameters(db,params); // some may be reset later di_fixed_parameters(db,params); di_general_parameters(db,global_params); // reset some params database_malloc(db->ndiphs,db); db->nindex = db->ndiphs; // we can trust that number this time fread(&strsize,sizeof(int),1,db->gfd); // number of chars in diph names if (db->swap) strsize = SWAPINT(strsize); diphnames = walloc(char,strsize); fread(diphnames,sizeof(char),strsize,db->gfd); for (j=i=0;inindex;i++) { db->indx[i]->diph = &diphnames[j]; db->indx[i]->file = 0; for ( ; diphnames[j] != '\0'; j++) // skip to next diphname if (j > strsize) { cerr << "Diphone: group file diphone name table corrupted" << endl; festival_error(); } j++; } // Diphone signals di_group_load_signal(db); // Diphone LPC parameters if (db->type == di_lpc) di_group_load_lpc_params(db); // Diphone Pitch marks di_group_load_pm(db); if (db->sig_access_type == di_direct) { fclose(db->gfd); // read eveything db->gfd = 0; } } static void di_group_load_signal(DIPHONE_DATABASE *db) { int i; unsigned short *samp_counts; int sample_offset,totsamples; samp_counts = walloc(unsigned short,db->nindex); fread(samp_counts,sizeof(unsigned short),db->nindex,db->gfd); if (db->swap) swap_bytes_ushort(samp_counts,db->nindex); fread(&totsamples,sizeof(int),1,db->gfd); if (db->swap) totsamples = SWAPINT(totsamples); if (db->sig_access_type == di_direct) { if (db->group_encoding == di_raw) { db->allsignal = walloc(short,totsamples); fread(db->allsignal,sizeof(short),totsamples,db->gfd); if (db->swap) swap_bytes_short(db->allsignal,totsamples); } else if (db->group_encoding == di_ulaw) { db->allulawsignal = walloc(unsigned char,totsamples); fread(db->allulawsignal,sizeof(unsigned char),totsamples,db->gfd); } } else { db->gsignalbase = ftell(db->gfd); db->offsets = walloc(int,db->nindex); } sample_offset = 0; for (i=0; i < db->nindex; i++) { db->vox[i]->nsamples = samp_counts[i]; if (db->sig_access_type == di_direct) { if (db->group_encoding == di_raw) db->vox[i]->signal = &db->allsignal[sample_offset]; else if (db->group_encoding == di_ulaw) { db->vox[i]->signal = walloc(short,samp_counts[i]); ulaw_to_short(&db->allulawsignal[sample_offset], db->vox[i]->signal,samp_counts[i]); } else { cerr << "Diphone: unknown group type to unpack" << endl; festival_error(); } } else { db->offsets[i] = sample_offset; db->vox[i]->signal = 0; } sample_offset += samp_counts[i]; } if (db->sig_access_type != di_direct) if (db->group_encoding == di_ulaw) fseek(db->gfd,(long)sample_offset,SEEK_CUR); else fseek(db->gfd,(long)sample_offset*sizeof(short),SEEK_CUR); wfree(samp_counts); } static void di_group_load_lpc_params(DIPHONE_DATABASE *db) { // LPC params are always fully loaded int totframes; int i,j,k; unsigned short *frame_counts; int frame_offset; int this_frame; frame_counts = walloc(unsigned short, db->nindex); fread(frame_counts,sizeof(unsigned short),db->nindex,db->gfd); if (db->swap) swap_bytes_ushort(frame_counts,db->nindex); fread(&totframes,sizeof(int),1,db->gfd); if (db->swap) totframes = SWAPINT(totframes); if (db->group_encoding == di_raw) // its as floats { db->allframes = walloc(float,totframes*db->lpc_order); fread(db->allframes,sizeof(float), totframes*db->lpc_order,db->gfd); if (db->swap) swap_bytes_float(db->allframes,totframes*db->lpc_order); } else if (db->group_encoding == di_ulaw) // its as shorts { db->allframesshort = walloc(short,totframes*db->lpc_order); fread(db->allframesshort,sizeof(short), totframes*db->lpc_order,db->gfd); if (db->swap) swap_bytes_short(db->allframesshort, totframes*db->lpc_order); } frame_offset = 0; for (i=0; i < db->nindex; i++) { db->lpc[i]->nframes = frame_counts[i]; db->lpc[i]->f = walloc(float *,frame_counts[i]); if (db->group_encoding == di_raw) for (j=0;jlpc[i]->nframes;j++) db->lpc[i]->f[j] = &db->allframes[(frame_offset+j)*db->lpc_order]; else if (db->group_encoding == di_ulaw) { int fixedpoint = FALSE; if (siod_get_lval("lpc_fixedpoint",NULL) != NIL) fixedpoint = TRUE; for (j=0;jlpc[i]->nframes;j++) { db->lpc[i]->f[j] = walloc(float,db->lpc_order); this_frame = (frame_offset+j)*db->lpc_order; if (fixedpoint) for (k=0;klpc_order;k++) db->lpc[i]->f[j][k] = (float)db->allframesshort[this_frame+k]; else for (k=0;klpc_order;k++) db->lpc[i]->f[j][k] = (float)db->allframesshort[this_frame+k]/32766.0; } } else { cerr << "Diphone: unknown group type to unpack" << endl; festival_error(); } frame_offset += frame_counts[i]; } wfree(db->allframesshort); db->allframesshort = 0; // don't really need this any more wfree(frame_counts); } static void di_group_load_pm(DIPHONE_DATABASE *db) { unsigned short *pm_info; int i,j; pm_info = walloc(unsigned short,db->nindex*3); if (fread(pm_info,sizeof(unsigned short),db->nindex*3,db->gfd) != (unsigned int)(db->nindex*3)) { cerr << "DIPHONE: short group file, can't read pm\n"; festival_error(); } if (db->swap) for (i=0; i < db->nindex*3; i++) pm_info[i] = SWAPSHORT(pm_info[i]); for (i=0; i < db->nindex; i++) { db->pm[i]->mark = walloc(unsigned short,pm_info[i*3]); db->pm[i]->nmark = pm_info[i*3]; db->pm[i]->lmark = pm_info[(i*3)+1]; db->pm[i]->rmark = pm_info[(i*3)+2]; fread(db->pm[i]->mark,sizeof(unsigned short),db->pm[i]->nmark,db->gfd); if (db->swap) for (j=0; j < db->pm[i]->nmark; j++) db->pm[i]->mark[j] = SWAPSHORT(db->pm[i]->mark[j]); } } static LISP di_enlispen_params(DIPHONE_DATABASE *db) { // Return lisp representation of the parameters in db return cons(make_param_str("name",db->name), cons(make_param_str("type",db->type_str), cons(make_param_str("index_file",db->index_file), cons(make_param_str("signal_dir",db->signal_dir), cons(make_param_str("signal_ext",db->signal_ext), cons(make_param_str("signal_type",db->signal_type), cons(make_param_str("pitch_dir",db->pitch_dir), cons(make_param_str("pitch_ext",db->pitch_ext), cons(make_param_str("lpc_dir",db->lpc_dir), cons(make_param_str("lpc_ext",db->lpc_ext), cons(make_param_str("lpc_res_ext",db->lpc_res_ext), cons(make_param_str("lpc_type",db->lpc_type), cons(make_param_str("lpc_res_type",db->lpc_res_type), cons(make_param_float("lpc_res_offset",db->lpc_res_offset), cons(make_param_int("lpc_frame_offset",db->lpc_frame_offset), cons(make_param_int("lpc_order",db->lpc_order), cons(make_param_float("lpc_frame_shift",db->lpc_frame_shift), cons(make_param_int("samp_freq",db->samp_freq), cons(make_param_str("phoneset",db->phoneset), cons(make_param_str("access_type",db->sig_access_type_str), cons(make_param_str("group_encoding",db->group_encoding_str), cons(make_param_int("num_diphones",db->nindex), cons(make_param_int("sig_band",db->sig_band), cons(make_param_int("def_f0",db->def_f0), cons(make_param_str("default_diphone",db->default_diphone), cons(make_param_lisp("alternates_before",db->alternates_before), cons(make_param_lisp("alternates_after",db->alternates_after), NIL))))))))))))))))))))))))))); } void di_save_grouped_db(const EST_Pathname &filename, DIPHONE_DATABASE *db) { // Get index file and saved data from grouped file FILE *fd; LISP params; int strsize,totsamples,totframes; int i,j,k; if ((fd=fopen(filename,"wb")) == NULL) { cerr << "Diphone: cannot open group file " << filename << " for saving" << endl; festival_error(); } fwrite(&DIPHONE_MAGIC,sizeof(int),1,fd); params = di_enlispen_params(db); // get lisp representation of parameters lprin1f(params,fd); // Only need to dump the diphone names, not the rest of the indx info strsize = 0; for (i=0;inindex;i++) strsize += strlen(db->indx[i]->diph)+1; fwrite(&strsize,sizeof(int),1,fd); for (i=0;inindex;i++) fwrite(db->indx[i]->diph,sizeof(char),strlen(db->indx[i]->diph)+1,fd); // Diphone Signals // Dump the signal sizes first to make reading easier totsamples = 0; for (i=0;inindex;i++) { if (db->vox[i]->signal == 0) // in case it isn't loaded yet { load_pitch_file(db,i,di_direct); load_signal_file(db,i,di_direct); } fwrite(&db->vox[i]->nsamples,sizeof(unsigned short),1,fd); totsamples += db->vox[i]->nsamples; } fwrite(&totsamples,sizeof(int),1,fd); // Dump signals (compressed if necessary) for (i=0;inindex;i++) { if (db->group_encoding == di_raw) fwrite(db->vox[i]->signal,sizeof(short),db->vox[i]->nsamples,fd); else if (db->group_encoding == di_ulaw) { unsigned char *ulaw = walloc(unsigned char,db->vox[i]->nsamples); short_to_ulaw(db->vox[i]->signal,ulaw,db->vox[i]->nsamples); fwrite(ulaw,sizeof(unsigned char),db->vox[i]->nsamples,fd); wfree(ulaw); } else { cerr << "Diphone: unknown group type for dumping" << endl; festival_error(); } } // Diphone LPC parameters if (db->type == di_lpc) { for (i=0;inindex;i++) // ensure they are all loaded load_lpc_file(db,i,di_direct); totframes = 0; for (i=0;inindex;i++) { fwrite(&db->lpc[i]->nframes,sizeof(unsigned short),1,fd); totframes += db->lpc[i]->nframes; } fwrite(&totframes,sizeof(int),1,fd); for (i=0;inindex;i++) { if (db->group_encoding == di_raw) // saved as floats { for (j=0; jlpc[i]->nframes; j++) fwrite(db->lpc[i]->f[j],sizeof(float),db->lpc_order,fd); } else if (db->group_encoding == di_ulaw) // saved as shorts { short *sh = new short[db->lpc_order]; for (j=0; jlpc[i]->nframes; j++) { for (k=0; klpc_order; k++) sh[k] = (short)(db->lpc[i]->f[j][k]*32766.0); fwrite(sh,sizeof(short),db->lpc_order,fd); } delete sh; } else { cerr << "Diphone: unknown group type for dumping" << endl; festival_error(); } } } // Diphone Pitch Marks for (i=0;inindex;i++) { fwrite(&db->pm[i]->nmark,sizeof(unsigned short),1,fd); fwrite(&db->pm[i]->lmark,sizeof(unsigned short),1,fd); fwrite(&db->pm[i]->rmark,sizeof(unsigned short),1,fd); } for (i=0;inindex;i++) { fwrite(db->pm[i]->mark,sizeof(unsigned short),db->pm[i]->nmark,fd); i=i; } fclose(fd); } festival/src/modules/diphone/oc.cc0000644000175000017500000002450310126474330015600 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : December 1996 */ /*-----------------------------------------------------------------------*/ /* Optimal Coupling for two pieces of speech */ /* */ /* Given two Tracks find the best point (minimised weighted euclidean */ /* distance between two vectors). */ /* */ /*=======================================================================*/ #include #include "festival.h" #define LEFT_PHONE(x) (car(x)) #define RIGHT_PHONE(x) (car(cdr(x))) #define FILEID(x) (car(cdr(cdr(x)))) #define START(x) (car(cdr(cdr(cdr(x))))) #define MID(x) (car(cdr(cdr(cdr(cdr(x)))))) #define END(x) (car(cdr(cdr(cdr(cdr(cdr(x))))))) static int get_track_for_phone(EST_Track &a,const EST_String &fileid,float st,float end); static LISP before_ds(LISP d1, LISP ds); static LISP after_ds(LISP d1, LISP ds); static float find_best_left(LISP d,LISP ds,LISP weights); static float find_best_right(LISP d,LISP ds,LISP weights); static float frametoms(int frame,float frame_shift); static int mstoframe(float ms,float frame_shift); static float frame_distance(EST_Track &a, int fa, EST_Track &b, int fb, int size, double *weights); static EST_String coeffs_dir = "coeffs/"; static EST_String coeffs_ext = ".cep"; LISP find_optimal_coupling(LISP table, LISP weights) { // For each diphone description in table find the best overall // join point between it and all other diphone sthat can join with it LISP d,newtab,newent; float best_left,best_right; newtab = NIL; gc_protect(&newtab); coeffs_dir = get_c_string(siod_get_lval("oc_coeffs_dir","no coeffs dir")); coeffs_ext = get_c_string(siod_get_lval("oc_coeffs_ext","no coeffs ext")); for (d=table; d != NIL; d=cdr(d)) { pprint(car(d)); if (ph_is_silence(get_c_string(LEFT_PHONE(car(d))))) best_left = get_c_float(START(car(d))); else best_left=find_best_left(car(d),before_ds(car(d),table),weights); if (ph_is_silence(get_c_string(RIGHT_PHONE(car(d))))) best_right = get_c_float(END(car(d))); else best_right=find_best_right(car(d),after_ds(car(d),table),weights); newent = cons(LEFT_PHONE(car(d)), // left phone cons(RIGHT_PHONE(car(d)), // right phone cons(FILEID(car(d)), // file_id cons(flocons(best_left), // left cut point cons(MID(car(d)), // mid point cons(flocons(best_right),NIL)))))); // right cut point newtab = cons(newent,newtab); } newtab = reverse(newtab); gc_unprotect(&newtab); return newtab; } static int get_track_for_phone(EST_Track &a, const EST_String &fileid, float st, float end) { // Load EST_Track from fileid from st to end (in ms) EST_Track whole; int start_frame, end_frame, i; if (whole.load(coeffs_dir+fileid+coeffs_ext) != 0) return -1; start_frame = mstoframe(st-12.8,whole.shift()); end_frame = mstoframe(end-12.8,whole.shift())+1; a.resize(end_frame-start_frame,whole.num_channels()); for (i=start_frame; i < end_frame; i++) a.copy_frame_in(i-start_frame, whole, i, 0, 0, whole.num_channels()); a.fill_time(whole.shift()); return 0; } static LISP before_ds(LISP d1, LISP ds) { // Return all entries in ds whose cadr equals d1's car // i.e. all diphones which match d1's left phone LISP m=NIL,l; for (l=ds; l != NIL; l=cdr(l)) if (streq(get_c_string(car(d1)),get_c_string(car(cdr(car(l)))))) m=cons(car(l),m); return m; } static LISP after_ds(LISP d1, LISP ds) { // Return all entries in ds whose car equals d1's cadr // i..e all diphones which match d1's right phone LISP m=NIL,l; for (l=ds; l != NIL; l=cdr(l)) if (streq(get_c_string(car(cdr(d1))),get_c_string(car(car(l))))) m=cons(car(l),m); return m; } static float find_best_left(LISP d,LISP ds,LISP weights) { // Find the best join point with each of phones described // in d EST_Track a,b; LISP l; int i,j,best,bestj;; double b_dist,dist; float best_pos; get_track_for_phone(a,get_c_string(FILEID(d)), get_c_float(START(d)),get_c_float(MID(d))); // Cummulate the costs for each possible cut point double *counts = new double[a.num_frames()]; for (i=0; i < a.num_frames(); i++) counts[i] = 0; double *w = new double[siod_llength(weights)]; for (l=weights,i=0; i < siod_llength(weights); i++,l=cdr(l)) w[i] = get_c_float(car(l)); for (l=ds; l != NIL; l=cdr(l)) { // for each matching phone get_track_for_phone(b,get_c_string(FILEID(car(l))), get_c_float(MID(car(l))),get_c_float(END(car(l)))); best=1; b_dist = frame_distance(a, 1, b, 0, a.num_channels(),w); for (i=1; i < a.num_frames()-1; i++) { for (j=0; j < b.num_frames(); j++) { dist = frame_distance(a, i, b, j, a.num_channels(),w); if (dist < b_dist) { b_dist = dist; best = i; bestj = j; } } } // You should probably find minimise the std // printf("best pos %d %s-%s %s-%s\n",best, // get_c_string(LEFT_PHONE(d)),get_c_string(RIGHT_PHONE(d)), // get_c_string(LEFT_PHONE(car(l))),get_c_string(RIGHT_PHONE(car(l)))); counts[best] += 1; // sum the best possible // counts[best] += b_dist; // sum the best possible } // Now find out the best position best = 0; for (i=0; i < a.num_frames(); i++) { if (counts[i] > counts[best]) best = i; } // Change frame number back to ms offset best_pos = frametoms(mstoframe(get_c_float(START(d)),a.shift()) + best,a.shift()); delete counts; delete w; return best_pos; } static float find_best_right(LISP d,LISP ds,LISP weights) { // Find the best join point with each of phones described // in d EST_Track a,b; LISP l; int i,j,best,bestj;; double b_dist,dist; float best_pos; get_track_for_phone(a,get_c_string(FILEID(d)), get_c_float(MID(d)),get_c_float(END(d))); // Cummulate the costs for each possible cut point double *counts = new double[a.num_frames()]; for (i=0; i < a.num_frames(); i++) counts[i] = 0; double *w = new double[siod_llength(weights)]; for (l=weights,i=0; i < siod_llength(weights); i++,l=cdr(l)) w[i] = get_c_float(car(l)); for (l=ds; l != NIL; l=cdr(l)) { // for each matching phone get_track_for_phone(b,get_c_string(FILEID(car(l))), get_c_float(START(car(l))), get_c_float(MID(car(l)))); best=1; b_dist = frame_distance( a, 1, b, 0, a.num_channels(),w); for (i=1; i < a.num_frames()-1; i++) { for (j=0; j < b.num_frames(); j++) { dist = frame_distance( a, i, b, j, a.num_channels(),w); if (dist < b_dist) { b_dist = dist; best = i; bestj = j; } } } // You should probably find minimise the std counts[best] += 1; // sum the best possible // counts[best] += b_dist; // sum the best possible } // Now find out the best position best = 0; for (i=0; i < a.num_frames(); i++) { if (counts[i] > counts[best]) best = i; } // Change frame number back to ms offset best_pos = frametoms(mstoframe(get_c_float(MID(d)),a.shift()) + best,a.shift()); delete counts; delete w; return best_pos; } static float frametoms(int frame,float frame_shift) { return (frame*frame_shift)*1000.0; } static int mstoframe(float ms,float frame_shift) { return (int)((ms/1000.0)/frame_shift); } // RJC - change for Track reorg. static float frame_distance(EST_Track &a, int fa, EST_Track &b, int fb, int size, double *weights) { float cost = 0.0,diff; int i; for (i=0; i < size; i++) { if (weights[i] != 0.0) { diff = (a(fa,i)-b(fb,i)); cost += diff*diff*weights[i]; } } return cost; } festival/src/modules/diphone/diphone.h0000644000175000017500000001555307262605650016504 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alistair Conkie */ /* Date : August 1996 */ /*-----------------------------------------------------------------------*/ #ifndef __DIPHONE_H_ #define __DIPHONE_H_ #include "EST_Pathname.h" enum di_sigaccess_t {di_direct, di_dynamic, di_ondemand}; enum di_db_type_t {di_pcm, di_lpc}; enum di_group_encode_t {di_raw, di_ulaw }; enum di_group_t {di_grouped, di_ungrouped}; typedef struct { char *diph; char *file; float beg; float mid; float end; } DI_INDEX; typedef struct { unsigned short nsamples; short *signal; } DI_VOX; typedef struct { unsigned short nmark; unsigned short lmark; unsigned short rmark; unsigned short *mark; } DI_PM; typedef struct { unsigned short nframes; float **f; } DI_LPC; typedef struct DD_STRUCT { char *name; char *type_str; enum di_db_type_t type; /* pcm or lpc */ enum di_group_t gtype; /* grouped or ungrouped */ char *index_file; char *signal_dir; char *signal_ext; char *signal_type; char *lpc_dir; char *lpc_ext; char *lpc_res_ext; char *lpc_res_type; float lpc_res_offset; int lpc_frame_offset; EST_String lpc_type; int lpc_order; float lpc_frame_shift; /* in seconds */ char *pitch_dir; char *pitch_ext; char *sig_access_type_str; enum di_sigaccess_t sig_access_type; char *group_encoding_str; enum di_group_encode_t group_encoding; LISP alternates_before; LISP alternates_after; int def_f0; FILE *xfd; FILE *gfd; /* the group file */ int gsignalbase; /* offset in group file where signals start */ int gframebase; /* offset in group file where frames start */ int samp_freq; int sig_band; /* the safety zone in samples round each signal */ char *phoneset; int ndiphs; int swap; char *default_diphone; /* when all else fails */ int lpc_pitch_synch; /* True if lpc frames are pitch synchronous */ short *allsignal; /* used in group files */ unsigned char *allulawsignal; float *allframes; short *allframesshort; int nindex; int zone; /* bytes */ int *offsets; /* used for indexing signals in group files */ int *frameoffsets; /* used for indexing frames in group files */ DI_INDEX **indx; DI_VOX **vox; DI_PM **pm; DI_LPC **lpc; } DIPHONE_DATABASE; typedef struct { int p_sz; int p_max; int p_tmp; int *pitch; int *cum_pitch; int *phon_ref; int *phon_mark; int *diph_ref; int *diph_mark; int *inwidth; int *outwidth; } DIPHONE_ACOUSTIC; typedef struct { int p_sz; int p_max; char **phons; int *duration; int *cum_dur; int *ref; int *pm_req; int *pm_giv; int *pm_chg_diph; /* phoneme-based info */ char **diphs; int t_sz; int t_max; int *pc_targs; int *targ_phon; int *targ_freq; int *abs_targ; /* maybe in samples */ } DIPHONE_SPN; typedef struct { int o_sz; int o_max; short *track; } DIPHONE_OUTPUT; void di_load_database(DIPHONE_DATABASE *database); void di_calc_pitch(DIPHONE_DATABASE *config, DIPHONE_SPN *ps, DIPHONE_ACOUSTIC *as); void di_psola_tm(DIPHONE_DATABASE *db, DIPHONE_ACOUSTIC *as, DIPHONE_OUTPUT *output); void di_reslpc(DIPHONE_DATABASE *db, DIPHONE_ACOUSTIC *as, DIPHONE_OUTPUT *output); void reslpc_resynth(const EST_String &file, DIPHONE_DATABASE *db, DIPHONE_ACOUSTIC *as, DIPHONE_OUTPUT *output); void di_frame_select(DIPHONE_DATABASE *db, DIPHONE_SPN *ps, DIPHONE_ACOUSTIC *as); short *di_get_diph_signal(int diph,DIPHONE_DATABASE *db); float *di_get_diph_lpc_mark(int diph,int mark,DIPHONE_DATABASE *db); //short *di_get_diph_res_mark(int diph,int mark,DIPHONE_DATABASE *db); short *di_get_diph_res_mark(int diph,int mark,int size,DIPHONE_DATABASE *db); void di_save_grouped_db(const EST_Pathname &filename, DIPHONE_DATABASE *db); void di_load_grouped_db(const EST_Pathname &filename, DIPHONE_DATABASE *db, LISP global_params); void di_fixed_parameters(DIPHONE_DATABASE *db,LISP params); void di_general_parameters(DIPHONE_DATABASE *db,LISP params); void load_pitch_file(DIPHONE_DATABASE *database, int i, int mode); LISP find_optimal_coupling(LISP table, LISP weights); void di_add_diphonedb(DIPHONE_DATABASE *db); DIPHONE_DATABASE *make_diphone_db(void); #endif /* __DIPHONE_H__ */ festival/src/modules/clunits/Makefile0000644000175000017500000001011511337416545016366 0ustar awbawb########################################################################### ## ## ## Carnegie Mellon University and ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1998-2010 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH, CARNEGIE MELLON UNIVERSITY AND THE ## ## CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH REGARD TO ## ## THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY ## ## AND FITNESS, IN NO EVENT SHALL THE UNIVERSITY OF EDINBURGH, CARNEGIE ## ## MELLON UNIVERSITY NOR THE CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, ## ## INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER ## ## RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION ## ## OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF ## ## OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Alan W Black ## ## Date: April 1998 ## ## with substantial tidyup Spring 2000 ## ########################################################################### ## ## ## Implementation of Black, A. and Taylor, P. (1997). Automatically ## ## clustering similar units for unit selection in speech synthesis ## ## Proceedings of Eurospeech 97, vol2 pp 601-604, Rhodes, Greece. ## ## ## ## postscript: http://www.cs.cmu.edu/~awb/papers/ES97units.ps ## ## http://www.cs.cmu.edu/~awb/papers/ES97units/ES97units.html ## ## ## ## This also serves as an example unit selection algorithm within the ## ## the new festival unit selection/signal processing architecture ## ## ## ########################################################################### TOP=../../.. DIRNAME=src/modules/clunits SCMS = acost.scm H = clunits.h TSRCS = SRCSCXX = acost.cc clunits.cc cldb.cc cljoin.cc SRCS = $(SRCSCXX) $(TSRCS) OBJS = $(SRCSCXX:.cc=.o) FILES=Makefile $(SRCS) $(H) $(SCMS) $(WORKINGON) LOCAL_INCLUDES = -I../include -I../UniSyn INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/clunits/acost.cc0000644000175000017500000003125711465010357016351 0ustar awbawb/*************************************************************************/ /* */ /* Carnegie Mellon University and */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998-2001 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH, CARNEGIE MELLON UNIVERSITY AND THE */ /* CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH REGARD TO */ /* THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY */ /* AND FITNESS, IN NO EVENT SHALL THE UNIVERSITY OF EDINBURGH, CARNEGIE */ /* MELLON UNIVERSITY NOR THE CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, */ /* INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER */ /* RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION */ /* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF */ /* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* Find the "acoustic" distance between two units. There are a number */ /* of various params to this but it falls downs to a weighted summed */ /* difference of parameters for the two units. */ /* */ /* A linear interpolation of the smallest to largest is done, and an */ /* optional penality is factored. */ /* */ /* This is independent of any particular Unit Database */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "clunits.h" static void find_unit_distances(LISP units, const EST_String &fname); static float duration_penalty_weight=1.0; static float f0_penalty_weight=0.0; static EST_FVector weights; static LISP get_stds_per_unit = NIL; static EST_String directory = ""; LISP acost_utt_load_coeffs(LISP utt, LISP params) { // Load all the coefficient file and create sub_track for each // segment EST_Utterance *u = get_c_utt(utt); EST_Track *track = new EST_Track; EST_String coefffilename = EST_String(get_param_str("db_dir",params,"./"))+ get_param_str("coeffs_dir",params,"coeffs/")+ u->f("fileid").string()+ get_param_str("coeffs_ext",params,".coeffs"); float ac_left_context = get_param_float("ac_left_context",params,0.0); EST_String segrelation = EST_String(get_param_str("clunit_relation",params,"Segment")); if (track->load(coefffilename) != format_ok) { cerr << "ACOST: failed to read track from \"" << coefffilename << "\"" << endl; festival_error(); } cl_maybe_fix_pitch_c0(track); // Add the whole track to a new relation EST_Item *c_si = u->create_relation("Acoustic_Coeffs")->append(); c_si->set_val("Acoustic_Coeffs", est_val(track)); // Now add subtracks for each segment for (EST_Item *s=u->relation(segrelation)->first(); s != 0; s=s->next()) { EST_Track *st = new EST_Track; float start = ffeature(s,"segment_start"); float end = ffeature(s,"segment_end"); if (s->prev()) start -= ac_left_context* ffeature(s,"p.segment_duration").Float(); int startf = track->index(start); int nframes = track->index(end)-startf; if (startf >= track->num_frames()) { cerr << "ACOST: utterances longer than coeffs file \n " << coefffilename << endl; festival_error(); } else if ((startf + nframes) > track->num_frames()) nframes = track->num_frames() - startf; track->sub_track(*st,startf,nframes,0); s->set_val("Acoustic_Coeffs",est_val(st)); } return utt; } LISP make_unit_distance_tables(LISP unittypes, LISP params) { // Build distance tables for given lists of lists of items // Each item should have a coeffs feature (a track). Also // the order the items are give is the order they will // be put into the saved distance matrix. Care should be // taken to ensure you save related features in the same order LISP ut; for (ut=unittypes; ut != NIL; ut = cdr(ut)) { acost_dt_params(params); EST_String unit_name = get_c_string(car(car(ut))); EST_String fname = EST_String(get_param_str("db_dir",params,"./"))+ get_param_str("disttabs_dir",params,"disttabs/")+ unit_name+".disttab"; cout << "Making unit distance table for " << unit_name << " (" << siod_llength(cdr(car(ut))) << ")" << endl; find_unit_distances(cdr(car(ut)),fname); } return NIL; } LISP ac_distance_tracks(LISP filename1, LISP filename2, LISP lweights) { // Return Unit type distance between two full files. EST_Track a,b; if (a.load(get_c_string(filename1)) != format_ok) { cerr << "CLUNITS: distance tracks: \"" << get_c_string(filename1) << "\" unloadable." << endl; festival_error(); } if (b.load(get_c_string(filename2)) != format_ok) { cerr << "CLUNITS: distance tracks: \"" << get_c_string(filename2) << "\" unloadable." << endl; festival_error(); } LISP l; int i; float dist; duration_penalty_weight = get_c_float(car(lweights)); EST_FVector tweights(siod_llength(cdr(lweights))); for (l=cdr(lweights),i=0; l!=NIL;l=cdr(l),i++) tweights[i] = get_c_float(car(l)); dist = ac_unit_distance(a,b,tweights); return flocons(dist); } void acost_dt_params(LISP params) { // Extract parameters from description LISP lweights,l; int i; directory = get_param_str("disttab_dir",params,"disttabs"); lweights = get_param_lisp("ac_weights",params,NIL); weights.resize(siod_llength(lweights)); for (l=lweights,i=0; l != NIL; l=cdr(l),i++) weights[i] = get_c_float(car(l)); duration_penalty_weight = get_param_float("dur_pen_weight",params,1.0); f0_penalty_weight = get_param_float("f0_pen_weight",params,0.0); get_stds_per_unit = get_param_lisp("get_stds_per_unit",params,NIL); } static void cumulate_ss_frames(EST_Track *a,EST_SuffStats *ss_frames) { // Gather sufficient statistics on the parameters in each frame int i,j; double p; for (i=0; i < a->num_frames(); i++) for (j=0; j < a->num_channels(); j++) { p = a->a_no_check(i,j); if (!finite(p)) { p = 1.0e5; a->a_no_check(i,j) = p; } ss_frames[j] += p; } } static EST_Track *acost_get_coefficients(EST_Item *si) { EST_Val c = si->f("Acoustic_Coeffs"); if (c == 0) { cerr << "ACOST: failed to find coefficients on items\n"; festival_error(); } return track(c); } static void find_unit_distances(LISP units, const EST_String &fname) { // Find all distances between units of type. int i,j; LISP u,v; EST_FMatrix dist(siod_llength(units),siod_llength(units)); EST_SuffStats *ss_frames = new EST_SuffStats[weights.length()]; // Find stddev for this unit for (i=0,u=units; u != 0; u=cdr(u),i++) { dist.a_no_check(0,i) = 0; if (get_stds_per_unit != NIL) // need to calculate stds for this unit { EST_Item *si = get_c_item(car(u)); EST_Track *a=acost_get_coefficients(si); if (a->num_channels() != weights.length()) { cerr << "ACOST: number of weights " << weights.length() << " does not match mcep param width " << a->num_channels() << endl; festival_error(); } cumulate_ss_frames(a,ss_frames); } } if (get_stds_per_unit != NIL) // modify weights with stds for (i=0; i < weights.length(); i++) weights[i] /= (ss_frames[i].stddev() * ss_frames[i].stddev()); for (i=1,u=cdr(units); u != 0; u=cdr(u),i++) { EST_Track *a=acost_get_coefficients(get_c_item(car(u))); // Only fill lower half of matrix for (v=units,j=0; j < i; j++,v=cdr(v)) { EST_Track *b=acost_get_coefficients(get_c_item(car(v))); dist.a_no_check(i,j) = ac_unit_distance(*a,*b,weights); } for ( ; j < dist.num_rows(); j++) dist.a_no_check(i,j) = 0.0; } delete [] ss_frames; if (dist.save(fname,"est_ascii") != write_ok) { cerr << "ACOST: failed to save distance data in \"" << fname << endl; festival_error(); } } float ac_unit_distance(const EST_Track &unit1, const EST_Track &unit2, const EST_FVector wghts) { // Find distance between two units, unit1 will be smaller than unit2 float distance = 0.0; int i,j,k; float fj,incr,dur_penalty; float cost,diff; float score; int nc = unit1.num_channels(); if (unit1.end() > unit2.end()) return ac_unit_distance(unit2,unit1,wghts); // unit1 is smaller if (unit1.num_frames() == 0) // return 1.0e20; // HUGE_VAL is too HUGE return 100; if ((unit1.num_channels() != unit2.num_channels()) || (unit1.num_channels() != wghts.length())) { cerr << "ac_unit_distance: unit1 (" << unit1.num_channels() << "), unit2 (" << unit2.num_channels() << ") and wghts (" << wghts.length() << ") are of different size" << endl; festival_error(); } // printf("unit1 nf %d end %f unit2 nf %d end %f \n", // unit1.num_frames(), // unit1.end(), // unit2.num_frames(), // unit2.end()); incr = unit1.end()/ unit2.end(); for (fj=0.0,j=i=0; i < unit2.num_frames(); i++,fj+=incr) { while ((j < unit1.num_frames()-1) && (unit1.t(j) < unit2.t(i)*incr) ) j++; // printf("unit1 j %d %f unit2 %d %f %f\n", // j,unit1.t(j), // i,unit2.t(i), // unit2.t(i)*incr); cost = f0_penalty_weight * fabs((unit1.t(j)-((j > 0) ? unit1.t(j-1) : 0))- (unit2.t(i)-((i > 0) ? unit2.t(i-1) : 0))); for (k=0; k < nc; k++) if (wghts.a_no_check(k) != 0.0) { diff = unit2.a_no_check(i,k)-unit1.a_no_check(j,k); diff *= diff; cost += diff*wghts.a_no_check(k); } distance += cost; } dur_penalty = (float)unit2.end()/(float)unit1.end(); score = (distance/(float)i)+(dur_penalty*duration_penalty_weight); return score; } // Maybe can reduce this to an FVector pulled out of the track float frame_distance(const EST_Track &a, int ai, const EST_Track &b, int bi, const EST_FVector &wghts, float f0_weight) { float cost = 0.0,diff; if ((a.num_channels() != b.num_channels()) || (a.num_channels() != wghts.length())) { cerr << "frame_distance: unit1, unit2 and wghts" << " are of different size" << endl; festival_error(); } if ((ai < 0) || (ai >= a.num_frames()) || (bi < 0) || (bi >= b.num_frames())) { cerr << "frame_distance: frames out of range" << endl; festival_error(); } if (f0_weight > 0) { cost = f0_weight * fabs((a.t(ai)-((ai > 0) ? a.t(ai-1) : 0))- (b.t(bi)-((bi > 0) ? b.t(bi-1) : 0))); } for (int k=0; k < a.num_channels(); k++) { if (wghts.a_no_check(k) != 0.0) { diff = a.a_no_check(ai,k)-b.a_no_check(bi,k); diff *= wghts.a_no_check(k); cost += diff*diff; } } return sqrt(cost); } festival/src/modules/clunits/clunits.cc0000644000175000017500000006406111465010357016720 0ustar awbawb/*************************************************************************/ /* */ /* Carnegie Mellon University and */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998-2001 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH, CARNEGIE MELLON UNIVERSITY AND THE */ /* CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH REGARD TO */ /* THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY */ /* AND FITNESS, IN NO EVENT SHALL THE UNIVERSITY OF EDINBURGH, CARNEGIE */ /* MELLON UNIVERSITY NOR THE CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, */ /* INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER */ /* RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION */ /* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF */ /* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* Yet another unit selection method. */ /* */ /* Using an acoustic measure find the distance between all units in the */ /* db. Try to minimise the mean difference between units in a cluster */ /* using CART technology, based on features like phonetic and prosodic */ /* context. This gives a bunch of CARTs for each unit type in the db */ /* which are acoustically close. Use these as candidates and optimise */ /* a path through them minimising join using a viterbi search. */ /* */ /* Advantages: */ /* requires little or no measurements at selection time */ /* allows for clear method of pruning */ /* no weights need to be generated (well, except where they do) */ /* will optimise appropriately with varying numbers of example units */ /* */ /* Disadvantages: */ /* Units can't cross between clusters */ /* */ /* Implementation of Black, A. and Taylor, P. (1997). Automatically */ /* clustering similar units for unit selection in speech synthesis */ /* Proceedings of Eurospeech 97, vol2 pp 601-604, Rhodes, Greece. */ /* */ /* postscript: http://www.cs.cmu.edu/~awb/papers/ES97units.ps */ /* http://www.cs.cmu.edu/~awb/papers/ES97units/ES97units.html */ /* */ /* Comments: */ /* */ /* This is a new implementation using the newer unit selection/signal */ /* processing archtecture in festival */ /* */ /* This is still in development but become more stable. It is robust */ /* for many cases, though a lot depends on the db and parameters */ /* you use */ /* */ /* This had significant new work (and bug fixes) done on it when awb */ /* moved to CMU */ /* */ /*=======================================================================*/ #include #include "EST_math.h" #include "festival.h" #include "clunits.h" static EST_String static_unit_prev_move = "unit_prev_move"; static EST_String static_unit_this_move = "unit_this_move"; static EST_String static_jscore = "local_join_cost"; static EST_String static_tscore = "local_target_cost"; static EST_String static_cscore = "cummulative_unit_score"; static void setup_clunits_params(); static EST_VTCandidate *TS_candlist(EST_Item *s,EST_Features &f); static EST_VTPath *TS_npath(EST_VTPath *p,EST_VTCandidate *c,EST_Features &f); static float naive_join_cost(CLunit *unit0, CLunit *unit1, EST_Item *s, float &u0_move, float &u1_move); static float optimal_couple(CLunit *u0, CLunit *u1, float &u0_move, float &u1_move, int type, float different_prev_pen, float non_consecutive_pen); static void cl_parse_diphone_times(EST_Relation &diphone_stream, EST_Relation &source_lab); VAL_REGISTER_CLASS_NODEL(vtcand,EST_VTCandidate); VAL_REGISTER_CLASS_NODEL(clunit,CLunit); LISP selection_trees = NIL; LISP clunits_params = NIL; static int optimal_coupling = 0; static int extend_selections = 0; static int clunits_debug = 0; static int clunits_log_scores = 0; static int clunits_smooth_frames = 0; float continuity_weight = 1; float f0_join_weight = 0.0; float different_prev_pen = 1000.0; float non_consecutive_pen = 100.0; static EST_String clunit_name_feat = "name"; static CLDB *cldb; static LISP clunits_select(LISP utt) { // Select units from db using CARTs to index into clustered unit groups EST_Utterance *u = get_c_utt(utt); EST_Item *s, *f; cldb = check_cldb(); // make sure there is one loaded setup_clunits_params(); f = u->relation("Segment")->head(); for (s=f; s; s=s->next()) s->set_val("clunit_name",ffeature(s,clunit_name_feat)); if (f) { EST_Viterbi_Decoder v(TS_candlist,TS_npath,-1); v.set_big_is_good(FALSE); // big is bad v.initialise(u->relation("Segment")); v.search(); if (!v.result("unit_id")) { cerr << "CLUNIT: failed to find path\n"; return utt; } v.copy_feature(static_unit_this_move); v.copy_feature(static_unit_prev_move); v.copy_feature(static_jscore); v.copy_feature(static_tscore); v.copy_feature(static_cscore); } return utt; } static LISP clunits_get_units(LISP utt) { // Create unit stream and loading params EST_Utterance *u = get_c_utt(utt); EST_Relation *units,*ss; EST_Item *s; cldb = check_cldb(); // make sure there is one loaded units = u->create_relation("Unit"); for (s=u->relation("Segment")->head(); s != 0; s=s->next()) { EST_Item *unit = units->append(); CLunit *db_unit = clunit(s->f("unit_id")); float st,e; unit->set_name(db_unit->name); unit->set("fileid",db_unit->fileid); // These should be modified from the optimal coupling if ((s->prev()) && (s->f_present("unit_this_move"))) st = s->F("unit_this_move"); else st = db_unit->start; if (s->next() && (s->next()->f_present("unit_prev_move"))) e = s->next()->F("unit_prev_move"); else e = db_unit->end; if ((e-st) < 0.011) e = st + 0.011; unit->set("start",st); unit->set("middle",db_unit->start); unit->set("end",e); unit->set("unit_start",st); unit->set("unit_middle",db_unit->start); unit->set("unit_end",e); unit->set("seg_start",db_unit->start); unit->set("seg_end",db_unit->end); cldb->load_coefs_sig(unit); if (clunits_debug) printf("unit: %s fileid %s start %f end %f\n", (const char *)db_unit->name, (const char *)db_unit->fileid, st,e); } // Make it look as much like the diphones as possible for // the rest of the code ss = u->create_relation("SourceSegments"); for (s = u->relation("Segment")->head(); s != 0 ; s = s->next()) { EST_Item *d = ss->append(); d->set_name(ffeature(s,"clunit_name")); } cl_parse_diphone_times(*units,*ss); return utt; } static void cl_parse_diphone_times(EST_Relation &diphone_stream, EST_Relation &source_lab) { EST_Item *s, *u; EST_Track *pm; int e_frame, m_frame = 0; float dur_1 = 0.0, dur_2 = 0.0, p_time; float t_time = 0.0, end; p_time = 0.0; for (s = source_lab.head(), u = diphone_stream.head(); u; u = u->next(), s = s->next()) { pm = track(u->f("coefs")); if (pm == 0) { cerr << "CLUNIT: couldn't get pitchmarks for " << u->name() << endl; festival_error(); } e_frame = pm->num_frames() - 1; m_frame = u->I("middle_frame"); dur_1 = pm->t(m_frame); dur_2 = pm->t(e_frame) - dur_1; s->set("end", (dur_1 + p_time)); p_time = s->F("end") + dur_2; end = dur_1 + dur_2 + t_time; t_time = end; u->set("end", t_time); } if (s) s->set("end", (dur_2 + p_time)); } static LISP clunits_simple_wave(LISP utt) { // Naive joining of waveforms EST_Utterance *u = get_c_utt(utt); EST_Wave *w = new EST_Wave; EST_Wave *w1 = 0; EST_Item *witem = 0; EST_Item *s; int size,i,k,c; for (size=0,s=u->relation("Unit")->head(); s != 0; s = s->next()) size += wave(s->f("sig"))->num_samples(); if (u->relation("Unit")->head()) { // This will copy the necessary wave features across s = u->relation("Unit")->head(); *w = *(wave(s->f("sig"))); } i = w->num_samples(); w->resize(size); // its maximum size for (s=u->relation("Unit")->head()->next(); s; s=s->next()) { w1 = wave(s->f("sig")); // Find last zero crossing for (c=0; ((i > 0) && (c < 40)); c++,i--) if (((w->a_no_check(i) < 0) && (w->a_no_check(i-1) >= 0)) || ((w->a_no_check(i) >= 0) && (w->a_no_check(i-1) < 0))) break; if (c == 40) i += 40; // Find next zero crossing for (c=0,k=1; ((k < w1->num_samples()) && (c < 40)); k++,i++) if (((w1->a_no_check(k) < 0) && (w1->a_no_check(k-1) >= 0)) || ((w1->a_no_check(k) >= 0) && (w1->a_no_check(k-1) < 0))) break; if (c == 40) k -= 40; for (; k < w1->num_samples(); k++,i++) w->a_no_check(i) = w1->a_no_check(k); } w->resize(i); witem = u->create_relation("Wave")->append(); witem->set_val("wave",est_val(w)); return utt; } static LISP clunits_windowed_wave(LISP utt) { // windowed join, no prosodic modification EST_Utterance *u = get_c_utt(utt); EST_Wave *w = new EST_Wave; EST_Wave *w1 = 0; EST_Track *t1 = 0; EST_Item *witem = 0; EST_Item *s; int size,i,k,wi,samp_idx, l_samp_idx; int width, lwidth; EST_Wave *www=0; for (size=0,s=u->relation("Unit")->head(); s != 0; s = s->next()) size += wave(s->f("sig"))->num_samples(); if (u->relation("Unit")->head()) { // This will copy the necessary wave features across s = u->relation("Unit")->head(); www = wave(s->f("sig")); *w = *www; } w->resize(size); // its maximum size wi=0; lwidth = width = 0; for (s=u->relation("Unit")->head(); s; s=s->next()) { w1 = wave(s->f("sig")); t1 = track(s->f("coefs")); l_samp_idx = 0; for (i=0; i < t1->num_frames()-1; i++) { samp_idx = (int)(t1->t(i)*w->sample_rate()); width = samp_idx - l_samp_idx; if (clunits_smooth_frames && (i==0) && (lwidth != 0)) width = (width+lwidth)/2; // not sure if this is worth it wi += width; for (k=-width; ((knum_samples())) ;k++) w->a(wi+k) += (int)(0.5*(1+cos((PI/(double)(width))*(double)k))* w1->a(samp_idx+k)); l_samp_idx = samp_idx; } lwidth = width; } w->resize(wi); witem = u->create_relation("Wave")->append(); witem->set_val("wave",est_val(w)); return utt; } static LISP clunits_smoothedjoin_wave(LISP utt) { // Actually not very smoothed yet, just joined EST_Utterance *u = get_c_utt(utt); EST_Wave *w = new EST_Wave; EST_Wave *w1 = 0; EST_Track *t1 = 0; EST_Item *witem = 0; EST_Item *s; int size,i,wi; int samp_end, samp_start; EST_Wave *www=0; for (size=0,s=u->relation("Unit")->head(); s != 0; s = s->next()) { samp_end = s->I("samp_end"); samp_start = s->I("samp_start"); size += samp_end-samp_start; } if (u->relation("Unit")->head()) { // This will copy the necessary wave features across s = u->relation("Unit")->head(); www = wave(s->f("sig")); *w = *www; } w->resize(size); // its maximum size wi=0; for (s=u->relation("Unit")->head(); s; s=s->next()) { samp_end = s->I("samp_end"); samp_start = s->I("samp_start"); w1 = wave(s->f("sig")); /* printf("%s %s %f %f %d %d\n", (const char *)s->S("name"), (const char *)s->S("fileid"), (float)samp_start/(float)w->sample_rate(), (float)samp_end/(float)w->sample_rate(), w1->num_samples(), samp_end); */ t1 = track(s->f("coefs")); for (i=samp_start; ia_no_check(wi) = w1->a_no_check(i); /* printf("%d %f\n",wi,(float)wi/(float)w->sample_rate()); */ } w->resize(wi); witem = u->create_relation("Wave")->append(); witem->set_val("wave",est_val(w)); return utt; } static void setup_clunits_params() { // Set up params clunits_params = siod_get_lval("clunits_params", "CLUNITS: no parameters set for module"); optimal_coupling = get_param_int("optimal_coupling",clunits_params,0); different_prev_pen = get_param_float("different_prev_pen",clunits_params,1000.0); non_consecutive_pen = get_param_float("non_consectutive_pen",clunits_params,100.0); extend_selections = get_param_int("extend_selections",clunits_params,0); continuity_weight = get_param_float("continuity_weight",clunits_params,1); f0_join_weight = get_param_float("f0_join_weight",clunits_params,0.0); clunits_debug = get_param_int("clunits_debug",clunits_params,0); clunits_log_scores = get_param_int("log_scores",clunits_params,0); clunits_smooth_frames = get_param_int("smooth_frames",clunits_params,0); clunit_name_feat = get_param_str("clunit_name_feat",clunits_params,"name"); selection_trees = siod_get_lval("clunits_selection_trees", "CLUNITS: clunits_selection_trees unbound"); } static EST_VTCandidate *TS_candlist(EST_Item *s,EST_Features &f) { // Return a list of candidate units for target s // Use the appropriate CART to select a small group of candidates EST_VTCandidate *all_cands = 0; EST_VTCandidate *c, *gt; LISP tree,group,l,pd,cc,ls; EST_String name; EST_String lookingfor; CLunit *u; int bbb,ccc; float cluster_mean; (void)f; bbb=ccc=0; lookingfor = s->S("clunit_name"); ls = siod(s); cc = siod_get_lval("clunits_cand_hooks",NULL); if (cc) pd = apply_hooks(siod_get_lval("clunits_cand_hooks",NULL), ls); else { tree = car(cdr(siod_assoc_str(lookingfor,selection_trees))); pd = wagon_pd(s,tree); } if (pd == NIL) { cerr << "CLUNITS: no predicted class for " << s->S("clunit_name") << endl; festival_error(); } group = car(pd); cluster_mean = get_c_float(car(cdr(pd))); for (bbb=0,l=group; l != NIL; l=cdr(l),bbb++) { c = new EST_VTCandidate; name = s->S("clunit_name")+"_"+get_c_string(car(car(l))); u = cldb->get_unit(name); if (u == 0) { cerr << "CLUNITS: failed to find unit " << name << " in index" << endl; festival_error(); } cldb->load_join_coefs(u); c->name = est_val(u); c->s = s; // Mean distance from others in cluster (could be precalculated) c->score = get_c_float(car(cdr(car(l))))-cluster_mean; c->score *= c->score; // Maybe this should be divided by overall mean of set // to normalise this figure (?) c->next = all_cands; all_cands = c; } if (extend_selections) { // An experiment, for all candidates of the previous // item whose following is of this phone type, include // them as a candidate EST_Item *ppp = s->prev(); if (ppp) { EST_VTCandidate *lc = vtcand(ppp->f("unit_cands")); for (ccc=0 ; lc && (ccc < extend_selections); lc = lc->next) { CLunit *unit = clunit(lc->name); CLunit *next_unit; if (unit->next_unit) next_unit = unit->next_unit; else continue; EST_String ss; ss = next_unit->name.before("_"); if (ss.matches(".*_.*_.*")) { ss += "_"; ss += next_unit->name.after("_").before("_"); } /* printf("%s %s\n",(const char *)ss, (const char *)lookingfor); */ for (gt=all_cands; gt; gt=gt->next) if (clunit(gt->name)->name == next_unit->name) break; /* got this one already */ if ((ss == lookingfor) && (gt == 0)) { // its the right type so add it c = new EST_VTCandidate; c->name = est_val(next_unit); cldb->load_join_coefs(next_unit); c->s = s; c->score = 0; c->next = all_cands; all_cands = c; bbb++; ccc++; } } } s->set_val("unit_cands",est_val(all_cands)); } if (clunits_debug) printf("cands %d (extends %d) %s\n",bbb,ccc,(const char *)lookingfor); return all_cands; } static EST_VTPath *TS_npath(EST_VTPath *p,EST_VTCandidate *c,EST_Features &f) { // Combine candidate c with previous path updating score // with join cost float cost; EST_VTPath *np = new EST_VTPath; CLunit *u0, *u1; float u0_move=0.0, u1_move=0.0; (void)f; np->c = c; np->from = p; if ((p == 0) || (p->c == 0)) cost = 0; // nothing previous to join to else { u0 = clunit(p->c->name); u1 = clunit(c->name); // printf("u0 %s u1 %s\n", // (const char *)u0->name, // (const char *)u1->name); if (optimal_coupling) cost = optimal_couple(u0,u1,u0_move,u1_move, optimal_coupling, different_prev_pen, non_consecutive_pen); else // naive measure cost = naive_join_cost(u0,u1,c->s,u0_move,u1_move); // When optimal_coupling == 2 the moves will be 0, just the scores // are relevant if (optimal_coupling == 1) { np->f.set(static_unit_prev_move,u0_move); // new (prev) end np->f.set(static_unit_this_move,u1_move); // new start } } // printf("cost %f continuity_weight %f\n", cost, continuity_weight); cost *= continuity_weight; np->state = c->pos; // "state" is candidate number if (clunits_log_scores && (cost != 0)) cost = log(cost); np->f.set(static_jscore,cost); np->f.set(static_tscore,c->score); if (p==0) np->score = (c->score+cost); else np->score = (c->score+cost) + p->score; np->f.set(static_cscore,np->score); if (clunits_debug > 1) printf("joining cost %f\n",np->score); return np; } static float optimal_couple(CLunit *u0, CLunit *u1, float &u0_move, float &u1_move, int type, float different_prev_pen, float non_consecutive_pen ) { // Find combination cost of u0 to u1, checking for best // frame up to n frames back in u0 and u1. // Note this checks the u0 with u1's predecessor, which may or may not // be of the same type // There is some optimisation here in unit coeff access EST_Track *u0_cep, *u1_p_cep; float dist, best_val; int i,eee; int u0_st, u0_end; int u1_p_st, u1_p_end; int best_u0, best_u1; CLunit *u1_p; float f; u1_p = u1->prev_unit; u0_move = u0->end; if (u1_p == 0) u1_move = 0; else u1_move = u1_p->end; if (u1_p == u0) // they are consecutive return 0.0; if (u1_p == 0) // hacky condition, when there is no previous we'll return 0.0; // assume a good join (should be silence there) if (u1_p->join_coeffs == 0) cldb->load_join_coefs(u1_p); // Get indexes into full cep for utterances rather than sub ceps u0_cep = u0->join_coeffs; u1_p_cep = u1_p->join_coeffs; u0_end = u0_cep->num_frames(); u1_p_end = u1_p_cep->num_frames(); if (!streq(u1_p->base_name,u0->base_name)) { /* prev(u1) is a different phone from u0 so don't slide */ f = different_prev_pen; u0_st = u0_cep->num_frames()-1; u1_p_st = u1_p_cep->num_frames()-1; } else if (type == 2) { /* we'll only check the edge for the join */ u0_st = u0_cep->num_frames()-1; u1_p_st = u1_p_cep->num_frames()-1; f = 1; } else { u0_st = (int)(u0_cep->num_frames() * 0.33); u1_p_st = (int)(u1_p_cep->num_frames() * 0.33); f = 1; } best_u0=u0_end; best_u1=u1_p_end; best_val = HUGE_VAL; // Here we look for the best join without sliding the windows if ((u0_end-u0_st) < (u1_p_end-u1_p_st)) eee = u0_end-u0_st; else eee = u1_p_end-u1_p_st; for (i=0; i < eee; i++) { dist = frame_distance(*u0_cep,i+u0_st, *u1_p_cep,i+u1_p_st, cldb->cweights, f0_join_weight); if (dist < best_val) { best_val = dist; best_u0 = i+u0_st; best_u1 = i+u1_p_st; } } #if 0 // This tries *all* possible matches in the pair, its slow // and has a tendency to shorten things more than you'd like // so we just use the more simple test above. int j; for (i=u0_st; i < u0_end; i++) { for (j=u1_p_st; j < u1_p_end; j++) { dist = frame_distance(*u0_cep,i, *u1_p_cep,j, cldb->cweights); if (dist < best_val) { best_val = dist; best_u0 = i; best_u1 = j; } } } #endif if (type == 1) { u0_move = u0_cep->t(best_u0); u1_move = u1_p_cep->t(best_u1); } return non_consecutive_pen+(best_val*f); } static float naive_join_cost(CLunit *unit0, CLunit *unit1, EST_Item *s, float &u0_move, float &u1_move) { // A naive join cost, because I haven't ported the info yet u0_move = unit0->end; u1_move = unit1->start; if (unit0 == unit1) return 0; else if (unit1->prev_unit->name == unit0->name) return 0; else if (ph_is_silence(s->name())) return 0; else if (ph_is_stop(s->name())) return 0.2; else if (ph_is_fricative(s->name())) return 0.3; else return 1.0; } static LISP cldb_load_all_coeffs(LISP filelist) { LISP f; cldb = check_cldb(); for (f=filelist; f; f=cdr(f)) { cldb->get_file_coefs_sig(get_c_string(car(f))); cldb->get_file_join_coefs(get_c_string(car(f))); } return NIL; } void festival_clunits_init(void) { // Initialization for clunits selection proclaim_module("clunits", "Copyright (C) University of Edinburgh and CMU 1997-2010\n"); gc_protect(&clunits_params); gc_protect(&selection_trees); festival_def_utt_module("Clunits_Select",clunits_select, "(Clunits_Select UTT)\n\ Select units from current databases using cluster selection method."); festival_def_utt_module("Clunits_Get_Units",clunits_get_units, "(Clunits_Get_Units UTT)\n\ Construct Unit relation from the selected units in Segment and extract\n\ their parameters from the clunit db."); festival_def_utt_module("Clunits_Simple_Wave",clunits_simple_wave, "(Clunits_Simple_Wave UTT)\n\ Naively concatenate signals together into a single wave (for debugging)."); festival_def_utt_module("Clunits_Windowed_Wave",clunits_windowed_wave, "(Clunits_Windowed_Wave UTT)\n\ Use hamming window over edges of units to join them, no prosodic \n\ modification though."); festival_def_utt_module("Clunits_SmoothedJoin_Wave",clunits_smoothedjoin_wave, "(Clunits_SmoothedJoin_Wave UTT)\n\ smoothed join."); init_subr_1("clunits:load_db",cl_load_db, "(clunits:load_db PARAMS)\n\ Load index file for cluster database and set up params, and select it."); init_subr_1("clunits:select",cldb_select, "(clunits:select NAME)\n\ Select a previously loaded cluster database."); init_subr_1("clunits:load_all_coefs",cldb_load_all_coeffs, "(clunits:load_all_coefs FILEIDLIST)\n\ Load in coefficients, signal and join coefficients for each named\n\ fileid. This is can be called at startup to to reduce the load time\n\ during synthesis (though may make the image large)."); init_subr_0("clunits:list",cldb_list, "(clunits:list)\n\ List names of currently loaded cluster databases."); init_subr_2("acost:build_disttabs",make_unit_distance_tables, "(acost:build_disttabs UTTTYPES PARAMS)\n\ Built matrices of distances between each ling_item in each each list\n\ of ling_items in uttypes. Uses acoustic weights in PARAMS and save\n\ the result as a matrix for later use."); init_subr_2("acost:utt.load_coeffs",acost_utt_load_coeffs, "(acost:utt.load_coeffs UTT PARAMS)\n\ Load in the acoustic coefficients into UTT and set the Acoustic_Coeffs\n\ feature for each segment in UTT."); init_subr_3("acost:file_difference",ac_distance_tracks, "(acost:file_difference FILENAME1 FILENAME2 PARAMS)\n\ Load in the two named tracks and find the acoustic difference over all\n\ based on the weights in PARAMS."); init_subr_2("cl_mapping", l_cl_mapping, "(cl_mapping UTT PARAMS)\n\ Impose prosody upto some percentage, and not absolutely."); } festival/src/modules/clunits/cldb.cc0000644000175000017500000002745011332061167016142 0ustar awbawb/*************************************************************************/ /* */ /* Carnegie Mellon University and */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998-2001 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH, CARNEGIE MELLON UNIVERSITY AND THE */ /* CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH REGARD TO */ /* THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY */ /* AND FITNESS, IN NO EVENT SHALL THE UNIVERSITY OF EDINBURGH, CARNEGIE */ /* MELLON UNIVERSITY NOR THE CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, */ /* INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER */ /* RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION */ /* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF */ /* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : April 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* A quick database structure */ /* */ /*=======================================================================*/ #include #include #include "festival.h" #include "EST_FileType.h" #include "clunits.h" VAL_REGISTER_CLASS(clunitsdb,CLDB) SIOD_REGISTER_CLASS(clunitsdb,CLDB) static void cl_load_catalogue(CLDB *cldb,EST_String &indexfile); static LISP CLDB_list = NIL; static CLDB *current_cldb = 0; static void cldb_add(const EST_String &name, CLDB *cldb) { // Add lexicon to list of lexicons LISP lpair; lpair = siod_assoc_str(name,CLDB_list); if (CLDB_list == NIL) gc_protect(&CLDB_list); if (lpair == NIL) { CLDB_list = cons(cons(strintern(name), cons(siod(cldb),NIL)), CLDB_list); } else { cwarn << "CLDB " << name << " recreated" << endl; // old one will be garbage collected setcar(cdr(lpair),siod(cldb)); } return; } LISP cl_load_db(LISP params) { EST_String indexfile; int i; LISP w; CLDB *cldb = new CLDB; cldb->params = params; indexfile = EST_String("") + get_param_str("db_dir",params,"./")+ get_param_str("catalogue_dir",params,"./")+ get_param_str("index_name",params,"catalogue")+ ".catalogue"; cl_load_catalogue(cldb,indexfile); cldb->cweights.resize(siod_llength(get_param_lisp("join_weights",params,NIL))); for (i=0,w=get_param_lisp("join_weights",params,NIL); w; w=cdr(w),i++) cldb->cweights[i] = get_c_float(car(w)); cldb_add(get_param_str("index_name",params,"catalogue"),cldb); current_cldb = cldb; return NIL; } static void cl_load_catalogue(CLDB *cldb,EST_String &indexfile) { EST_TokenStream ts; EST_EstFileType t; EST_Option hinfo; EST_String v; bool ascii; EST_read_status r; if (((indexfile == "-") ? ts.open(cin) : ts.open(indexfile)) != 0) { cerr << "CLUNITS: Can't open catalogue file " << indexfile << endl; festival_error(); } if (((r = read_est_header(ts, hinfo, ascii, t)) != format_ok) || (t != est_file_index)) { cerr << "CLUNITS: " << indexfile << " is not an indexfile" << endl; festival_error(); } CLunit *ls = 0; while(!ts.eof()) { CLunit *s = new CLunit; s->name = ts.get().string(); s->base_name = s->name.before("_"); s->fileid = ts.get().string(); s->start = atof(ts.get().string()); s->mid = atof(ts.get().string()); s->end = atof(ts.get().string()); if ((ls != 0) && (ls->fileid == s->fileid) && (ls->end == s->start)) { s->prev_unit = ls; ls->next_unit = s; } cldb->index.add(s->name,s); ls = s; } } CLDB *check_cldb() { if (current_cldb == 0) { cerr << "CLDB: no database loaded\n"; festival_error(); } return current_cldb; } void cl_maybe_fix_pitch_c0(EST_Track *c) { // If its pitch synchronous, trash the first coefficient with // the pitch value, there should be a cleaner way to do this int i; float ltime = 0; if (!c->equal_space()) { for (i=0; i < c->num_frames(); i++) { c->a_no_check(i,0) = 1/(c->t(i)-ltime); ltime = c->t(i); } } } void CLDB::load_join_coefs(CLunit *unit) { // Load in the coefficients and signal for this unit. CLfile *fileitem; EST_Track *join_coeffs; if (unit->join_coeffs != 0) return; fileitem = get_file_join_coefs(unit->fileid); EST_Track *unit_join_coeffs = new EST_Track; join_coeffs = fileitem->join_coeffs; int pm_start = join_coeffs->index(unit->start); int pm_end = join_coeffs->index(unit->end); join_coeffs->sub_track(*unit_join_coeffs, pm_start, pm_end-pm_start+1,0); unit->join_coeffs = unit_join_coeffs; } CLfile *CLDB::get_file_join_coefs(const EST_String &fileid) { CLfile *fileitem; fileitem = get_fileitem(fileid); if (fileitem == 0) { // even the file isn't here fileitem = new CLfile; fileindex.add(fileid,fileitem); } if (fileitem->join_coeffs == 0) { EST_Track *join_coeffs = new EST_Track; EST_String jc_filename = EST_String("") + get_param_str("db_dir",params,"./") + get_param_str("coeffs_dir",params,"wav/") + fileid+ get_param_str("coeffs_ext",params,".dcoeffs"); if (join_coeffs->load(jc_filename) != format_ok) { delete join_coeffs; cerr << "CLUNITS: failed to load join coeffs file " << jc_filename << endl; festival_error(); } // cl_maybe_fix_pitch_c0(join_coeffs); fileitem->join_coeffs = join_coeffs; } return fileitem; } CLfile *CLDB::get_file_coefs_sig(const EST_String &fileid) { CLfile *fileitem = get_fileitem(fileid); if (fileitem == 0) { // even the file isn't here fileitem = new CLfile; fileindex.add(fileid,fileitem); } if (fileitem->sig == 0) { EST_Track *track = new EST_Track; EST_String coef_filename = EST_String("") + get_param_str("db_dir",params,"./") + get_param_str("pm_coeffs_dir",params,"pm/") + fileid+ get_param_str("pm_coeffs_ext",params,".pm"); if (track->load(coef_filename) != format_ok) { delete track; cerr << "CLUNITS: failed to load coeffs file " << coef_filename << endl; festival_error(); } fileitem->coefs = track; EST_Wave *sig = new EST_Wave; EST_String sig_filename = EST_String("") + get_param_str("db_dir",params,"./") + get_param_str("sig_dir",params,"wav/") + fileid+ get_param_str("sig_ext",params,".wav"); if (sig->load(sig_filename) != format_ok) { delete sig; cerr << "CLUNITS: failed to load signal file " << sig_filename << endl; festival_error(); } fileitem->sig = sig; } return fileitem; } void CLDB::load_coefs_sig(EST_Item *unit) { // Load in the coefficients and signal for this unit. EST_String fileid = unit->f("fileid"); CLfile *fileitem; fileitem = get_file_coefs_sig(fileid); EST_Track *coeffs = fileitem->coefs; EST_Wave *sig = fileitem->sig; EST_Track u1; EST_Wave *unit_sig = new EST_Wave; int pm_start = coeffs->index(unit->F("start")); int pm_middle = coeffs->index(unit->F("middle")); int pm_end = coeffs->index(unit->F("end")); // coeffs->sub_track(u1,Gof((pm_start-1),0), pm_end - pm_start + 1); coeffs->sub_track(u1,pm_start, pm_end - pm_start + 1,0); EST_Track *unit_coeffs = new EST_Track(u1); for (int j = 0; j < u1.num_frames(); ++j) unit_coeffs->t(j) = u1.t(j) - coeffs->t(Gof((pm_start - 1), 0)); /* printf("coefs %s: pm_start %d pm_end %d pm_length %d\n", (const char *)fileid, pm_start, pm_end, pm_end - pm_start + 1); */ unit->set_val("coefs",est_val(unit_coeffs)); if ((pm_middle-pm_start-1) < 1) unit->set("middle_frame", 1); else unit->set("middle_frame", pm_middle - pm_start -1); int samp_start = (int)(coeffs->t(Gof((pm_start - 1), 0)) * (float)sig->sample_rate()); int samp_end; if ((pm_end + 1) < coeffs->num_frames()) samp_end = (int)(coeffs->t(pm_end + 1) * (float)sig->sample_rate()); else samp_end = (int)(coeffs->t(pm_end) * (float)sig->sample_rate()); int real_samp_start = (int)(unit->F("start") * (float)sig->sample_rate()); int real_samp_end = (int)(unit->F("end") * (float)sig->sample_rate()); if (samp_end-samp_start < 1) sig->sub_wave(*unit_sig,samp_start, 1); else sig->sub_wave(*unit_sig,samp_start, samp_end-samp_start); if (real_samp_start-samp_start<0) unit->set("samp_start",0); else unit->set("samp_start",real_samp_start-samp_start); unit->set("samp_end",real_samp_end-samp_start); /* Need to preserve where the phone boundary is (which may actually */ /* be past the end of this unit */ unit->set("samp_seg_start", (int)(unit->F("seg_start") * (float)sig->sample_rate())-samp_start); unit->set_val("sig",est_val(unit_sig)); } CLunit::CLunit() { start=0; mid=0; end=0; prev_unit = 0; next_unit = 0; samp_start = 0; samp_end = 0; join_coeffs = 0; coefs = 0; sig = 0; } CLunit::~CLunit() { delete join_coeffs; delete coefs; delete sig; } CLfile::CLfile() { join_coeffs = 0; coefs = 0; sig = 0; } CLfile::~CLfile() { delete join_coeffs; delete coefs; delete sig; } CLDB::CLDB() { gc_protect(¶ms); } static void del_clunit(void *s) { delete (CLunit *)s; } static void del_clfile(void *s) { delete (CLfile *)s; } CLDB::~CLDB() { index.clear(del_clunit); fileindex.clear(del_clfile); gc_unprotect(¶ms); } LISP cldb_list(void) { // List names of all current defined cluster dbs LISP d = NIL; LISP l; for (l=CLDB_list; l != NIL; l=cdr(l)) d = cons(car(car(l)),d); return d; } LISP cldb_select(LISP dbname) { // Select named cldb and make it current EST_String name = get_c_string(dbname); LISP lpair; lpair = siod_assoc_str(name,CLDB_list); if (lpair == NIL) { cerr << "CLDB " << name << " not defined" << endl; festival_error(); } else current_cldb = clunitsdb(car(cdr(lpair))); return dbname; } festival/src/modules/clunits/cljoin.cc0000644000175000017500000001611111465010357016506 0ustar awbawb/*************************************************************************/ /* */ /* Language Technologies Institute */ /* Carnegie Mellon University */ /* Copyright (c) 1999-2004 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author: Alan W Black */ /* Date: July 2001 */ /* --------------------------------------------------------------------- */ /* Some more interesting join methods */ /* based on the mapping code in UniSyn */ /* */ /*************************************************************************/ #include "EST_error.h" #include "us_synthesis.h" #include "festival.h" #if 0 static int awb_voiced(EST_Track &pm,int i) { // Hacky guess at voicedness return TRUE; if ((i < 2) || (i+3 > pm.num_frames())) return FALSE; else if (((pm.t(i) - pm.t(i-1)) == (pm.t(i+1) - pm.t(i))) && ((pm.t(i-1) - pm.t(i-2)) == (pm.t(i) - pm.t(i-1)))) return FALSE; else return TRUE; } static float frame_duration(EST_Track &pm, int i) { if (i <= 0) return frame_duration(pm,i+1); else if (i >= pm.num_frames()) return frame_duration(pm,pm.num_frames()-1); else return pm.t(i)-pm.t(i-1); } static float awb_smoothed(EST_Track &pm, int i) { // Returned smoothed pitch period at i return (frame_duration(pm,i-4)+ frame_duration(pm,i-3)+ frame_duration(pm,i-2)+ frame_duration(pm,i-1)+ frame_duration(pm,i)+ frame_duration(pm,i+1)+ frame_duration(pm,i+2))/7.0; } #endif static void make_segment_varied_mapping(EST_Relation &source_lab, EST_Track &source_pm, EST_Track &target_pm, EST_IVector &map, float dur_impose_factor, float f0_impose_factor) { int n_i, s_i, u_i, u_frames; int spp; float stime, ttime, ltime, dratio, n_frames; int max_frames; EST_Item *u; EST_Track ntarget_pm; ntarget_pm = target_pm; if (target_pm.num_frames() > source_pm.num_frames()) max_frames = target_pm.num_frames()+100; else max_frames = source_pm.num_frames()+100; ntarget_pm.resize(max_frames,target_pm.num_channels()); map.resize(max_frames); // printf("source_lab relations is %s\n",(const char *)source_lab.name()); if (target_pm.t(target_pm.num_frames() - 1) < source_lab.tail()->F("end",0)) { EST_warning("Target pitchmarks end before end of target segment " "timings (%f vs %f). Expect a truncated utterance\n", target_pm.t(target_pm.num_frames() - 1), source_lab.tail()->F("end",0.0)); } n_i = 0; s_i = 0; ltime = 0; for (u = source_lab.head(); u; u = u->next()) { u_frames = u->I("num_frames"); // stime = source_pm.t(s_i+u_frames-1) - source_pm.t(s_i); stime = u->F("unit_duration"); ttime = ffeature(u,"segment_duration"); if (streq("+",(EST_String)ffeature(u,"ph_vc"))) dratio = stime / (stime + ((ttime-stime)*dur_impose_factor)); else dratio = 1; n_frames = (float)u_frames / dratio; /* printf("unit %s dratio %f %d %f time %f, stime %f\n", (const char *)u->name(),dratio,u_frames,n_frames, ttime,stime); */ for (u_i = 0; u_i < n_frames; u_i++,n_i++) { spp = (int)((float)u_i*dratio); if (s_i + spp == 0) ntarget_pm.t(n_i) = ltime; else ntarget_pm.t(n_i) = ltime + source_pm.t(s_i+spp)- source_pm.t(s_i+spp-1); map[n_i] = s_i+spp; ltime = ntarget_pm.t(n_i); if (n_i+1 == ntarget_pm.num_frames()) break; } s_i += u_frames; } ntarget_pm.resize(n_i,ntarget_pm.num_channels()); /* printf("target_pm.end() = %f ntarget_pm.end() = %f\n", target_pm.end(), ntarget_pm.end()); */ target_pm = ntarget_pm; /* printf("target_pm.end() = %f ntarget_pm.end() = %f\n", target_pm.end(), ntarget_pm.end()); */ if (n_i == 0) map.resize(0); // nothing to synthesize else map.resize(n_i - 1); } void cl_mapping(EST_Utterance &utt, LISP params) { EST_Relation *target_lab; EST_IVector *map; EST_Track *source_coef=0, *target_coef=0; float dur_impose_factor, f0_impose_factor; source_coef = track(utt.relation("SourceCoef")->head()->f("coefs")); target_coef = track(utt.relation("TargetCoef")->head()->f("coefs")); target_lab = utt.relation("Segment"); map = new EST_IVector; dur_impose_factor = get_param_float("dur_impose_factor",params,0.0); f0_impose_factor = get_param_float("f0_impose_factor",params,0.0); make_segment_varied_mapping(*target_lab, *source_coef, *target_coef, *map, dur_impose_factor, f0_impose_factor); utt.create_relation("US_map"); EST_Item *item = utt.relation("US_map")->append(); item->set_val("map", est_val(map)); } LISP l_cl_mapping(LISP utt, LISP params) { EST_Utterance *u = get_c_utt(utt); cl_mapping(*u,params); return utt; } festival/src/modules/clunits/clunits.h0000644000175000017500000001030010453726511016547 0ustar awbawb/*************************************************************************/ /* */ /* Carnegie Mellon University and */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998-2000 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH, CARNEGIE MELLON UNIVERSITY AND THE */ /* CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH REGARD TO */ /* THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY */ /* AND FITNESS, IN NO EVENT SHALL THE UNIVERSITY OF EDINBURGH, CARNEGIE */ /* MELLON UNIVERSITY NOR THE CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, */ /* INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER */ /* RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION */ /* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF */ /* OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ /* */ /*************************************************************************/ #ifndef __CLUNITS_H__ #define __CLUNITS_H__ #include "EST_StringTrie.h" class CLunit { public: CLunit(); ~CLunit(); EST_String fileid; EST_String name; EST_String base_name; float start; float mid; float end; class CLunit *prev_unit; class CLunit *next_unit; int samp_start; int samp_end; int middle_frame; EST_Track *join_coeffs; EST_Track *coefs; EST_Wave *sig; }; class CLfile { public: CLfile(); ~CLfile(); EST_Track *join_coeffs; EST_Track *coefs; EST_Wave *sig; }; class CLDB { public: CLDB(); ~CLDB(); LISP params; EST_StringTrie index; EST_StringTrie fileindex; EST_FVector cweights; CLunit *get_unit(const EST_String &name) { return (CLunit *)index.lookup(name); } CLfile *get_fileitem(const EST_String &name) { return (CLfile *)fileindex.lookup(name); } void load_coefs_sig(EST_Item *unit); CLfile *get_file_coefs_sig(const EST_String &fileid); void load_join_coefs(CLunit *unit); CLfile *get_file_join_coefs(const EST_String &fileid); }; LISP cl_load_db(LISP params); LISP acost_utt_load_coeffs(LISP utt, LISP params); LISP make_unit_distance_tables(LISP unittypes, LISP params); LISP ac_distance_tracks(LISP filename1, LISP filename2, LISP lweights); void acost_dt_params(LISP params); float ac_unit_distance(const EST_Track &unit1, const EST_Track &unit2, const EST_FVector wghts); float frame_distance(const EST_Track &a, int ai, const EST_Track &b, int bi, const EST_FVector &wghts, float f0_weight); void cl_maybe_fix_pitch_c0(EST_Track *c); CLDB *check_cldb(); LISP cldb_list(void); LISP cldb_select(LISP dbname); LISP l_cl_mapping(LISP utt, LISP method); #endif festival/src/modules/clunits/acost.scm0000644000175000017500000000636207306543623016553 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Carnegie Mellon University and ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1998-2001 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH, CARNEGIE MELLON UNIVERSITY AND THE ;; ;;; CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH REGARD TO ;; ;;; THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY ;; ;;; AND FITNESS, IN NO EVENT SHALL THE UNIVERSITY OF EDINBURGH, CARNEGIE ;; ;;; MELLON UNIVERSITY NOR THE CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, ;; ;;; INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER ;; ;;; RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION ;; ;;; OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF ;; ;;; OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Finding features and acoustic distance measture for a set of ;;; segments in a database of utterances ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; This is primarily implement for the cluster unit selection method ;;; but may be use in other unit selection schemes. ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; There are five stages to this ;;; Load in all utterances ;;; Load in their coefficients ;;; Collect together the units of the same type ;;; build distance tables from them ;;; dump features for them ;;; ;; Everything is moved to lib/clunits_build.scm and ;; lib/clunits.scm (require_module 'clunits) (require 'clunits_build) (provide 'acost) festival/src/modules/clustergen/Makefile0000644000175000017500000000774112073264745017074 0ustar awbawb########################################################################### ## ## ## --------------------------------------------------------------- ## ## The HMM-Based Speech Synthesis System (HTS): version 1.1b ## ## HTS Working Group ## ## ## ## Department of Computer Science ## ## Nagoya Institute of Technology ## ## and ## ## Interdisciplinary Graduate School of Science and Engineering ## ## Tokyo Institute of Technology ## ## Copyright (c) 2001-2003 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and ## ## distribute this software and its documentation without ## ## restriction, including without limitation the rights to use, ## ## copy, modify, merge, publish, distribute, sublicense, and/or ## ## sell copies of this work, and to permit persons to whom this ## ## work is furnished to do so, subject to the following conditions: ## ## ## ## 1. The code must retain the above copyright notice, this list ## ## of conditions and the following disclaimer. ## ## ## ## 2. Any modifications must be clearly marked as such. ## ## ## ## NAGOYA INSTITUTE OF TECHNOLOGY, TOKYO INSITITUTE OF TECHNOLOGY, ## ## HTS WORKING GROUP, AND THE CONTRIBUTORS TO THIS WORK DISCLAIM ## ## ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL ## ## IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL NAGOYA INSTITUTE OF TECHNOLOGY, TOKYO INSITITUTE OF ## ## TECHNOLOGY, HTS WORKING GROUP, NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY ## ## DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, ## ## WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTUOUS ## ## ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR ## ## PERFORMANCE OF THIS SOFTWARE. ## ## ## ########################################################################### ## This directory contains some low level support for statitsical ## ## parametric speech synthesis, mostly take from Nagoya Institutes of ## ## Technologies HTS Engine (though often with significant modification ## ## ## ## These are specifically designed to support the Clustergen synthesis ## ## technique. ## ## ## ## Alan W Black (awb@cs.cmu.edu) ## ########################################################################### TOP=../../.. DIRNAME=src/modules/clustergen H = mlsa_resynthesis.h HTS_vocoder_me.h vc.h simple_mlpg.h CPPSRCS = clustergen.cc mlsa_resynthesis.cc vc.cc simple_mlpg.cc HTS_vocoder_me.cc SRCS = $(CPPSRCS) OBJS = $(CPPSRCS:.cc=.o) FILES=Makefile $(SRCS) $(H) LOCAL_INCLUDES = -I../include -I ../hts_engine -DFESTIVAL INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/clustergen/clustergen.cc0000644000175000017500000000753412432156007020104 0ustar awbawb/*************************************************************************/ /* */ /* Language Technologies Institute */ /* Carnegie Mellon University */ /* Copyright (c) 2005-2010 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* CARNEGIE MELLON UNIVERSITY AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL CARNEGIE MELLON UNIVERSITY NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Most of the key functions in this directory are derived from HTS */ /* and are only provided with Scheme wraparounds for Clustergen */ /* */ /*************************************************************************/ /* Standard C Libraries */ #include #include #include #include #include "festival.h" LISP mlsa_resynthesis(LISP ltrack, LISP strtrack, LISP filtertrack); LISP mlpg(LISP ltrack); LISP me_mlsa(LISP ltrack, LISP strtrack) { return mlsa_resynthesis(ltrack, strtrack, NULL); } void festival_clustergen_init(void) { proclaim_module("clustergen_engine", "Copyright (C) Carnegie Mellon University 2005-2014\n"); init_subr_3("mlsa_resynthesis", mlsa_resynthesis, "(mlsa_resynthesis TRACK STRTRACK FILTERTRACK)\n\ Return a WAVE synthesized from the F0/MCEP TRACK, STRTRACK is non-nil, use mixed excitation.\n\ If FILTERTRACK is non-nil, it has filters for excitation"); init_subr_1("mlpg",mlpg, "(mlpg TRACK)\n\ Return a track suitable for mlsa from a TRACK with dynamics in it."); init_subr_2("me_mlsa",me_mlsa, "(me_mlsa TRACK STRTRACK)\n\ Return a WAVE synthesized from the F0/MCEP TRACK, STRTRACK is non-nil, use mixed excitation.\n\ Deprecated! Use mlsa_resynthesis instead"); } festival/src/modules/clustergen/mlsa_resynthesis.cc0000644000175000017500000002341612132762232021322 0ustar awbawb/*********************************************************************/ /* */ /* Language Technologies Institute */ /* Carnegie Mellon University */ /* Copyright (c) 2013 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and */ /* distribute this software and its documentation without */ /* restriction, including without limitation the rights to use, */ /* copy, modify, merge, publish, distribute, sublicense, and/or */ /* sell copies of this work, and to permit persons to whom this */ /* work is furnished to do so, subject to the following conditions: */ /* 1. The code must retain the above copyright notice, this list */ /* of conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote */ /* products derived from this software without specific */ /* prior written permission. */ /* */ /* CARNEGIE MELLON UNIVERSITY AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN */ /* NO EVENT SHALL CARNEGIE MELLON UNIVERSITY NOR THE CONTRIBUTORS */ /* BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES */ /* OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA */ /* OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR */ /* OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH */ /* THE USE OR PERFORMANCE OF THIS SOFTWARE. */ /* */ /*********************************************************************/ /* Author: Alok Parlikar (aup@cs.cmu.edu) */ /* Date: January 2013 */ /*********************************************************************/ /* Interface to the MLSA code within HTS Engine API (in the src/modules/hts_engine directory) Written for HTS Engine API 1.07 (Jan 2013) */ #include #include #include #include #include "./vc.h" #include "./HTS_vocoder_me.h" DVECTOR synthesis_body(DMATRIX mcep, // input mel-cep sequence DVECTOR f0v, // input F0 sequence EST_Track *str, // str for mixed excitation EST_Track *filtertrack, double framerate, // sampling frequency (Hz) int framem, // FFT length double alpha, double beta); LISP mlsa_resynthesis(LISP ltrack, LISP strtrack, LISP filtrack) { EST_Track *t; EST_Track *str = 0; EST_Track *filter_track = 0; EST_Wave *wave = 0; DVECTOR w; DMATRIX mcep; DVECTOR f0v; int framerate = 16000; int i, j; int shift; double ALPHA = 0.42; double BETA = 0.0; LISP lispframerate = siod_get_lval("framerate", NULL); if (lispframerate == NIL) { framerate = 16000; } else { framerate = FLONM(lispframerate); } ALPHA = FLONM(siod_get_lval("mlsa_alpha_param", "mlsa: mlsa_alpha_param not set")); BETA = FLONM(siod_get_lval("mlsa_beta_param", "mlsa: mlsa_beta_param not set")); if ((ltrack == NULL) || (TYPEP(ltrack, tc_string) && (streq(get_c_string(ltrack), "nil")))) return siod(new EST_Wave(0, 1, framerate)); t = track(ltrack); if (strtrack != NULL) { /* We have to do mixed-excitation */ str = track(strtrack); } if (filtrack != NULL) { filter_track = track(filtrack); } f0v = xdvalloc(t->num_frames()); mcep = xdmalloc(t->num_frames(), t->num_channels()-1); for (i = 0; i < t->num_frames(); i++) { f0v->data[i] = t->a(i, 0); for (j = 1; j < t->num_channels(); j++) mcep->data[i][j-1] = t->a(i, j); } if (t->num_frames() > 1) { // Hacky way to get floats doing the right thing // Seems to work for 16K and 48KHz -- aup shift = (100000.0 * t->t(1)) - (100000.0 * t->t(0)); shift = shift/10; if (shift % 10 > 5) shift = ceil(shift/10.0); else shift = floor(shift/10.0); } else { shift = 5.0; } if (shift == 0) shift = 5.0; w = synthesis_body(mcep, f0v, str, filter_track, framerate, shift, ALPHA, BETA); wave = new EST_Wave(w->length, 1, framerate); for (i = 0; i < w->length; i++) wave->a(i) = (int16_t)w->data[i]; //NOLINT xdmfree(mcep); xdvfree(f0v); xdvfree(w); return siod(wave); } DVECTOR synthesis_body(DMATRIX mcep, // input mel-cep sequence DVECTOR f0v, // input F0 sequence EST_Track *str, // str for mixed excitation EST_Track *filter_track, // Track for LPF or // Mixed Excitation // Filters double framerate, // sampling frequency (Hz) int framem, // FFT length double alpha, double beta) { int64_t t, pos; int framel; double f0; HTS_Vocoder v; HTS_Vocoder_ME v_me; DVECTOR xd = NODATA; HTS_Boolean use_log_gain = FALSE; size_t stage = 0; // MGC or LSP size_t nlpf = 0; double *lpf = NULL; double volume = 1.0; // Mixed Excitation Stuff LISP filters; LISP f; int fl; int i, j; int me_num_filters = 0; int me_filter_order = 0; double **me_filter = NULL; double *xp_sig = NULL; double *xn_sig = NULL; double *hp = NULL; double *hn = NULL; double *strengths; if (str == NULL) { // Not using mixed excitation, so use the lpf that HTS uses // in the SLT voice as of Jan 2013 -- aup // Check if LPF filter is to be applied to signal. if (filter_track != NULL) { if (filter_track->num_frames() != 1) { printf("Warning: Wrong filter passed. Ignoring LPF\n"); printf("Expected single row for pulse-noise-excitation voice"); } else { nlpf = (filter_track->num_channels() - 1)/2; lpf = (double*) calloc(filter_track->num_channels(), sizeof(double)); for (i = 0; i < filter_track->num_channels(); i++) { lpf[i] = filter_track->a(0,i); } // printf("aup_debug: Using LPF filter of nlpf %d\n", nlpf); } } } // floats may not do the right thing, but this seems to work // for 16KHz and 48KHz -- aup framel = framem * framerate/1000.0; if (str == NULL) { // Not Mixed Excitation HTS_Vocoder_initialize(&v, mcep->col - 1, stage, use_log_gain, framerate, framel); } else { // Mixed Excitation if (filter_track != NULL) { me_num_filters = filter_track->num_frames(); me_filter_order = filter_track->num_channels(); me_filter = walloc(double*, me_num_filters); for (i = 0; i < me_num_filters; i++) { me_filter[i] = walloc(double, me_filter_order); for (j = 0; j < me_filter_order; j++) { me_filter[i][j] = filter_track->a(i, j); } } } else { printf("Warning: Attempting to use Mixed Excitation without Filters"); } xp_sig = (double*) calloc(me_filter_order, sizeof(double)); xn_sig = (double*) calloc(me_filter_order, sizeof(double)); hp = (double*) calloc(me_filter_order, sizeof(double)); hn = (double*) calloc(me_filter_order, sizeof(double)); v_me.v = &v; HTS_Vocoder_initialize_me(&v_me, mcep->col -1, stage, use_log_gain, framerate, framel, me_num_filters, me_filter_order, me_filter, xp_sig, xn_sig, hp, hn); } // synthesize waveforms by MLSA filter xd = xdvalloc(mcep->row * (framel + 2)); for (t = 0, pos = 0; t < mcep->row; t++) { if (t >= f0v->length) f0 = LZERO; else f0 = f0v->data[t]; if (f0 == 0) f0 = LZERO; else f0 = log(f0); if (str == NULL) { // Not Mixed Excitation // printf("aup_debug %d %d %d\n", t, pos, framel); HTS_Vocoder_synthesize(&v, mcep->col - 1, f0, mcep->data[t], nlpf, lpf, alpha, beta, volume, &xd->data[pos], NULL); } else { strengths = (double*) calloc(me_filter_order, sizeof(double)); for (i = 0; i < me_num_filters; i++) { strengths[i] = str->a((int)t, i); } HTS_Vocoder_synthesize_me(&v_me, mcep->col - 1, f0, mcep->data[t], strengths, nlpf, lpf, alpha, beta, volume, &xd->data[pos], NULL); free(strengths); } pos += framel; } if (lpf != NULL) free(lpf); if (str != NULL) { // Mixed Excitation free(xp_sig); free(xn_sig); free(hp); free(hn); } HTS_Vocoder_clear(&v); return xd; } festival/src/modules/clustergen/vc.cc0000644000175000017500000001547211332061167016341 0ustar awbawb/* --------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis System (HTS): version 1.1b */ /* HTS Working Group */ /* */ /* Department of Computer Science */ /* Nagoya Institute of Technology */ /* and */ /* Interdisciplinary Graduate School of Science and Engineering */ /* Tokyo Institute of Technology */ /* Copyright (c) 2001-2003 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and */ /* distribute this software and its documentation without */ /* restriction, including without limitation the rights to use, */ /* copy, modify, merge, publish, distribute, sublicense, and/or */ /* sell copies of this work, and to permit persons to whom this */ /* work is furnished to do so, subject to the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list */ /* of conditions and the following disclaimer. */ /* */ /* 2. Any modifications must be clearly marked as such. */ /* */ /* NAGOYA INSTITUTE OF TECHNOLOGY, TOKYO INSITITUTE OF TECHNOLOGY, */ /* HTS WORKING GROUP, AND THE CONTRIBUTORS TO THIS WORK DISCLAIM */ /* ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL */ /* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL NAGOYA INSTITUTE OF TECHNOLOGY, TOKYO INSITITUTE OF */ /* TECHNOLOGY, HTS WORKING GROUP, NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY */ /* DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, */ /* WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTUOUS */ /* ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR */ /* PERFORMANCE OF THIS SOFTWARE. */ /* */ /* --------------------------------------------------------------- */ /* This is Zen's MLSA filter as ported by Toda to festvox vc */ /* and back ported into hts/festival so we can do MLSA filtering */ /* If I took more time I could probably make this use the same as */ /* as the other code in this directory -- awb@cs.cmu.edu 03JAN06 */ /* --------------------------------------------------------------- */ /* and then ported into Flite (November 2007 awb@cs.cmu.edu) */ /*********************************************************************/ /* */ /* vector (etc) code common to mlpg and mlsa */ /*-------------------------------------------------------------------*/ #include "festival.h" #include "vc.h" /* from vector.cc */ #define cst_free wfree #define cst_alloc walloc LVECTOR xlvalloc(long length) { LVECTOR x; length = MAX(length, 0); x = cst_alloc(struct LVECTOR_STRUCT,1); x->data = cst_alloc(long,MAX(length, 1)); x->imag = NULL; x->length = length; return x; } void xlvfree(LVECTOR x) { if (x != NULL) { if (x->data != NULL) { cst_free(x->data); } if (x->imag != NULL) { cst_free(x->imag); } cst_free(x); } return; } DVECTOR xdvalloc(long length) { DVECTOR x; length = MAX(length, 0); x = cst_alloc(struct DVECTOR_STRUCT,1); x->data = cst_alloc(double,MAX(length, 1)); x->imag = NULL; x->length = length; return x; } void xdvfree(DVECTOR x) { if (x != NULL) { if (x->data != NULL) { cst_free(x->data); } if (x->imag != NULL) { cst_free(x->imag); } cst_free(x); } return; } void dvialloc(DVECTOR x) { if (x->imag != NULL) { cst_free(x->imag); } x->imag = cst_alloc(double,x->length); return; } DVECTOR xdvcut(DVECTOR x, long offset, long length) { long k; long pos; DVECTOR y; y = xdvalloc(length); if (x->imag != NULL) { dvialloc(y); } for (k = 0; k < y->length; k++) { pos = k + offset; if (pos >= 0 && pos < x->length) { y->data[k] = x->data[pos]; if (y->imag != NULL) { y->imag[k] = x->imag[pos]; } } else { y->data[k] = 0.0; if (y->imag != NULL) { y->imag[k] = 0.0; } } } return y; } DMATRIX xdmalloc(long row, long col) { DMATRIX matrix; int i; matrix = cst_alloc(struct DMATRIX_STRUCT,1); matrix->data = cst_alloc(double *,row); for (i=0; idata[i] = cst_alloc(double,col); matrix->imag = NULL; matrix->row = row; matrix->col = col; return matrix; } void xdmfree(DMATRIX matrix) { int i; if (matrix != NULL) { if (matrix->data != NULL) { for (i=0; irow; i++) cst_free(matrix->data[i]); cst_free(matrix->data); } if (matrix->imag != NULL) { for (i=0; irow; i++) cst_free(matrix->imag[i]); cst_free(matrix->imag); } cst_free(matrix); } return; } DVECTOR xdvinit(double j, double incr, double n) { long k; long num; DVECTOR x; if ((incr > 0.0 && j > n) || (incr < 0.0 && j < n)) { x = xdvnull(); return x; } if (incr == 0.0) { num = (long)n; if (num <= 0) { x = xdvnull(); return x; } } else { num = LABS((long)((n - j) / incr)) + 1; } /* memory allocate */ x = xdvalloc(num); /* initailize data */ for (k = 0; k < x->length; k++) { x->data[k] = j + (k * incr); } return x; } /* from voperate.cc */ double dvmax(DVECTOR x, long *index) { long k; long ind; double max; ind = 0; max = x->data[ind]; for (k = 1; k < x->length; k++) { if (max < x->data[k]) { ind = k; max = x->data[k]; } } if (index != NULL) { *index = ind; } return max; } double dvmin(DVECTOR x, long *index) { long k; long ind; double min; ind = 0; min = x->data[ind]; for (k = 1; k < x->length; k++) { if (min > x->data[k]) { ind = k; min = x->data[k]; } } if (index != NULL) { *index = ind; } return min; } double dvsum(DVECTOR x) { long k; double sum; for (k = 0, sum = 0.0; k < x->length; k++) { sum += x->data[k]; } return sum; } festival/src/modules/clustergen/simple_mlpg.cc0000644000175000017500000007554111332061167020244 0ustar awbawb/* --------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis System (HTS): version 1.1.1 */ /* HTS Working Group */ /* */ /* Department of Computer Science */ /* Nagoya Institute of Technology */ /* and */ /* Interdisciplinary Graduate School of Science and Engineering */ /* Tokyo Institute of Technology */ /* Copyright (c) 2001-2003 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and */ /* distribute this software and its documentation without */ /* restriction, including without limitation the rights to use, */ /* copy, modify, merge, publish, distribute, sublicense, and/or */ /* sell copies of this work, and to permit persons to whom this */ /* work is furnished to do so, subject to the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list */ /* of conditions and the following disclaimer. */ /* */ /* 2. Any modifications must be clearly marked as such. */ /* */ /* NAGOYA INSTITUTE OF TECHNOLOGY, TOKYO INSITITUTE OF TECHNOLOGY, */ /* HTS WORKING GROUP, AND THE CONTRIBUTORS TO THIS WORK DISCLAIM */ /* ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL */ /* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL NAGOYA INSTITUTE OF TECHNOLOGY, TOKYO INSITITUTE OF */ /* TECHNOLOGY, HTS WORKING GROUP, NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY */ /* DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, */ /* WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS */ /* ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR */ /* PERFORMANCE OF THIS SOFTWARE. */ /* */ /* --------------------------------------------------------------- */ /* mlpg.c : speech parameter generation from pdf sequence */ /* */ /* 2003/12/26 by Heiga Zen */ /* --------------------------------------------------------------- */ /*********************************************************************/ /* */ /* Nagoya Institute of Technology, Aichi, Japan, */ /* and */ /* Carnegie Mellon University, Pittsburgh, PA */ /* Copyright (c) 2003-2004,2008 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and */ /* distribute this software and its documentation without */ /* restriction, including without limitation the rights to use, */ /* copy, modify, merge, publish, distribute, sublicense, and/or */ /* sell copies of this work, and to permit persons to whom this */ /* work is furnished to do so, subject to the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list */ /* of conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* */ /* NAGOYA INSTITUTE OF TECHNOLOGY, CARNEGIE MELLON UNIVERSITY, AND */ /* THE CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH */ /* REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL NAGOYA INSTITUTE */ /* OF TECHNOLOGY, CARNEGIE MELLON UNIVERSITY, NOR THE CONTRIBUTORS */ /* BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR */ /* ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR */ /* PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER */ /* TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE */ /* OR PERFORMANCE OF THIS SOFTWARE. */ /* */ /*********************************************************************/ /* */ /* Author : Tomoki Toda (tomoki@ics.nitech.ac.jp) */ /* Date : June 2004 */ /* */ /* Modified as a single file for inclusion in festival/flite */ /* May 2008 awb@cs.cmu.edu */ /*-------------------------------------------------------------------*/ /* */ /* ML-Based Parameter Generation */ /* */ /*-------------------------------------------------------------------*/ #include "festival.h" #include "simple_mlpg.h" #define mlpg_alloc(X,Y) (walloc(Y,X)) #define mlpg_free wfree #define cst_errmsg EST_error #define cst_error() #ifdef CMUFLITE #include "cst_alloc.h" #include "cst_string.h" #include "cst_math.h" #include "cst_vc.h" #include "cst_track.h" #include "cst_wave.h" #include "cst_mlpg.h" #define mlpg_alloc(X,Y) (cst_alloc(Y,X)) #define mlpg_free cst_free #endif static MLPGPARA xmlpgpara_init(int dim, int dim2, int dnum, int clsnum); static void xmlpgparafree(MLPGPARA param); static double get_like_pdfseq_vit(int dim, int dim2, int dnum, int clsnum, MLPGPARA param, EST_Track *model, XBOOL dia_flag); static void get_dltmat(DMATRIX mat, MLPG_DWin *dw, int dno, DMATRIX dmat); static double *dcalloc(int x, int xoff); static double **ddcalloc(int x, int y, int xoff, int yoff); /***********************************/ /* ML using Choleski decomposition */ /***********************************/ /* Diagonal Covariance Version */ static void InitDWin(PStreamChol *pst, const float *dynwin, int fsize); static void InitPStreamChol(PStreamChol *pst, const float *dynwin, int fsize, int order, int T); static void mlgparaChol(DMATRIX pdf, PStreamChol *pst, DMATRIX mlgp); static void mlpgChol(PStreamChol *pst); static void calc_R_and_r(PStreamChol *pst, const int m); static void Choleski(PStreamChol *pst); static void Choleski_forward(PStreamChol *pst); static void Choleski_backward(PStreamChol *pst, const int m); static double get_gauss_full(long clsidx, DVECTOR vec, // [dim] DVECTOR detvec, // [clsnum] DMATRIX weightmat, // [clsnum][1] DMATRIX meanvec, // [clsnum][dim] DMATRIX invcovmat); // [clsnum * dim][dim] static double get_gauss_dia(long clsidx, DVECTOR vec, // [dim] DVECTOR detvec, // [clsnum] DMATRIX weightmat, // [clsnum][1] DMATRIX meanmat, // [clsnum][dim] DMATRIX invcovmat); // [clsnum][dim] static double cal_xmcxmc(long clsidx, DVECTOR x, DMATRIX mm, // [num class][dim] DMATRIX cm); // [num class * dim][dim] const float mlpg_dynwin[] = { -0.5, 0.0, 0.5 }; #define mlpg_dynwinsize 3 static MLPGPARA xmlpgpara_init(int dim, int dim2, int dnum, int clsnum) { MLPGPARA param; // memory allocation param = mlpg_alloc(1,struct MLPGPARA_STRUCT); param->ov = xdvalloc(dim); param->iuv = NODATA; param->iumv = NODATA; param->flkv = xdvalloc(dnum); param->stm = NODATA; param->dltm = xdmalloc(dnum, dim2); param->pdf = NODATA; param->detvec = NODATA; param->wght = xdmalloc(clsnum, 1); param->mean = xdmalloc(clsnum, dim); param->cov = NODATA; param->clsidxv = NODATA; /* dia_flag */ param->clsdetv = xdvalloc(1); param->clscov = xdmalloc(1, dim); param->vdet = 1.0; param->vm = NODATA; param->vv = NODATA; param->var = NODATA; return param; } static void xmlpgparafree(MLPGPARA param) { if (param != NODATA) { if (param->ov != NODATA) xdvfree(param->ov); if (param->iuv != NODATA) xdvfree(param->iuv); if (param->iumv != NODATA) xdvfree(param->iumv); if (param->flkv != NODATA) xdvfree(param->flkv); if (param->stm != NODATA) xdmfree(param->stm); if (param->dltm != NODATA) xdmfree(param->dltm); if (param->pdf != NODATA) xdmfree(param->pdf); if (param->detvec != NODATA) xdvfree(param->detvec); if (param->wght != NODATA) xdmfree(param->wght); if (param->mean != NODATA) xdmfree(param->mean); if (param->cov != NODATA) xdmfree(param->cov); if (param->clsidxv != NODATA) xlvfree(param->clsidxv); if (param->clsdetv != NODATA) xdvfree(param->clsdetv); if (param->clscov != NODATA) xdmfree(param->clscov); if (param->vm != NODATA) xdvfree(param->vm); if (param->vv != NODATA) xdvfree(param->vv); if (param->var != NODATA) xdvfree(param->var); mlpg_free(param); } return; } static double get_like_pdfseq_vit(int dim, int dim2, int dnum, int clsnum, MLPGPARA param, EST_Track *model, XBOOL dia_flag) { long d, c, k, l, j; double sumgauss; double like = 0.0; for (d = 0, like = 0.0; d < dnum; d++) { // read weight and mean sequences param->wght->data[0][0] = 0.9; /* FIXME weights */ for (j=0; jmean->data[0][j] = model->a((int)d,(int)((j+1)*2)); // observation vector for (k = 0; k < dim2; k++) { param->ov->data[k] = param->stm->data[d][k]; param->ov->data[k + dim2] = param->dltm->data[d][k]; } // mixture index c = d; param->clsdetv->data[0] = param->detvec->data[c]; // calculating likelihood if (dia_flag == XTRUE) { for (k = 0; k < param->clscov->col; k++) param->clscov->data[0][k] = param->cov->data[c][k]; sumgauss = get_gauss_dia(0, param->ov, param->clsdetv, param->wght, param->mean, param->clscov); } else { for (k = 0; k < param->clscov->row; k++) for (l = 0; l < param->clscov->col; l++) param->clscov->data[k][l] = param->cov->data[k + param->clscov->row * c][l]; sumgauss = get_gauss_full(0, param->ov, param->clsdetv, param->wght, param->mean, param->clscov); } if (sumgauss <= 0.0) param->flkv->data[d] = -1.0 * INFTY2; else param->flkv->data[d] = log(sumgauss); like += param->flkv->data[d]; // estimating U', U'*M if (dia_flag == XTRUE) { // PDF [U'*M U'] for (k = 0; k < dim; k++) { param->pdf->data[d][k] = param->clscov->data[0][k] * param->mean->data[0][k]; param->pdf->data[d][k + dim] = param->clscov->data[0][k]; } } else { // PDF [U'*M U'] for (k = 0; k < dim; k++) { param->pdf->data[d][k] = 0.0; for (l = 0; l < dim; l++) { param->pdf->data[d][k * dim + dim + l] = param->clscov->data[k][l]; param->pdf->data[d][k] += param->clscov->data[k][l] * param->mean->data[0][l]; } } } } like /= (double)dnum; return like; } #if 0 static double get_like_gv(long dim2, long dnum, MLPGPARA param) { long k; double av = 0.0, dif = 0.0; double vlike = -INFTY; if (param->vm != NODATA && param->vv != NODATA) { for (k = 0; k < dim2; k++) calc_varstats(param->stm->data, k, dnum, &av, &(param->var->data[k]), &dif); vlike = log(get_gauss_dia5(param->vdet, 1.0, param->var, param->vm, param->vv)); } return vlike; } static void sm_mvav(DMATRIX mat, long hlen) { long k, l, m, p; double d, sd; DVECTOR vec = NODATA; DVECTOR win = NODATA; vec = xdvalloc(mat->row); // smoothing window win = xdvalloc(hlen * 2 + 1); for (k = 0, d = 1.0, sd = 0.0; k < hlen; k++, d += 1.0) { win->data[k] = d; win->data[win->length - k - 1] = d; sd += d + d; } win->data[k] = d; sd += d; for (k = 0; k < win->length; k++) win->data[k] /= sd; for (l = 0; l < mat->col; l++) { for (k = 0; k < mat->row; k++) { for (m = 0, vec->data[k] = 0.0; m < win->length; m++) { p = k - hlen + m; if (p >= 0 && p < mat->row) vec->data[k] += mat->data[p][l] * win->data[m]; } } for (k = 0; k < mat->row; k++) mat->data[k][l] = vec->data[k]; } xdvfree(win); xdvfree(vec); return; } #endif static void get_dltmat(DMATRIX mat, MLPG_DWin *dw, int dno, DMATRIX dmat) { int i, j, k, tmpnum; tmpnum = (int)mat->row - dw->width[dno][WRIGHT]; for (k = dw->width[dno][WRIGHT]; k < tmpnum; k++) // time index for (i = 0; i < (int)mat->col; i++) // dimension index for (j = dw->width[dno][WLEFT], dmat->data[k][i] = 0.0; j <= dw->width[dno][WRIGHT]; j++) dmat->data[k][i] += mat->data[k + j][i] * dw->coef[dno][j]; for (i = 0; i < (int)mat->col; i++) { // dimension index for (k = 0; k < dw->width[dno][WRIGHT]; k++) // time index for (j = dw->width[dno][WLEFT], dmat->data[k][i] = 0.0; j <= dw->width[dno][WRIGHT]; j++) if (k + j >= 0) dmat->data[k][i] += mat->data[k + j][i] * dw->coef[dno][j]; else dmat->data[k][i] += (2.0 * mat->data[0][i] - mat->data[-k - j][i]) * dw->coef[dno][j]; for (k = tmpnum; k < (int)mat->row; k++) // time index for (j = dw->width[dno][WLEFT], dmat->data[k][i] = 0.0; j <= dw->width[dno][WRIGHT]; j++) if (k + j < (int)mat->row) dmat->data[k][i] += mat->data[k + j][i] * dw->coef[dno][j]; else dmat->data[k][i] += (2.0 * mat->data[mat->row - 1][i] - mat->data[mat->row - k - j + mat->row - 2][i]) * dw->coef[dno][j]; } return; } static double *dcalloc(int x, int xoff) { double *ptr; ptr = mlpg_alloc(x,double); /* ptr += xoff; */ /* Just not going to allow this */ return(ptr); } static double **ddcalloc(int x, int y, int xoff, int yoff) { double **ptr; register int i; ptr = mlpg_alloc(x,double *); for (i = 0; i < x; i++) ptr[i] = dcalloc(y, yoff); /* ptr += xoff; */ /* Just not going to allow this */ return(ptr); } ///////////////////////////////////// // ML using Choleski decomposition // ///////////////////////////////////// static void InitDWin(PStreamChol *pst, const float *dynwin, int fsize) { int i,j; int leng; pst->dw.num = 1; // only static if (dynwin) { pst->dw.num = 2; // static + dyn } // memory allocation pst->dw.width = mlpg_alloc(pst->dw.num,int *); for (i = 0; i < pst->dw.num; i++) pst->dw.width[i] = mlpg_alloc(2,int); pst->dw.coef = mlpg_alloc(pst->dw.num, double *); pst->dw.coef_ptrs = mlpg_alloc(pst->dw.num, double *); // window for static parameter WLEFT = 0, WRIGHT = 1 pst->dw.width[0][WLEFT] = pst->dw.width[0][WRIGHT] = 0; pst->dw.coef_ptrs[0] = mlpg_alloc(1,double); pst->dw.coef[0] = pst->dw.coef_ptrs[0]; pst->dw.coef[0][0] = 1.0; // set delta coefficients for (i = 1; i < pst->dw.num; i++) { pst->dw.coef_ptrs[i] = mlpg_alloc(fsize, double); pst->dw.coef[i] = pst->dw.coef_ptrs[i]; for (j=0; jdw.coef[i][j] = (double)dynwin[j]; // set pointer leng = fsize / 2; // L (fsize = 2 * L + 1) pst->dw.coef[i] += leng; // [L] -> [0] center pst->dw.width[i][WLEFT] = -leng; // -L left pst->dw.width[i][WRIGHT] = leng; // L right if (fsize % 2 == 0) pst->dw.width[i][WRIGHT]--; } pst->dw.maxw[WLEFT] = pst->dw.maxw[WRIGHT] = 0; for (i = 0; i < pst->dw.num; i++) { if (pst->dw.maxw[WLEFT] > pst->dw.width[i][WLEFT]) pst->dw.maxw[WLEFT] = pst->dw.width[i][WLEFT]; if (pst->dw.maxw[WRIGHT] < pst->dw.width[i][WRIGHT]) pst->dw.maxw[WRIGHT] = pst->dw.width[i][WRIGHT]; } return; } static void InitPStreamChol(PStreamChol *pst, const float *dynwin, int fsize, int order, int T) { // order of cepstrum pst->order = order; // windows for dynamic feature InitDWin(pst, dynwin, fsize); // dimension of observed vector pst->vSize = (pst->order + 1) * pst->dw.num; // odim = dim * (1--3) // memory allocation pst->T = T; // number of frames pst->width = pst->dw.maxw[WRIGHT] * 2 + 1; // width of R pst->mseq = ddcalloc(T, pst->vSize, 0, 0); // [T][odim] pst->ivseq = ddcalloc(T, pst->vSize, 0, 0); // [T][odim] pst->R = ddcalloc(T, pst->width, 0, 0); // [T][width] pst->r = dcalloc(T, 0); // [T] pst->g = dcalloc(T, 0); // [T] pst->c = ddcalloc(T, pst->order + 1, 0, 0); // [T][dim] return; } static void mlgparaChol(DMATRIX pdf, PStreamChol *pst, DMATRIX mlgp) { int t, d; // error check if (pst->vSize * 2 != pdf->col || pst->order + 1 != mlgp->col) { cst_errmsg("Error mlgparaChol: Different dimension\n"); cst_error(); } // mseq: U^{-1}*M, ifvseq: U^{-1} for (t = 0; t < pst->T; t++) { for (d = 0; d < pst->vSize; d++) { pst->mseq[t][d] = pdf->data[t][d]; pst->ivseq[t][d] = pdf->data[t][pst->vSize + d]; } } // ML parameter generation mlpgChol(pst); // extracting parameters for (t = 0; t < pst->T; t++) for (d = 0; d <= pst->order; d++) mlgp->data[t][d] = pst->c[t][d]; return; } // generate parameter sequence from pdf sequence using Choleski decomposition static void mlpgChol(PStreamChol *pst) { register int m; // generating parameter in each dimension for (m = 0; m <= pst->order; m++) { calc_R_and_r(pst, m); Choleski(pst); Choleski_forward(pst); Choleski_backward(pst, m); } return; } //------ parameter generation fuctions // calc_R_and_r: calculate R = W'U^{-1}W and r = W'U^{-1}M static void calc_R_and_r(PStreamChol *pst, const int m) { register int i, j, k, l, n; double wu; for (i = 0; i < pst->T; i++) { pst->r[i] = pst->mseq[i][m]; pst->R[i][0] = pst->ivseq[i][m]; for (j = 1; j < pst->width; j++) pst->R[i][j] = 0.0; for (j = 1; j < pst->dw.num; j++) { for (k = pst->dw.width[j][0]; k <= pst->dw.width[j][1]; k++) { n = i + k; if (n >= 0 && n < pst->T && pst->dw.coef[j][-k] != 0.0) { l = j * (pst->order + 1) + m; pst->r[i] += pst->dw.coef[j][-k] * pst->mseq[n][l]; wu = pst->dw.coef[j][-k] * pst->ivseq[n][l]; for (l = 0; l < pst->width; l++) { n = l-k; if (n <= pst->dw.width[j][1] && i + l < pst->T && pst->dw.coef[j][n] != 0.0) pst->R[i][l] += wu * pst->dw.coef[j][n]; } } } } } return; } // Choleski: Choleski factorization of Matrix R static void Choleski(PStreamChol *pst) { register int t, j, k; pst->R[0][0] = sqrt(pst->R[0][0]); for (j = 1; j < pst->width; j++) pst->R[0][j] /= pst->R[0][0]; for (t = 1; t < pst->T; t++) { for (j = 1; j < pst->width; j++) if (t - j >= 0) pst->R[t][0] -= pst->R[t - j][j] * pst->R[t - j][j]; pst->R[t][0] = sqrt(pst->R[t][0]); for (j = 1; j < pst->width; j++) { for (k = 0; k < pst->dw.maxw[WRIGHT]; k++) if (j != pst->width - 1) pst->R[t][j] -= pst->R[t - k - 1][j - k] * pst->R[t - k - 1][j + 1]; pst->R[t][j] /= pst->R[t][0]; } } return; } // Choleski_forward: forward substitution to solve linear equations static void Choleski_forward(PStreamChol *pst) { register int t, j; double hold; pst->g[0] = pst->r[0] / pst->R[0][0]; for (t=1; t < pst->T; t++) { hold = 0.0; for (j = 1; j < pst->width; j++) if (t - j >= 0 && pst->R[t - j][j] != 0.0) hold += pst->R[t - j][j] * pst->g[t - j]; pst->g[t] = (pst->r[t] - hold) / pst->R[t][0]; } return; } // Choleski_backward: backward substitution to solve linear equations static void Choleski_backward(PStreamChol *pst, const int m) { register int t, j; double hold; pst->c[pst->T - 1][m] = pst->g[pst->T - 1] / pst->R[pst->T - 1][0]; for (t = pst->T - 2; t >= 0; t--) { hold = 0.0; for (j = 1; j < pst->width; j++) if (t + j < pst->T && pst->R[t][j] != 0.0) hold += pst->R[t][j] * pst->c[t + j][m]; pst->c[t][m] = (pst->g[t] - hold) / pst->R[t][0]; } return; } //////////////////////////////////// // ML Considering Global Variance // //////////////////////////////////// #if 0 static void varconv(double **c, const int m, const int T, const double var) { register int n; double sd, osd; double oav = 0.0, ovar = 0.0, odif = 0.0; calc_varstats(c, m, T, &oav, &ovar, &odif); osd = sqrt(ovar); sd = sqrt(var); for (n = 0; n < T; n++) c[n][m] = (c[n][m] - oav) / osd * sd + oav; return; } static void calc_varstats(double **c, const int m, const int T, double *av, double *var, double *dif) { register int i; register double d; *av = 0.0; *var = 0.0; *dif = 0.0; for (i = 0; i < T; i++) *av += c[i][m]; *av /= (double)T; for (i = 0; i < T; i++) { d = c[i][m] - *av; *var += d * d; *dif += d; } *var /= (double)T; return; } // Diagonal Covariance Version static void mlgparaGrad(DMATRIX pdf, PStreamChol *pst, DMATRIX mlgp, const int max, double th, double e, double alpha, DVECTOR vm, DVECTOR vv, XBOOL nrmflag, XBOOL extvflag) { int t, d; // error check if (pst->vSize * 2 != pdf->col || pst->order + 1 != mlgp->col) { cst_errmsg("Error mlgparaChol: Different dimension\n"); cst_error(); } // mseq: U^{-1}*M, ifvseq: U^{-1} for (t = 0; t < pst->T; t++) { for (d = 0; d < pst->vSize; d++) { pst->mseq[t][d] = pdf->data[t][d]; pst->ivseq[t][d] = pdf->data[t][pst->vSize + d]; } } // ML parameter generation mlpgChol(pst); // extend variance if (extvflag == XTRUE) for (d = 0; d <= pst->order; d++) varconv(pst->c, d, pst->T, vm->data[d]); // estimating parameters mlpgGrad(pst, max, th, e, alpha, vm, vv, nrmflag); // extracting parameters for (t = 0; t < pst->T; t++) for (d = 0; d <= pst->order; d++) mlgp->data[t][d] = pst->c[t][d]; return; } // generate parameter sequence from pdf sequence using gradient static void mlpgGrad(PStreamChol *pst, const int max, double th, double e, double alpha, DVECTOR vm, DVECTOR vv, XBOOL nrmflag) { register int m, i, t; double diff, n, dth; if (nrmflag == XTRUE) n = (double)(pst->T * pst->vSize) / (double)(vm->length); else n = 1.0; // generating parameter in each dimension for (m = 0; m <= pst->order; m++) { calc_R_and_r(pst, m); dth = th * sqrt(vm->data[m]); for (i = 0; i < max; i++) { calc_grad(pst, m); if (vm != NODATA && vv != NODATA) calc_vargrad(pst, m, alpha, n, vm->data[m], vv->data[m]); for (t = 0, diff = 0.0; t < pst->T; t++) { diff += pst->g[t] * pst->g[t]; pst->c[t][m] += e * pst->g[t]; } diff = sqrt(diff / (double)pst->T); if (diff < dth || diff == 0.0) break; } } return; } // calc_grad: calculate -RX + r = -W'U^{-1}W * X + W'U^{-1}M void calc_grad(PStreamChol *pst, const int m) { register int i, j; for (i = 0; i < pst->T; i++) { pst->g[i] = pst->r[i] - pst->c[i][m] * pst->R[i][0]; for (j = 1; j < pst->width; j++) { if (i + j < pst->T) pst->g[i] -= pst->c[i + j][m] * pst->R[i][j]; if (i - j >= 0) pst->g[i] -= pst->c[i - j][m] * pst->R[i - j][j]; } } return; } static void calc_vargrad(PStreamChol *pst, const int m, double alpha, double n, double vm, double vv) { register int i; double vg, w1, w2; double av = 0.0, var = 0.0, dif = 0.0; if (alpha > 1.0 || alpha < 0.0) { w1 = 1.0; w2 = 1.0; } else { w1 = alpha; w2 = 1.0 - alpha; } calc_varstats(pst->c, m, pst->T, &av, &var, &dif); for (i = 0; i < pst->T; i++) { vg = -(var - vm) * (pst->c[i][m] - av) * vv * 2.0 / (double)pst->T; pst->g[i] = w1 * pst->g[i] / n + w2 * vg; } return; } #endif // diagonal covariance static DVECTOR xget_detvec_diamat2inv(DMATRIX covmat) // [num class][dim] { long dim, clsnum; long i, j; double det; DVECTOR detvec = NODATA; clsnum = covmat->row; dim = covmat->col; // memory allocation detvec = xdvalloc(clsnum); for (i = 0; i < clsnum; i++) { for (j = 0, det = 1.0; j < dim; j++) { det *= covmat->data[i][j]; if (det > 0.0) { covmat->data[i][j] = 1.0 / covmat->data[i][j]; } else { cst_errmsg("error:(class %ld) determinant <= 0, det = %f\n", i, det); xdvfree(detvec); return NODATA; } } detvec->data[i] = det; } return detvec; } static double get_gauss_full(long clsidx, DVECTOR vec, // [dim] DVECTOR detvec, // [clsnum] DMATRIX weightmat, // [clsnum][1] DMATRIX meanvec, // [clsnum][dim] DMATRIX invcovmat) // [clsnum * dim][dim] { double gauss; if (detvec->data[clsidx] <= 0.0) { cst_errmsg("#error: det <= 0.0\n"); cst_error(); } gauss = weightmat->data[clsidx][0] / sqrt(pow(2.0 * PI, (double)vec->length) * detvec->data[clsidx]) * exp(-1.0 * cal_xmcxmc(clsidx, vec, meanvec, invcovmat) / 2.0); return gauss; } static double cal_xmcxmc(long clsidx, DVECTOR x, DMATRIX mm, // [num class][dim] DMATRIX cm) // [num class * dim][dim] { long clsnum, k, l, b, dim; double *vec = NULL; double td, d; dim = x->length; clsnum = mm->row; b = clsidx * dim; if (mm->col != dim || cm->col != dim || clsnum * dim != cm->row) { cst_errmsg("Error cal_xmcxmc: different dimension\n"); cst_error(); } // memory allocation vec = mlpg_alloc((int)dim, double); for (k = 0; k < dim; k++) vec[k] = x->data[k] - mm->data[clsidx][k]; for (k = 0, d = 0.0; k < dim; k++) { for (l = 0, td = 0.0; l < dim; l++) td += vec[l] * cm->data[l + b][k]; d += td * vec[k]; } // memory free mlpg_free(vec); vec = NULL; return d; } #if 0 // diagonal covariance static double get_gauss_dia5(double det, double weight, DVECTOR vec, // dim DVECTOR meanvec, // dim DVECTOR invcovvec) // dim { double gauss, sb; long k; if (det <= 0.0) { cst_errmsg("#error: det <= 0.0\n"); cst_error(); } for (k = 0, gauss = 0.0; k < vec->length; k++) { sb = vec->data[k] - meanvec->data[k]; gauss += sb * invcovvec->data[k] * sb; } gauss = weight / sqrt(pow(2.0 * PI, (double)vec->length) * det) * exp(-gauss / 2.0); return gauss; } #endif static double get_gauss_dia(long clsidx, DVECTOR vec, // [dim] DVECTOR detvec, // [clsnum] DMATRIX weightmat, // [clsnum][1] DMATRIX meanmat, // [clsnum][dim] DMATRIX invcovmat) // [clsnum][dim] { double gauss, sb; long k; if (detvec->data[clsidx] <= 0.0) { cst_errmsg("#error: det <= 0.0\n"); cst_error(); } for (k = 0, gauss = 0.0; k < vec->length; k++) { sb = vec->data[k] - meanmat->data[clsidx][k]; gauss += sb * invcovmat->data[clsidx][k] * sb; } gauss = weightmat->data[clsidx][0] / sqrt(pow(2.0 * PI, (double)vec->length) * detvec->data[clsidx]) * exp(-gauss / 2.0); return gauss; } static void pst_free(PStreamChol *pst) { int i; for (i=0; idw.num; i++) mlpg_free(pst->dw.width[i]); mlpg_free(pst->dw.width); pst->dw.width = NULL; for (i=0; idw.num; i++) mlpg_free(pst->dw.coef_ptrs[i]); mlpg_free(pst->dw.coef); pst->dw.coef = NULL; mlpg_free(pst->dw.coef_ptrs); pst->dw.coef_ptrs = NULL; for (i=0; iT; i++) mlpg_free(pst->mseq[i]); mlpg_free(pst->mseq); for (i=0; iT; i++) mlpg_free(pst->ivseq[i]); mlpg_free(pst->ivseq); for (i=0; iT; i++) mlpg_free(pst->R[i]); mlpg_free(pst->R); mlpg_free(pst->r); mlpg_free(pst->g); for (i=0; iT; i++) mlpg_free(pst->c[i]); mlpg_free(pst->c); return; } LISP mlpg(LISP ltrack) { /* Generate an (mcep) track using Maximum Likelihood Parameter Generation */ MLPGPARA param = NODATA; EST_Track *param_track, *out; int dim, dim_st; float like; int i,j; int nframes; PStreamChol pst; if ((ltrack == NULL) || (TYPEP(ltrack,tc_string) && (streq(get_c_string(ltrack),"nil")))) return NIL; param_track = track(ltrack); nframes = param_track->num_frames(); dim = (param_track->num_channels()/2)-1; dim_st = dim/2; /* dim2 in original code */ out = new EST_Track(); out->copy_setup(*param_track); out->resize(nframes,dim_st+1); param = xmlpgpara_init(dim,dim_st,nframes,nframes); // mixture-index sequence param->clsidxv = xlvalloc(nframes); for (i=0; iclsidxv->data[i] = i; // initial static feature sequence param->stm = xdmalloc(nframes,dim_st); for (i=0; istm->data[i][j] = param_track->a(i,(j+1)*2); } /* Load cluster means */ for (i=0; imean->data[i][j] = param_track->a(i,(j+1)*2); /* GMM parameters diagonal covariance */ InitPStreamChol(&pst, mlpg_dynwin, mlpg_dynwinsize, dim_st-1, nframes); param->pdf = xdmalloc(nframes,dim*2); param->cov = xdmalloc(nframes,dim); for (i=0; icov->data[i][j] = param_track->a(i,(j+1)*2+1) * param_track->a(i,(j+1)*2+1); param->detvec = xget_detvec_diamat2inv(param->cov); /* global variance parameters */ /* TBD get_gv_mlpgpara(param, vmfile, vvfile, dim2, msg_flag); */ if (nframes > 0) { get_dltmat(param->stm, &pst.dw, 1, param->dltm); like = get_like_pdfseq_vit(dim, dim_st, nframes, nframes, param, param_track, XTRUE); /* vlike = get_like_gv(dim2, dnum, param); */ mlgparaChol(param->pdf, &pst, param->stm); } /* Put the answer back into the output track */ for (i=0; it(i) = param_track->t(i); out->a(i,0) = param_track->a(i,0); /* F0 */ for (j=0; ja(i,j+1) = param->stm->data[i][j]; } } // memory free xmlpgparafree(param); pst_free(&pst); return siod(out); } festival/src/modules/clustergen/HTS_vocoder_me.cc0000644000175000017500000003364412110442172020564 0ustar awbawb/************************************************************************************/ /* Copyright (c) 2012 The Department of Arts and Culture, */ /* The Government of the Republic of South Africa. */ /* */ /* Contributors: Meraka Institute, CSIR, South Africa. */ /* */ /* Permission is hereby granted, free of charge, to any person obtaining a copy */ /* of this software and associated documentation files (the "Software"), to deal */ /* in the Software without restriction, including without limitation the rights */ /* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell */ /* copies of the Software, and to permit persons to whom the Software is */ /* furnished to do so, subject to the following conditions: */ /* The above copyright notice and this permission notice shall be included in */ /* all copies or substantial portions of the Software. */ /* */ /* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR */ /* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, */ /* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE */ /* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER */ /* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, */ /* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN */ /* THE SOFTWARE. */ /* */ /************************************************************************************/ /* */ /* AUTHOR : Aby Louw */ /* DATE : 14 May 2012 */ /* */ /************************************************************************************/ /* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ /* */ /* Added mixed excitation, see code at end of file. */ /* Alok Parlikar modified this file for use with HTS Engine 1.07 */ /************************************************************************************/ #ifndef HTS_VOCODER_ME_C #define HTS_VOCODER_ME_C #ifdef __cplusplus #define HTS_VOCODER_ME_C_START extern "C" { #define HTS_VOCODER_ME_C_END } #else #define HTS_VOCODER_ME_C_START #define HTS_VOCODER_ME_C_END #endif /* __CPLUSPLUS */ HTS_VOCODER_ME_C_START; #include "./HTS_vocoder_me.h" #include "../hts_engine/HTS_vocoder.c" /* HTS_Vocoder_initialize_me: initialize vocoder (mixed excitation) */ void HTS_Vocoder_initialize_me(HTS_Vocoder_ME * v_me, const int m, const int stage, HTS_Boolean use_log_gain, const int rate, const int fperiod, int num_filters, int filter_order, double **h, double *xp_sig, double *xn_sig, double *hp, double *hn) { int i; HTS_Vocoder *v = v_me->v; /* access to original HTS_Vocoder structure */ /* set parameter */ v->stage = stage; if (stage != 0) v->gamma = -1.0 / v->stage; else v->gamma = 0.0; v->use_log_gain = use_log_gain; v->fprd = fperiod; v->next = SEED; v->gauss = GAUSS; v->rate = rate; v->p1 = -1.0; v->sw = 0; v->x = 0x55555555; /* init buffer */ v->freqt_buff = NULL; v->freqt_size = 0; v->gc2gc_buff = NULL; v->gc2gc_size = 0; v->lsp2lpc_buff = NULL; v->lsp2lpc_size = 0; v->postfilter_buff = NULL; v->postfilter_size = 0; v->spectrum2en_buff = NULL; v->spectrum2en_size = 0; if (v->stage == 0) { /* for MCP */ v->c = (double *) HTS_calloc(m * (3 + PADEORDER) + 5 * PADEORDER + 6, sizeof(double)); v->cc = v->c + m + 1; v->cinc = v->cc + m + 1; v->d1 = v->cinc + m + 1; } else { /* for LSP */ v->c = (double *) HTS_calloc((m + 1) * (v->stage + 3), sizeof(double)); v->cc = v->c + m + 1; v->cinc = v->cc + m + 1; v->d1 = v->cinc + m + 1; } /* ABY: think this is not required */ v->pulse_buff = NULL; v->pulse_size = 0; /* initialize mixed excitation variables */ /*---------------------------------------*/ v_me->num_filters = num_filters; v_me->filter_order = filter_order; v_me->xp_sig = xp_sig; v_me->xn_sig = xn_sig; /* initialise xp_sig and xn_sig */ for (i = 0; i < v_me->filter_order; i++) { v_me->xp_sig[i] = 0.0; v_me->xn_sig[i] = 0.0; } /* allocate memory for pulse and noise shaping filters */ v_me->hp = hp; v_me->hn = hn; /* get filter coefs */ v_me->h = h; } /* HTS_Vocoder_synthesize_me: mixed excitation and MLSA/MGLSA filster * based waveform synthesis */ void HTS_Vocoder_synthesize_me(HTS_Vocoder_ME * v_me, const int m, double lf0, double *spectrum, double *strengths, size_t nlpf, double *lpf, double alpha, double beta, double volume, double *rawdata, HTS_Audio * audio) { double x; int i, j; int k; short xs; int rawidx = 0; double p; HTS_Vocoder *v = v_me->v; /* access to original HTS_Vocoder struct */ double xpulse; double xnoise; double fxpulse; double fxnoise; double e1, e2; /* Copy in str's and build pulse and noise shaping filter for this frame */ for (i = 0; i < v_me->filter_order; i++) { v_me->hp[i] = v_me->hn[i] = 0.0; for (j = 0; j < v_me->num_filters; j++) { v_me->hp[i] += strengths[j] * v_me->h[j][i]; v_me->hn[i] += (1 - strengths[j]) * v_me->h[j][i]; } } /* lf0 -> pitch */ if (lf0 == LZERO) p = 0.0; else p = v->rate / exp(lf0); /* first time */ if (v->p1 < 0.0) { HTS_Vocoder_initialize_excitation(v, 0); if (v->stage == 0) { /* for MCP */ HTS_mc2b(spectrum, v->c, m, alpha); } else { /* for LSP */ if (v->use_log_gain) v->c[0] = LZERO; else v->c[0] = ZERO; for (i = 1; i <= m; i++) v->c[i] = i * PI / (m + 1); HTS_lsp2mgc(v, v->c, v->c, m, alpha); HTS_mc2b(v->c, v->c, m, alpha); HTS_gnorm(v->c, v->c, m, v->gamma); for (i = 1; i <= m; i++) v->c[i] *= v->gamma; } } HTS_Vocoder_start_excitation(v, p, 0); if (v->stage == 0) { /* for MCP */ HTS_Vocoder_postfilter_mcp(v, spectrum, m, alpha, beta); HTS_mc2b(spectrum, v->cc, m, alpha); for (i = 0; i <= m; i++) v->cinc[i] = (v->cc[i] - v->c[i]) * IPERIOD / v->fprd; } else { /* for LSP */ HTS_Vocoder_postfilter_lsp(v, spectrum, m, alpha, beta); HTS_check_lsp_stability(spectrum, m); HTS_lsp2mgc(v, spectrum, v->cc, m, alpha); HTS_mc2b(v->cc, v->cc, m, alpha); HTS_gnorm(v->cc, v->cc, m, v->gamma); for (i = 1; i <= m; i++) v->cc[i] *= v->gamma; for (i = 0; i <= m; i++) v->cinc[i] = (v->cc[i] - v->c[i]) * IPERIOD / v->fprd; } for (j = 0, i = (IPERIOD + 1) / 2; j < v->fprd; j++) { if (v->stage == 0) { /* for MCP */ if (v->p1 == 0.0) { x = HTS_white_noise(v); /* MIXED EXCITATION */ xnoise = x; xpulse = 0.0; } else { if ((v->pc += 1.0) >= v->p1) { x = sqrt(v->p1); v->pc = v->pc - v->p1; } else { x = 0.0; } /* MIXED EXCITATION */ xpulse = x; xnoise = HTS_mseq(v); /* ABY: plus or minus 1 */ } /* MIXED EXCITATION */ /* The real work -- apply shaping filters to pulse and noise */ fxpulse = fxnoise = 0.0; for (k = v_me->filter_order - 1; k > 0; k--) { fxpulse += v_me->hp[k] * v_me->xp_sig[k]; fxnoise += v_me->hn[k] * v_me->xn_sig[k]; v_me->xp_sig[k] = v_me->xp_sig[k-1]; v_me->xn_sig[k] = v_me->xn_sig[k-1]; } fxpulse += v_me->hp[0] * xpulse; fxnoise += v_me->hn[0] * xnoise; v_me->xp_sig[0] = xpulse; v_me->xn_sig[0] = xnoise; x = fxpulse + fxnoise; /* excitation is pulse plus noise */ x *= exp(v->c[0]); x = HTS_mlsadf(x, v->c, m, alpha, PADEORDER, v->d1); } else { /* for LSP */ if (!NGAIN) x *= v->c[0]; x = HTS_mglsadf(x, v->c, m, alpha, v->stage, v->d1); } x *= volume; /* output */ if (rawdata) rawdata[rawidx++] = x; if (audio) { if (x > 32767.0) xs = 32767; else if (x < -32768.0) xs = -32768; else xs = (short) x; HTS_Audio_write(audio, xs); } if (!--i) { for (i = 0; i <= m; i++) v->c[i] += v->cinc[i]; i = IPERIOD; } } HTS_Vocoder_end_excitation(v, nlpf); HTS_movem(v->cc, v->c, m + 1); } /* HTS_Vocoder_clear_me: clear vocoder (mixed excitation) */ void HTS_Vocoder_clear_me(HTS_Vocoder_ME * v_me) { HTS_Vocoder *v = v_me->v; /* access to original HTS_Vocoder structure */ if ((v_me != NULL) && (v != NULL)) { /* free buffer */ if (v->freqt_buff != NULL) { HTS_free(v->freqt_buff); v->freqt_buff = NULL; } v->freqt_size = 0; if (v->gc2gc_buff != NULL) { HTS_free(v->gc2gc_buff); v->gc2gc_buff = NULL; } v->gc2gc_size = 0; if (v->lsp2lpc_buff != NULL) { HTS_free(v->lsp2lpc_buff); v->lsp2lpc_buff = NULL; } v->lsp2lpc_size = 0; if (v->postfilter_buff != NULL) { HTS_free(v->postfilter_buff); v->postfilter_buff = NULL; } v->postfilter_size = 0; if (v->spectrum2en_buff != NULL) { HTS_free(v->spectrum2en_buff); v->spectrum2en_buff = NULL; } v->spectrum2en_size = 0; if (v->c != NULL) { HTS_free(v->c); v->c = NULL; } v->pulse_size = 0; if (v->pulse_buff != NULL){ HTS_free(v->pulse_buff); v->pulse_buff = NULL; } v_me->num_filters = 0; v_me->filter_order = 0; v_me->xp_sig = NULL; v_me->xn_sig = NULL; v_me->hp = NULL; v_me->hn = NULL; v_me->h = NULL; } } HTS_VOCODER_ME_C_END; #endif /* !HTS_VOCODER_ME_C */ festival/src/modules/clustergen/mlsa_resynthesis.h0000644000175000017500000001125012073372674021170 0ustar awbawb/*********************************************************************/ /* */ /* Nagoya Institute of Technology, Aichi, Japan, */ /* Nara Institute of Science and Technology, Nara, Japan */ /* and */ /* Carnegie Mellon University, Pittsburgh, PA */ /* Copyright (c) 2003-2004 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and */ /* distribute this software and its documentation without */ /* restriction, including without limitation the rights to use, */ /* copy, modify, merge, publish, distribute, sublicense, and/or */ /* sell copies of this work, and to permit persons to whom this */ /* work is furnished to do so, subject to the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list */ /* of conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* */ /* NAGOYA INSTITUTE OF TECHNOLOGY, NARA INSTITUTE OF SCIENCE AND */ /* TECHNOLOGY, CARNEGIE MELLON UNIVERSITY, AND THE CONTRIBUTORS TO */ /* THIS WORK DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, */ /* INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, */ /* IN NO EVENT SHALL NAGOYA INSTITUTE OF TECHNOLOGY, NARA */ /* INSTITUTE OF SCIENCE AND TECHNOLOGY, CARNEGIE MELLON UNIVERSITY, */ /* NOR THE CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, INDIRECT OR */ /* CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM */ /* LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, */ /* NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN */ /* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ /* */ /*********************************************************************/ /* */ /* Author : Tomoki Toda (tomoki@ics.nitech.ac.jp) */ /* Date : June 2004 */ /* */ /* Modified by Alan W Black (awb@cs.cmu.edu) Jan 2006 */ /* taken from festvox/src/vc/ back into Festival */ /*-------------------------------------------------------------------*/ /* */ /* Subroutine for Speech Synthesis */ /* */ /*-------------------------------------------------------------------*/ #ifndef __MLSA_RESYNTHESIS_H #define __MLSA_RESYNTHESIS_H typedef struct DVECTOR_STRUCT { long length; double *data; double *imag; } *DVECTOR; typedef struct DMATRIX_STRUCT { long row; long col; double **data; double **imag; } *DMATRIX; #define XBOOL int #define XTRUE 1 #define XFALSE 0 #define NODATA NULL #define FABS(x) ((x) >= 0.0 ? (x) : -(x)) #define MAX(a, b) ((a) > (b) ? (a) : (b)) DVECTOR synthesis_body(DMATRIX mcep, DVECTOR f0v, EST_Track *str, EST_Track *filtrack, double fs, double framem, double alpha, double beta); #define RANDMAX 32767 #define B0 0x00000001 #define B28 0x10000000 #define B31 0x80000000 #define B31_ 0x7fffffff #define Z 0x00000000 typedef enum {MFALSE, MTRUE} Boolean; typedef struct _VocoderSetup { int fprd; int iprd; int seed; int pd; unsigned long next; Boolean gauss; double p1; double pc; double pj; double pade[21]; double *ppade; double *c, *cc, *cinc, *d1; double rate; int sw; double r1, r2, s; int x; /* for postfiltering */ int size; double *d; double *g; double *mc; double *cep; double *ir; int o; int irleng; /* for MIXED EXCITATION */ int ME_order; int ME_num; double *hpulse; double *hnoise; double *xpulsesig; double *xnoisesig; double **h; } VocoderSetup; #endif /* __RESYNTHESIS_SUB_H */ festival/src/modules/clustergen/HTS_vocoder_me.h0000644000175000017500000001253612073155232020431 0ustar awbawb/************************************************************************************/ /* Copyright (c) 2012 The Department of Arts and Culture, */ /* The Government of the Republic of South Africa. */ /* */ /* Contributors: Meraka Institute, CSIR, South Africa. */ /* */ /* Permission is hereby granted, free of charge, to any person obtaining a copy */ /* of this software and associated documentation files (the "Software"), to deal */ /* in the Software without restriction, including without limitation the rights */ /* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell */ /* copies of the Software, and to permit persons to whom the Software is */ /* furnished to do so, subject to the following conditions: */ /* The above copyright notice and this permission notice shall be included in */ /* all copies or substantial portions of the Software. */ /* */ /* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR */ /* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, */ /* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE */ /* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER */ /* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, */ /* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN */ /* THE SOFTWARE. */ /* */ /************************************************************************************/ /* */ /* AUTHOR : Aby Louw */ /* DATE : 14 May 2012 */ /* */ /************************************************************************************/ /* */ /* Added mixed excitation, see code at end of file. */ /* */ /************************************************************************************/ #ifndef HTS_VOCODER_ME_H #define HTS_VOCODER_ME_H #ifdef __cplusplus #define HTS_VOCODER_ME_H_START extern "C" { #define HTS_VOCODER_ME_H_END } #else #define HTS_VOCODER_ME_H_START #define HTS_VOCODER_ME_H_END #endif /* __CPLUSPLUS */ HTS_VOCODER_ME_H_START; /* hts_engine libraries */ #include /* HTS_Vocoder_ME: structure for setting of vocoder (mixed excitation) */ typedef struct _HTS_Vocoder_ME { HTS_Vocoder *v; /* hts engine vocoder structure */ /* mixed excitation */ double *xp_sig; /* pulse signal, the size of this should be the filter order */ double *xn_sig; /* noise signal, the size of this should be the filter order */ double *hp; /* pulse shaping filter, size of the filter order */ double *hn; /* noise shaping filter, size of the filter order */ int num_filters; /* number of filters */ int filter_order; /* filter order */ double **h; /* filter coefficients */ } HTS_Vocoder_ME; /* HTS_Vocoder_initialize_me: initialize vocoder (mixed excitation) */ void HTS_Vocoder_initialize_me(HTS_Vocoder_ME * v_me, const int m, const int stage, HTS_Boolean use_log_gain, const int rate, const int fperiod, int num_filters, int filter_order, double **h, double *xp_sig, double *xn_sig, double *hp, double *hn); /* HTS_Vocoder_synthesize_me: mixed excitation and MLSA/MGLSA filster based waveform synthesis */ void HTS_Vocoder_synthesize_me(HTS_Vocoder_ME * v_me, const int m, double lf0, double *spectrum, double *strengths, size_t nlpf, double* lpf, double alpha, double beta, double volume, double *rawdata, HTS_Audio * audio); /* HTS_Vocoder_clear_me: clear vocoder (mixed excitation) */ void HTS_Vocoder_clear_me(HTS_Vocoder_ME * v_me); HTS_VOCODER_ME_H_END; #endif /* !HTS_VOCODER_ME_H */ festival/src/modules/clustergen/vc.h0000644000175000017500000001017411332061167016175 0ustar awbawb/*********************************************************************/ /* */ /* Nagoya Institute of Technology, Aichi, Japan, */ /* Nara Institute of Science and Technology, Nara, Japan */ /* and */ /* Carnegie Mellon University, Pittsburgh, PA */ /* Copyright (c) 2003-2004 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and */ /* distribute this software and its documentation without */ /* restriction, including without limitation the rights to use, */ /* copy, modify, merge, publish, distribute, sublicense, and/or */ /* sell copies of this work, and to permit persons to whom this */ /* work is furnished to do so, subject to the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list */ /* of conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* */ /* NAGOYA INSTITUTE OF TECHNOLOGY, NARA INSTITUTE OF SCIENCE AND */ /* TECHNOLOGY, CARNEGIE MELLON UNIVERSITY, AND THE CONTRIBUTORS TO */ /* THIS WORK DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, */ /* INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, */ /* IN NO EVENT SHALL NAGOYA INSTITUTE OF TECHNOLOGY, NARA */ /* INSTITUTE OF SCIENCE AND TECHNOLOGY, CARNEGIE MELLON UNIVERSITY, */ /* NOR THE CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, INDIRECT OR */ /* CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM */ /* LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, */ /* NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN */ /* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ /* */ /*********************************************************************/ /* */ /* Author : Tomoki Toda (tomoki@ics.nitech.ac.jp) */ /* Date : June 2004 */ /* */ /* Functions shared between mlpg and mlsa */ /*-------------------------------------------------------------------*/ #ifndef __CST_VC_H #define __CST_VC_H typedef struct LVECTOR_STRUCT { long length; long *data; long *imag; } *LVECTOR; typedef struct DVECTOR_STRUCT { long length; double *data; double *imag; } *DVECTOR; typedef struct DMATRIX_STRUCT { long row; long col; double **data; double **imag; } *DMATRIX; #define XBOOL int #define XTRUE 1 #define XFALSE 0 #define NODATA NULL #define FABS(x) ((x) >= 0.0 ? (x) : -(x)) #define LABS(x) ((x) >= 0 ? (x) : -(x)) #define MAX(a, b) ((a) > (b) ? (a) : (b)) #define xdvnull() xdvalloc(0) #define xdvnums(length, value) xdvinit((double)(value), 0.0, (double)(length)) #define xdvzeros(length) xdvnums(length, 0.0) LVECTOR xlvalloc(long length); void xlvfree(LVECTOR x); DVECTOR xdvalloc(long length); DVECTOR xdvcut(DVECTOR x, long offset, long length); void xdvfree(DVECTOR vector); double dvmax(DVECTOR x, long *index); double dvmin(DVECTOR x, long *index); DMATRIX xdmalloc(long row, long col); void xdmfree(DMATRIX matrix); DVECTOR xdvinit(double j, double incr, double n); double dvsum(DVECTOR x); #define RANDMAX 32767 #define B0 0x00000001 #define B28 0x10000000 #define B31 0x80000000 #define B31_ 0x7fffffff #define Z 0x00000000 typedef enum {MFALSE, MTRUE} Boolean; #endif /* __CST_VC_H */ festival/src/modules/clustergen/simple_mlpg.h0000644000175000017500000002311211332061167020071 0ustar awbawb/* --------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis System (HTS): version 1.1.1 */ /* HTS Working Group */ /* */ /* Department of Computer Science */ /* Nagoya Institute of Technology */ /* and */ /* Interdisciplinary Graduate School of Science and Engineering */ /* Tokyo Institute of Technology */ /* Copyright (c) 2001-2003 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and */ /* distribute this software and its documentation without */ /* restriction, including without limitation the rights to use, */ /* copy, modify, merge, publish, distribute, sublicense, and/or */ /* sell copies of this work, and to permit persons to whom this */ /* work is furnished to do so, subject to the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list */ /* of conditions and the following disclaimer. */ /* */ /* 2. Any modifications must be clearly marked as such. */ /* */ /* NAGOYA INSTITUTE OF TECHNOLOGY, TOKYO INSITITUTE OF TECHNOLOGY, */ /* HTS WORKING GROUP, AND THE CONTRIBUTORS TO THIS WORK DISCLAIM */ /* ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL */ /* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL NAGOYA INSTITUTE OF TECHNOLOGY, TOKYO INSITITUTE OF */ /* TECHNOLOGY, HTS WORKING GROUP, NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY */ /* DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, */ /* WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS */ /* ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR */ /* PERFORMANCE OF THIS SOFTWARE. */ /* */ /* --------------------------------------------------------------- */ /* mlpg.c : speech parameter generation from pdf sequence */ /* */ /* 2003/12/26 by Heiga Zen */ /* --------------------------------------------------------------- */ /*********************************************************************/ /* */ /* Nagoya Institute of Technology, Aichi, Japan, */ /* and */ /* Carnegie Mellon University, Pittsburgh, PA */ /* Copyright (c) 2003-2004 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and */ /* distribute this software and its documentation without */ /* restriction, including without limitation the rights to use, */ /* copy, modify, merge, publish, distribute, sublicense, and/or */ /* sell copies of this work, and to permit persons to whom this */ /* work is furnished to do so, subject to the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list */ /* of conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* */ /* NAGOYA INSTITUTE OF TECHNOLOGY, CARNEGIE MELLON UNIVERSITY, AND */ /* THE CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH */ /* REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL NAGOYA INSTITUTE */ /* OF TECHNOLOGY, CARNEGIE MELLON UNIVERSITY, NOR THE CONTRIBUTORS */ /* BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR */ /* ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR */ /* PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER */ /* TORTUOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE */ /* OR PERFORMANCE OF THIS SOFTWARE. */ /* */ /*********************************************************************/ /* */ /* ML-Based Parameter Generation */ /* 2003/12/26 by Heiga Zen */ /* */ /* Basic functions are extracted from HTS and */ /* modified by Tomoki Toda (tomoki@ics.nitech.ac.jp) */ /* June 2004 */ /* Integrate as a Voice Conversion module */ /* */ /*-------------------------------------------------------------------*/ /* */ /* Author : Tomoki Toda (tomoki@ics.nitech.ac.jp) */ /* Date : June 2004 */ /* */ /*-------------------------------------------------------------------*/ /* Integrated into a single C file for festival/flite integration */ /* May 2008 awb@cs.cmu.edu */ /*-------------------------------------------------------------------*/ #ifndef _MLPG_H #define _MLPG_H #include "vc.h" #define INFTY ((double) 1.0e+38) #define INFTY2 ((double) 1.0e+19) #define INVINF ((double) 1.0e-38) #define INVINF2 ((double) 1.0e-19) #ifdef PI /* ok */ #elif defined(M_PI) #define PI M_PI #else #define PI 3.1415926535897932385 #endif #define WLEFT 0 #define WRIGHT 1 typedef struct _DWin { int num; /* number of static + deltas */ int **width; /* width [0..num-1][0(left) 1(right)] */ double **coef; /* coefficient [0..num-1][length[0]..length[1]] */ double **coef_ptrs; /* keeps the pointers so we can free them */ int maxw[2]; /* max width [0(left) 1(right)] */ } MLPG_DWin; typedef struct _PStreamChol { int vSize; // size of ovserved vector int order; // order of cepstrum int T; // number of frames int width; // width of WSW MLPG_DWin dw; double **mseq; // sequence of mean vector double **ivseq; // sequence of invarsed covariance vector double ***ifvseq; // sequence of invarsed full covariance vector double **R; // WSW[T][range] double *r; // WSM [T] double *g; // g [T] double **c; // parameter c } PStreamChol; typedef struct MLPGPARA_STRUCT { DVECTOR ov; DVECTOR iuv; DVECTOR iumv; DVECTOR flkv; DMATRIX stm; DMATRIX dltm; DMATRIX pdf; DVECTOR detvec; DMATRIX wght; DMATRIX mean; DMATRIX cov; LVECTOR clsidxv; DVECTOR clsdetv; DMATRIX clscov; double vdet; DVECTOR vm; DVECTOR vv; DVECTOR var; } *MLPGPARA; #if 0 static double get_like_gv(long dim2, long dnum, MLPGPARA param); static void sm_mvav(DMATRIX mat, long hlen); #endif #if 0 /* Full Covariance Version */ static void InitPStreamCholFC(PStreamChol *pst, char *dynwinf, char *accwinf, int order, int T); static void mlgparaCholFC(DMATRIX pdf, PStreamChol *pst, DMATRIX mlgp); static void mlpgCholFC(PStreamChol *pst); static void calc_R_and_r_FC(PStreamChol *pst); static void CholeskiFC(PStreamChol *pst); static void Choleski_forwardFC(PStreamChol *pst); static void Choleski_backwardFC(PStreamChol *pst); #endif /**********************************/ /* ML Considering Global Variance */ /**********************************/ #if 0 static void varconv(double **c, const int m, const int T, const double var); static void calc_varstats(double **c, const int m, const int T, double *av, double *var, double *dif); /* Diagonal Covariance Version */ static void mlgparaGrad(DMATRIX pdf, PStreamChol *pst, DMATRIX mlgp, const int max, double th, double e, double alpha, DVECTOR vm, DVECTOR vv, XBOOL nrmflag, XBOOL extvflag); static void mlpgGrad(PStreamChol *pst, const int max, double th, double e, double alpha, DVECTOR vm, DVECTOR vv, XBOOL nrmflag); static void calc_grad(PStreamChol *pst, const int m); static void calc_vargrad(PStreamChol *pst, const int m, double alpha, double n, double vm, double vv); static double get_gauss_dia5(double det, double weight, DVECTOR vec, // dim DVECTOR meanvec, // dim DVECTOR invcovvec); // dim #endif #if 0 static void get_gv_mlpgpara(MLPGPARA param, char *vmfile, char *vvfile, long dim2, XBOOL msg_flag); #endif #endif /* _MLPG_H */ festival/src/modules/java/Makefile0000644000175000017500000000513107262605650015627 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../../.. DIRNAME=src/modules/java BUILD_DIRS = cstr NEED_JAVA=1 FILES = Makefile java.mak java_cpp.mak java_media.mak jsapi.mak NoInit ALL = .sub_directories .javalib include $(TOP)/config/common_make_rules festival/src/modules/java/java.mak0000644000175000017500000000716707262605650015615 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Make rules for Java module ## ## ## ########################################################################### INCLUDE_JAVA=1 MOD_DESC_JAVA=Festival Java Support # PROJECT_LIBRARIES := festjava $(PROJECT_LIBRARIES) PROJECT_LIBRARY_DIR_festjava = $(TOP)/src/lib # PROJECT_SHARED_LIBRARIES := festjava $(PROJECT_SHARED_LIBRARIES) EST_JAVA_VERSION=basic ifeq ($(DIRNAME),src/modules) LIB_BUILD_DIRS := java $(LIB_BUILD_DIRS) endif ifeq ($(DIRNAME),src/scripts) SCRIPTS := festival_client_java.sh $(SCRIPTS) endif include $(EST)/config/compilers/$(JAVA_COMPILER).mak ifndef SHARED .config_error:: FORCE @echo "+--------------------------------------------------" @echo "| Must compile speech tools shared for Java support" @echo "+--------------------------------------------------" @exit 1 endif festival/src/modules/java/java_cpp.mak0000644000175000017500000000710307262605650016445 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Make rules for Java module ## ## ## ########################################################################### INCLUDE_JAVA_CPP=1 MOD_DESC_JAVA_CPP=Festival Java Support with C++ Interface # PROJECT_LIBRARIES := festjava $(PROJECT_LIBRARIES) # PROJECT_LIBRARY_DIR_festjava = $(TOP)/src/lib # PROJECT_SHARED_LIBRARIES := festjava $(PROJECT_SHARED_LIBRARIES) EST_JAVA_VERSION=cpp # ifeq ($(DIRNAME),src/modules) # LIB_BUILD_DIRS := java $(LIB_BUILD_DIRS) # endif include $(EST)/config/compilers/$(JAVA_COMPILER).mak ifndef INCLUDE_JAVA .config_error:: FORCE @echo "+--------------------------------------------------" @echo "| Must include module JAVA to use JAVA_CPP " @echo "+--------------------------------------------------" @exit 1 endif festival/src/modules/java/java_media.mak0000644000175000017500000000735407262605650016752 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Make rules for Java module using the media API version ## ## ## ########################################################################### INCLUDE_JAVA_MEDIA=1 MOD_DESC_JAVA_MEDIA=Festival Java Support using Java Media API # PROJECT_LIBRARIES := festjava $(PROJECT_LIBRARIES) # PROJECT_LIBRARY_DIR_festjava = $(TOP)/src/lib # PROJECT_SHARED_LIBRARIES := festjava $(PROJECT_SHARED_LIBRARIES) EST_JAVA_VERSION=media include $(EST)/config/compilers/$(JAVA_COMPILER).mak SYSTEM_JAVA_CLASSPATH:=$(SYSTEM_JAVA_MEDIA_CLASSPATH):$(SYSTEM_JAVA_CLASSPATH) SYSTEM_LD_LIBRARY_PATH:=$(SYSTEM_JAVA_MEDIA_LIBRARY):$(SYSTEM_LD_LIBRARY_PATH) # ifeq ($(DIRNAME),src/modules) # LIB_BUILD_DIRS := java $(LIB_BUILD_DIRS) # endif ifndef INCLUDE_JAVA .config_error:: FORCE @echo "+--------------------------------------------------" @echo "| Must include module JAVA to use JAVA_MEDIA " @echo "+--------------------------------------------------" @exit 1 endif festival/src/modules/java/jsapi.mak0000644000175000017500000000764307262605650016001 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: Wed May 27 1998 ## ## -------------------------------------------------------------------- ## ## Make rules for Java module ## ## ## ########################################################################### INCLUDE_JSAPI=1 MOD_DESC_JSAPI=Preliminary JSAPI Support ifeq ($(DIRNAME),src/modules/java/cstr/testPrograms) JAVA_CLASSES += SayHelloWorld endif ifeq ($(DIRNAME),src/modules/java/cstr/festival) BUILD_DIRS := jsapi $(BUILD_DIRS) endif ifeq ($(DIRNAME),src/scripts) SCRIPTS := jsapi_example.sh $(SCRIPTS) endif ifdef JSAPI_HOME SYSTEM_JAVA_CLASSPATH := $(JSAPI_HOME)/lib/jsapi.jar:$(SYSTEM_JAVA_CLASSPATH) else .config_error:: FORCE @echo "+--------------------------------------------------" @echo "| Please set JSAPI_HOME in config or environment @echo "+--------------------------------------------------" @exit 1 endif ifdef JSUTILS_HOME SYSTEM_JAVA_CLASSPATH := $(JSUTILS_HOME)/lib/jsutilsall.jar:$(SYSTEM_JAVA_CLASSPATH) endif ifndef INCLUDE_JAVA .config_error:: FORCE @echo "+--------------------------------------------------" @echo "| Must compile Java support to get JSAPI support @echo "+--------------------------------------------------" @exit 1 endif festival/src/modules/java/NoInit0000644000175000017500000000000107262605650015301 0ustar awbawb festival/src/modules/java/cstr/Makefile0000644000175000017500000000506607262605650016611 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../../../.. DIRNAME=src/modules/java/cstr BUILD_DIRS = festival testPrograms NEED_JAVA=1 FILES = Makefile ALL = .sub_directories include $(TOP)/config/common_make_rules festival/src/modules/java/cstr/festival/Makefile0000644000175000017500000000521307262605650020420 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../../../../.. DIRNAME=src/modules/java/cstr/festival BUILD_DIRS = client scheme ALL_DIRS = $(BUILD_DIRS) jsapi JAVA_CLASSES = Client NEED_JAVA=1 FILES = Makefile $(JAVA_CLASSES:=.java) ALL = .java .sub_directories include $(TOP)/config/common_make_rules festival/src/modules/java/cstr/festival/Client.java0000644000175000017500000001546407262605650021052 0ustar awbawb //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Centre for Speech Technology Research \\ // University of Edinburgh, UK \\ // Copyright (c) 1996,1997 \\ // All Rights Reserved. \\ // Permission is hereby granted, free of charge, to use and distribute \\ // this software and its documentation without restriction, including \\ // without limitation the rights to use, copy, modify, merge, publish, \\ // distribute, sublicense, and/or sell copies of this work, and to \\ // permit persons to whom this work is furnished to do so, subject to \\ // the following conditions: \\ // 1. The code must retain the above copyright notice, this list of \\ // conditions and the following disclaimer. \\ // 2. Any modifications must be clearly marked as such. \\ // 3. Original authors' names are not deleted. \\ // 4. The authors' names are not used to endorse or promote products \\ // derived from this software without specific prior written \\ // permission. \\ // THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK \\ // DISCLAIM ALL WARRANTIES With REGARD TO THIS SOFTWARE, INCLUDING \\ // ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT \\ // SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE \\ // FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES \\ // WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN \\ // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, \\ // ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF \\ // THIS SOFTWARE. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Author: Richard Caley (rjc@cstr.ed.ac.uk) \\ // -------------------------------------------------------------------- \\ // Mainline for Java festival client. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ package cstr.festival ; import java.lang.*; import java.util.*; import java.awt.*; import java.io.*; import java.net.*; import cstr.festival.client.*; import cstr.festival.scheme.*; class RequestHandler implements RequestListener { Thread mainline; public RequestHandler(Thread main) { mainline=main; } public void requestRunning(Request r) { System.out.println("Request "+r.command+" running"); } public void requestResult(Request r, Object res) { System.out.println("Request "+r.command+" result="+res); } public void requestError(Request r, String mes) { System.out.println("Request "+r.command+" error "+mes); } public void requestFinished(Request r) { System.out.println("Request "+r.command+" finished"); } } public class Client { public static void useage(String error) { if (error != null) { System.out.println(""); System.out.println(" "+error); } System.out.println(""); System.out.println("Useage: festival_client_java [options] file..."); System.out.println(""); System.out.println(" --help Show this message."); System.out.println(" --server Connect to named host."); System.out.println(" --port Use the given port."); System.out.println(" --sync Wait for each expression to finish"); System.out.println(" executing before reading next."); System.out.println(" --wait Wait for all expressions to finish"); System.out.println(" executing before exiting."); System.exit(error!=null?1:0); } public static void main (String[] args) { String server="localhost"; int port=1314; int i; boolean waitAtEnd=false; boolean sync=false; for(i=0; i=args.length) useage("Need name after --server"); server = args[i]; } else if (args[i].equals("--port")) { i++; if (i>=args.length) useage("Need name after --server"); try { port = Integer.parseInt(args[i]); } catch (NumberFormatException ex) { useage("Not a valid port number '" + args[i] + "'"); } } else if (args[i].equals("--wait")) waitAtEnd=true; else if (args[i].equals("--sync")) sync=true; else if (args[i].startsWith("--")) useage("Unknown argument "+args[i]); else break; } System.out.println("Server='"+server+":"+port+"'"); Session s=null; try { s = new Session(server, port); } catch (UnknownHostException ex) { useage("Unknown host '"+ex.getMessage()+"'"); } catch (IOException ex) { useage("Can't connect '"+ex.getMessage()+"'"); } s.initialise(); System.out.println("Connected"); if (i == args.length) { args = new String[] { "-" }; i=0; } RequestListener handler = new RequestHandler(Thread.currentThread()); file: for(; i "); System.out.flush(); } exp=sreader.nextExprString(); if (exp==null) break; System.out.println(": "+exp); Request r = s.request(exp, handler); if (sync) while (!r.isFinished()) r.waitForUpdate(); } } catch (IOException ioex) { System.out.println(" Error "+ioex.getMessage()); break file; } System.out.println("--EOF--"); } System.out.println("call terminate"); s.terminate(waitAtEnd); } }; festival/src/modules/java/cstr/festival/client/Makefile0000644000175000017500000000520707262605650021701 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../../../../../.. DIRNAME=src/modules/java/cstr/festival/client JAVA_CLASSES = Session Festival JobQueue Request RequestListener NEED_JAVA=1 FILES = Makefile $(JAVA_CLASSES:=.java) ALL = .java .sub_directories include $(TOP)/config/common_make_rules festival/src/modules/java/cstr/festival/client/Session.java0000644000175000017500000001271507262605650022531 0ustar awbawb //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Centre for Speech Technology Research \\ // University of Edinburgh, UK \\ // Copyright (c) 1996,1997 \\ // All Rights Reserved. \\ // Permission is hereby granted, free of charge, to use and distribute \\ // this software and its documentation without restriction, including \\ // without limitation the rights to use, copy, modify, merge, publish, \\ // distribute, sublicense, and/or sell copies of this work, and to \\ // permit persons to whom this work is furnished to do so, subject to \\ // the following conditions: \\ // 1. The code must retain the above copyright notice, this list of \\ // conditions and the following disclaimer. \\ // 2. Any modifications must be clearly marked as such. \\ // 3. Original authors' names are not deleted. \\ // 4. The authors' names are not used to endorse or promote products \\ // derived from this software without specific prior written \\ // permission. \\ // THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK \\ // DISCLAIM ALL WARRANTIES With REGARD TO THIS SOFTWARE, INCLUDING \\ // ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT \\ // SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE \\ // FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES \\ // WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN \\ // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, \\ // ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF \\ // THIS SOFTWARE. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Author: Richard Caley (rjc@cstr.ed.ac.uk) \\ // -------------------------------------------------------------------- \\ // Objects representing sessions on festival servers. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ package cstr.festival.client ; import java.lang.*; import java.util.*; import java.awt.*; import java.io.*; import java.net.*; import cstr.est.*; class RequestRecorder implements RequestListener { Object result; public RequestRecorder() { } public void requestRunning(Request r) { } public void requestResult(Request r, Object res) { result=res; } public void requestError(Request r, String mes) { } public void requestFinished(Request r) { } } public class Session { protected Festival festival; private int lastID=0; protected Hashtable requests; public Session(Socket sock) { festival = new Festival(sock); requests = new Hashtable(10); } public Session(InetAddress addr, int p) throws IOException { festival = new Festival(addr, p); requests = new Hashtable(10); } public Session(String host, int p) throws IOException, UnknownHostException { festival = new Festival(InetAddress.getByName(host), p); requests = new Hashtable(10); } protected void finalize() { terminate(false); } public void initialise() { festival.connect(); } public void terminate(boolean carefully) { festival.disconnect(carefully); } public Request request(String c, RequestListener l) { Integer id = new Integer(newID()); Request r = new Request(c,id, this); if (l!= null) r.addRequestListener(l); requests.put(id, r); festival.newJob(id, c, this); return r; } public Object synchronousRequest(String c) { RequestRecorder l = new RequestRecorder(); Request r = request(c, l); while (!r.isFinished()) r.waitForUpdate(); return l.result; } public void notifyRunning(Integer id) { Request r = (Request)requests.get(id); r.notifyRunning(); } public void notifyError(Integer id, String message) { Request r = (Request)requests.get(id); r.notifyError(message); } public void notifyResult(Integer id, int type, byte [] result ) { Request r = (Request)requests.get(id); Object oRes = result; if (type == Festival.FT_WAVE) try { Wave wv = new Wave(result); oRes = wv; } catch (UnsupportedEncodingException ex) { oRes = ex; } else if (type == Festival.FT_SCHEME) { oRes = new String(result); } r.notifyResult(oRes); } public void notifyFinished(Integer id) { Request r = (Request)requests.get(id); // it's finished, so we can forget about it. requests.remove(id); r.notifyFinished(); } private int newID() { return ++lastID; } } festival/src/modules/java/cstr/festival/client/Festival.java0000644000175000017500000002120507262605650022655 0ustar awbawb //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Centre for Speech Technology Research \\ // University of Edinburgh, UK \\ // Copyright (c) 1996,1997 \\ // All Rights Reserved. \\ // Permission is hereby granted, free of charge, to use and distribute \\ // this software and its documentation without restriction, including \\ // without limitation the rights to use, copy, modify, merge, publish, \\ // distribute, sublicense, and/or sell copies of this work, and to \\ // permit persons to whom this work is furnished to do so, subject to \\ // the following conditions: \\ // 1. The code must retain the above copyright notice, this list of \\ // conditions and the following disclaimer. \\ // 2. Any modifications must be clearly marked as such. \\ // 3. Original authors' names are not deleted. \\ // 4. The authors' names are not used to endorse or promote products \\ // derived from this software without specific prior written \\ // permission. \\ // THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK \\ // DISCLAIM ALL WARRANTIES With REGARD TO THIS SOFTWARE, INCLUDING \\ // ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT \\ // SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE \\ // FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES \\ // WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN \\ // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, \\ // ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF \\ // THIS SOFTWARE. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Author: Richard Caley (rjc@cstr.ed.ac.uk) \\ // -------------------------------------------------------------------- \\ // The thread which actually communicates with festival. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ package cstr.festival.client ; import java.lang.*; import java.util.*; import java.awt.*; import java.io.*; import java.net.*; import cstr.est.*; class Job { Integer id; String command; Session session; public Job(Integer i, String c, Session s) { id=i; command=c; session=s; } } class Festival extends Thread { public static final int FT_SCHEME = 1; public static final int FT_WAVE = 2; protected static byte [] endm; protected Socket s; protected String hostname=null; protected InetAddress address = null; protected int port=-1; protected boolean closeOnExit; protected PrintWriter out; protected InputStream in; protected boolean ok; protected JobQueue jobs; private byte buffer[]; private int buffered_p; private int buffered_e; static { String end="ft_StUfF_key"; endm = new byte[end.length()]; for(int i=0;i=0) { val.append(line.substring(0, endm)); pending = line.substring(endm+12); break; } else { val.append(line); val.append("\n"); } line = in.readLine(); } return val.toString(); */ } } festival/src/modules/java/cstr/festival/client/JobQueue.java0000644000175000017500000000723507262605650022626 0ustar awbawb //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Centre for Speech Technology Research \\ // University of Edinburgh, UK \\ // Copyright (c) 1996,1997 \\ // All Rights Reserved. \\ // Permission is hereby granted, free of charge, to use and distribute \\ // this software and its documentation without restriction, including \\ // without limitation the rights to use, copy, modify, merge, publish, \\ // distribute, sublicense, and/or sell copies of this work, and to \\ // permit persons to whom this work is furnished to do so, subject to \\ // the following conditions: \\ // 1. The code must retain the above copyright notice, this list of \\ // conditions and the following disclaimer. \\ // 2. Any modifications must be clearly marked as such. \\ // 3. Original authors' names are not deleted. \\ // 4. The authors' names are not used to endorse or promote products \\ // derived from this software without specific prior written \\ // permission. \\ // THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK \\ // DISCLAIM ALL WARRANTIES With REGARD TO THIS SOFTWARE, INCLUDING \\ // ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT \\ // SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE \\ // FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES \\ // WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN \\ // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, \\ // ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF \\ // THIS SOFTWARE. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Author: Richard Caley (rjc@cstr.ed.ac.uk) \\ // -------------------------------------------------------------------- \\ // A thread safe queue based on Vector. Could be faster. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ package cstr.festival.client ; import java.lang.*; import java.util.*; import java.awt.*; public class JobQueue { Vector q; public JobQueue(int space) { q = new Vector(space); } public JobQueue() { q = new Vector(); } public synchronized void add(Object thing) { q.addElement(thing); } public synchronized void remove(Object thing) { q.removeElement(thing); } public synchronized boolean isEmpty() { return q.isEmpty(); } public synchronized Object top() { Object o; if (q.isEmpty()) o = null; else o = q.firstElement(); return o; } public synchronized Object get() { Object o; if (q.isEmpty()) o = null; else { o = q.firstElement(); q.removeElementAt(0); } return o; } public synchronized Enumeration elements() { return q.elements(); } } festival/src/modules/java/cstr/festival/client/Request.java0000644000175000017500000001110207262605650022523 0ustar awbawb //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Centre for Speech Technology Research \\ // University of Edinburgh, UK \\ // Copyright (c) 1996,1997 \\ // All Rights Reserved. \\ // Permission is hereby granted, free of charge, to use and distribute \\ // this software and its documentation without restriction, including \\ // without limitation the rights to use, copy, modify, merge, publish, \\ // distribute, sublicense, and/or sell copies of this work, and to \\ // permit persons to whom this work is furnished to do so, subject to \\ // the following conditions: \\ // 1. The code must retain the above copyright notice, this list of \\ // conditions and the following disclaimer. \\ // 2. Any modifications must be clearly marked as such. \\ // 3. Original authors' names are not deleted. \\ // 4. The authors' names are not used to endorse or promote products \\ // derived from this software without specific prior written \\ // permission. \\ // THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK \\ // DISCLAIM ALL WARRANTIES With REGARD TO THIS SOFTWARE, INCLUDING \\ // ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT \\ // SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE \\ // FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES \\ // WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN \\ // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, \\ // ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF \\ // THIS SOFTWARE. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Author: Richard Caley (rjc@cstr.ed.ac.uk) \\ // -------------------------------------------------------------------- \\ // Record of a request made to a session. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ package cstr.festival.client ; import java.lang.*; import java.util.*; import java.awt.*; public class Request { public String command; public Integer id; public Session session; private Vector results; protected boolean running=false; protected String error=null; protected boolean finished=false; private RequestListener listener; public Request(String c, Integer i, Session s) { command=c; id=i; session=s; results= new Vector(1); } public synchronized boolean isFinished() { return finished; } public synchronized void waitForUpdate() { try { wait(); } catch (InterruptedException e) { } // System.out.println("wait done"); } public synchronized void notifyResult(Object r) { results.addElement(r); if (listener != null) listener.requestResult(this, r); notifyAll(); } public synchronized void notifyRunning() { running = true; if (listener != null) listener.requestRunning(this); notifyAll(); } public synchronized void notifyError(String message) { error = message; if (listener != null) listener.requestError(this, message); notifyAll(); } public synchronized void notifyFinished() { finished = true; if (listener != null) listener.requestFinished(this); notifyAll(); } public synchronized void addRequestListener(RequestListener l) { listener = l; if (running) l.requestRunning(this); for(int i=0; i < results.size(); i++) l.requestResult(this, results.elementAt(i)); if (error != null) l.requestError(this, error); if (finished) l.requestFinished(this); } public synchronized void removeRequestListener(RequestListener l) { if (listener==l) listener = null; } } festival/src/modules/java/cstr/festival/client/RequestListener.java0000644000175000017500000000613107262605650024237 0ustar awbawb //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Centre for Speech Technology Research \\ // University of Edinburgh, UK \\ // Copyright (c) 1996,1997 \\ // All Rights Reserved. \\ // Permission is hereby granted, free of charge, to use and distribute \\ // this software and its documentation without restriction, including \\ // without limitation the rights to use, copy, modify, merge, publish, \\ // distribute, sublicense, and/or sell copies of this work, and to \\ // permit persons to whom this work is furnished to do so, subject to \\ // the following conditions: \\ // 1. The code must retain the above copyright notice, this list of \\ // conditions and the following disclaimer. \\ // 2. Any modifications must be clearly marked as such. \\ // 3. Original authors' names are not deleted. \\ // 4. The authors' names are not used to endorse or promote products \\ // derived from this software without specific prior written \\ // permission. \\ // THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK \\ // DISCLAIM ALL WARRANTIES With REGARD TO THIS SOFTWARE, INCLUDING \\ // ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT \\ // SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE \\ // FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES \\ // WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN \\ // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, \\ // ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF \\ // THIS SOFTWARE. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Author: Richard Caley (rjc@cstr.ed.ac.uk) \\ // -------------------------------------------------------------------- \\ // Interface for things which keep track of requests to the sewrver. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ package cstr.festival.client ; import java.lang.*; import java.util.*; import java.awt.*; public interface RequestListener { public abstract void requestRunning(Request r); public abstract void requestResult(Request r, Object res); public abstract void requestError(Request r, String mes); public abstract void requestFinished(Request r); } festival/src/modules/java/cstr/festival/scheme/Makefile0000644000175000017500000000517007262605650021666 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../../../../../.. DIRNAME=src/modules/java/cstr/festival/scheme JAVA_CLASSES = SchemeTokenizer SchemeReader ReflectingSchemeReader NEED_JAVA=1 FILES = Makefile $(JAVA_CLASSES:=.java) ALL = .java include $(TOP)/config/common_make_rules festival/src/modules/java/cstr/festival/scheme/SchemeTokenizer.java0000644000175000017500000001155707262605650024176 0ustar awbawb //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Centre for Speech Technology Research \\ // University of Edinburgh, UK \\ // Copyright (c) 1996,1997 \\ // All Rights Reserved. \\ // Permission is hereby granted, free of charge, to use and distribute \\ // this software and its documentation without restriction, including \\ // without limitation the rights to use, copy, modify, merge, publish, \\ // distribute, sublicense, and/or sell copies of this work, and to \\ // permit persons to whom this work is furnished to do so, subject to \\ // the following conditions: \\ // 1. The code must retain the above copyright notice, this list of \\ // conditions and the following disclaimer. \\ // 2. Any modifications must be clearly marked as such. \\ // 3. Original authors' names are not deleted. \\ // 4. The authors' names are not used to endorse or promote products \\ // derived from this software without specific prior written \\ // permission. \\ // THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK \\ // DISCLAIM ALL WARRANTIES With REGARD TO THIS SOFTWARE, INCLUDING \\ // ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT \\ // SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE \\ // FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES \\ // WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN \\ // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, \\ // ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF \\ // THIS SOFTWARE. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Author: Richard Caley (rjc@cstr.ed.ac.uk) \\ // -------------------------------------------------------------------- \\ // A tokeniser for scheme expressions. \\ // \\ // This should obviously be a subclass of java.io.StreamTokenizer, \\ // but that is as much use as a chocolate teapot so we have to go from \\ // first principles. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ package cstr.festival.scheme ; import java.lang.*; import java.util.*; import java.awt.*; import java.io.*; public class SchemeTokenizer { public static final int TT_EOF = StreamTokenizer.TT_EOF; public static final int TT_WORD = StreamTokenizer.TT_WORD; static final int TT_NOTHING = -4; protected Reader r; public String sval; public int ttype; protected int pending =-1; public SchemeTokenizer(Reader rd) { r=rd; ttype=TT_NOTHING; } public int nextToken() throws IOException { int c; if (pending >=0) { c=pending; pending = -1; } else c = r.read(); // Skip whitespace and comments while (true) { while (Character.isWhitespace((char)c)) c = r.read(); if (c == ';') { while (c != '\n' && c != '\r') c = r.read(); while (c == '\n' || c == '\r') c = r.read(); } else break; } if (c <0) ttype = TT_EOF; else if (c == '"') { ttype = c; StringBuffer b = new StringBuffer(100); boolean escape=false; while ((c=r.read()) != '"' || escape) { if (escape) { if (c == 'n') c='\n'; else if (c == 'r') c='\r'; else if (c == 't') c='\t'; b.append((char)c); escape=false; } else if (c == '\\') escape=true; else { b.append((char)c); escape=false; } } sval = b.toString(); } else if (Character.isLetterOrDigit((char)c) || c=='.' || c=='_' || c=='-' || c=='*' || c==':') { ttype = TT_WORD; StringBuffer b = new StringBuffer(100); b.append((char)c); while (Character.isLetterOrDigit((char)(c=r.read())) || c=='.' || c=='_' || c=='-' || c=='*' || c==':') b.append((char)c); pending=c; sval = b.toString(); } else ttype = c; return ttype; } } festival/src/modules/java/cstr/festival/scheme/SchemeReader.java0000644000175000017500000000625407262605650023424 0ustar awbawb //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Centre for Speech Technology Research \\ // University of Edinburgh, UK \\ // Copyright (c) 1996,1997 \\ // All Rights Reserved. \\ // Permission is hereby granted, free of charge, to use and distribute \\ // this software and its documentation without restriction, including \\ // without limitation the rights to use, copy, modify, merge, publish, \\ // distribute, sublicense, and/or sell copies of this work, and to \\ // permit persons to whom this work is furnished to do so, subject to \\ // the following conditions: \\ // 1. The code must retain the above copyright notice, this list of \\ // conditions and the following disclaimer. \\ // 2. Any modifications must be clearly marked as such. \\ // 3. Original authors' names are not deleted. \\ // 4. The authors' names are not used to endorse or promote products \\ // derived from this software without specific prior written \\ // permission. \\ // THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK \\ // DISCLAIM ALL WARRANTIES With REGARD TO THIS SOFTWARE, INCLUDING \\ // ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT \\ // SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE \\ // FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES \\ // WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN \\ // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, \\ // ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF \\ // THIS SOFTWARE. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Author: Richard Caley (rjc@cstr.ed.ac.uk) \\ // -------------------------------------------------------------------- \\ // A simple scheme reader. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ package cstr.festival.scheme ; import java.lang.*; import java.util.*; import java.awt.*; import java.io.*; public abstract class SchemeReader { static final int SE_EOF=0; static final int SE_ATOM=1; static final int SE_LIST=2; static final int SE_CB=3; SchemeTokenizer tk; public SchemeReader(Reader r) { tk = new SchemeTokenizer(r); } public abstract Object nextExpr() throws IOException ; } festival/src/modules/java/cstr/festival/scheme/ReflectingSchemeReader.java0000644000175000017500000001164707262605650025431 0ustar awbawb //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Centre for Speech Technology Research \\ // University of Edinburgh, UK \\ // Copyright (c) 1996,1997 \\ // All Rights Reserved. \\ // Permission is hereby granted, free of charge, to use and distribute \\ // this software and its documentation without restriction, including \\ // without limitation the rights to use, copy, modify, merge, publish, \\ // distribute, sublicense, and/or sell copies of this work, and to \\ // permit persons to whom this work is furnished to do so, subject to \\ // the following conditions: \\ // 1. The code must retain the above copyright notice, this list of \\ // conditions and the following disclaimer. \\ // 2. Any modifications must be clearly marked as such. \\ // 3. Original authors' names are not deleted. \\ // 4. The authors' names are not used to endorse or promote products \\ // derived from this software without specific prior written \\ // permission. \\ // THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK \\ // DISCLAIM ALL WARRANTIES With REGARD TO THIS SOFTWARE, INCLUDING \\ // ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT \\ // SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE \\ // FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES \\ // WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN \\ // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, \\ // ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF \\ // THIS SOFTWARE. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Author: Richard Caley (rjc@cstr.ed.ac.uk) \\ // -------------------------------------------------------------------- \\ // A Scheme reader which returns the s expression as s atring suitable \\ // for passing on to a scheme interpreter, eg for sending to festival. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ package cstr.festival.scheme ; import java.lang.*; import java.util.*; import java.awt.*; import java.io.*; public class ReflectingSchemeReader extends SchemeReader { public ReflectingSchemeReader(Reader r) { super(r); } private int parseSexp(StringBuffer b) throws IOException { tk.nextToken(); while (tk.ttype == StreamTokenizer.TT_EOL) { b.append(" "); tk.nextToken(); } if (tk.ttype == StreamTokenizer.TT_EOF) return SE_EOF; if (tk.ttype == '\'') { b.append("'"); int t = parseSexp(b); return t; } if (tk.ttype == ')') { b.append(") "); return SE_CB; } if (tk.ttype == '(') { b.append("("); while (true) { int se_type = parseSexp(b); if (se_type == SE_EOF || se_type == SE_CB) return SE_LIST; } } if (tk.ttype == StreamTokenizer.TT_WORD) { b.append(tk.sval); b.append(" "); return SE_ATOM; } if (tk.ttype == '"') { b.append('"'); int s=b.length(); b.append(tk.sval); for(int i=s; i= ' ' && tk.ttype <= '\u00ff') { b.append((char)tk.ttype); b.append(' '); return SE_ATOM; } System.out.println("UNEXPECTED "+tk.ttype+" "+tk.sval); return SE_EOF; } public Object nextExpr() throws IOException { return (Object)nextExprString(); } public String nextExprString() throws IOException { StringBuffer exp = new StringBuffer(80); int type = parseSexp(exp); return type == SE_EOF ? (String)null : exp.toString(); } } festival/src/modules/java/cstr/festival/jsapi/Makefile0000644000175000017500000000526207262605650021532 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../../../../../.. DIRNAME=src/modules/java/cstr/festival/jsapi JAVA_CLASSES = EngineCentral FestivalModeDesc \ FestivalSynthesizer FestivalQueueItem \ MessageQueue NEED_JAVA=1 FILES = Makefile ReadMe $(JAVA_CLASSES:=.java) ALL = .java .sub_directories include $(TOP)/config/common_make_rules festival/src/modules/java/cstr/festival/jsapi/ReadMe0000644000175000017500000001716610045556464021161 0ustar awbawb Festival Java Speech API (jsapi) Support ======================================== (preliminary) Contents -------- 1 Overview 2 Requirements 3 Limitations 4 Building 5 Use 6 Issues --o--oOo--o-- 1 Overview ---------- This is a preliminary implementation of the Java speech API using festival. It is built on top of the Java festival client code, meaning that it talks to an instance of festival running as a server, possibly on another machine. See the festival documentation for instructions on running such a server. Some parts of JSAPI are not implemented. See section 3 for details. The code is almost pure Java. Since Java 1.1 doesn't support audio in any real sense, the speech tools program `na_play' is used to play audio. In the near future a version of the speech tools java code which uses the Java media interface will be created (as part of the move of all of our Java code to Java 1.2) allowing a pure-Java client and JSAPI implementation. The basic structure chosen for this implementation was to use festival purely as a text-to-waveform engine (as opposed to, for instance, having festival play the waveforms). text to waveform conversion is requested as soon as a Speakable is placed on the Synthesizer's queue. This has the advantage that the waveform will be available as soon as possible, but the disadvantage that cancelling a request on the queue may not stop festival from working on it (in fact at the moment jobs on festival's job queue are never cancelled). --o--oOo--o-- 2 Requirements -------------- In order to use this JSAPI synthesizer you will need some things in addition to the normal requirements for compiling speech tools and festival: A Java compiler. The latest version of the Sun Java Development Kit for Java version 1.1 (currently we use jdk-1.1.7) An implementation of the JSAPI framework (jsapi.jar). We currently use a few classes from the Java Speech API Utilities from the Speech Group at Sun Microsystems. Unfortunately the third dependency makes this implementation useful only to those who have access to the JSAPI utilities, which are not generally distributed. We plan to remove this problem as soon as possible. --o--oOo--o-- 3 Limitations ------------- Some areas of the JSAPI have not been implemented, for limits of time, lack of support within festival or just because we are unsure about the correct mapping of JSAPI to festival functionality. We are working on reducing this list. * The SynthesizerProperties mechanism is not implemented. This should be done by talking to festival, the infrastructure of this exists. Temporarily the non-standard setVoice() method of FestivalSynthesizer can be used to do what is most likely to be needed. * The festival EngineCentral class has two modes hard wired into it. This should be replaced by code to query festival. There is an issue of when the system should commit to which server to connect to. * The SynthesizerListener mechanism is not implemented. * The Pause and Resume methods are not implemented. * WORD_STARTED and MARKER_REACHED Speakable events are not produced. This code has not been only lightly exercised and so there are no doubt other bugs and limitations. Please let us know of any you spot. --o--oOo--o-- 4 Building ---------- First you should build the speech tools library with (at least) JAVA support. JAVA_CPP support is not necessary, but won't hurt. To do this you will need to define JAVA_HOME in either your environment or the speech_tools config file as you would do for any java application. Copy festival/config/config-dist to festival/config/config (if you haven't done this already) and edit config. Add JSAPI to the ALSO_INCLUDE line, as follows: ALSO_INCLUDE += JSAPI Either in the config file or in your shell environment define JSAPI_HOME and JSUTILS_HOME. For example JSAPI_HOME=/local/install/place/jsutils-1.0/jsapi JSUTILS_HOME=/local/install/place/jsutils-1.0 now go to the top level festival directory and do gnumake info to check that JSAPI is included in the list of selected modules and that JSAPI_HOME is as you would expect. Now you should be able to compile festival with gnumake as normal. --o--oOo--o-- 5 Use ----- Before writing JSAPI code you should probably run the simple example JSAPI application we have included, bin/jsapi_example. The following steps are necessary for any use of festival via JSAPI. In order to use JSAPI and festival you need to register the festival synthesizer. See the JSAPI documentation for details, but basically there are two methods: One is to put a file listing available synthesizers in either your home directory or JAVA_HOME. The file is called `speech.properties' and an example containing just festival can be found in festival/lib/speech.properties. The second method is to have your code register the synthesizers it will use. The name of the festival `EngineCentral' class is `cstr.festival.jsapi.EngineCentral'. The jsapi_example program has a -r argument you can Next you must run a festival in server mode and let the JSAPI code know where it is. See the festival documentation for instruction on how to run it, when you have done that you need to set the following java system properties (eg with -D on the java command line) festival.server=myhost.yoyodyne.com festival.port=1314 The jsapi_example program has a `--server host:port' argument and also looks in the FESTIVAL_SERVER environment variable for information in the same format. You can now run the jsapi_example program. This program simply passes each command line argument to the synthesizer. For instance $ FESTIVAL_SERVER=myhost.yoyodyne.com:1314 $ export FESTIVAL_SERVER $ bin/jsapi_example "Say this. Then say this." "Wasn't that fun!" This gives two strings to the synthesizer, resulting in 3 waveforms being played (because the first contains two sentences). Give jsapi_example the -v argument to get more details of what is happening. $ bin/jsapi_example -v "Say this. Then say this." "Wasn't that fun!" engine=Festival [myhost.yoyodyne.com:1314] locale = en_GB mode = myhost.yoyodyne.com:1314 running = true voices = Roger Burroughs Text is 'Say this. Then say this.' SpeakableEvent 'Say this. Then say this.' speakableStarted SpeakableEvent 'Say this. Then say this.' topOfQueue Text is 'Wasn't that fun!' SpeakableEvent 'Say this. Then say this.' speakableEnded SpeakableEvent 'Wasn't that fun!' speakableStarted SpeakableEvent 'Wasn't that fun!' topOfQueue SpeakableEvent 'Wasn't that fun!' speakableEnded finishing The code for this program is in src/modules/java/cstr/testPrograms/SayHelloWorld.java 6 Issues -------- Some issues remain as to the mapping between festival and JSAPI. JSAPI defines a three level hierarchy Engine/Mode/Voice where an Engine/Mode pair defines a unique, single locale synthesizer. Festival Has a flat space of voices, which could of course be partitioned into locales, but introduces the additional question of which server we are connecting to. Is the Engine `Festival' or is it `the Festival running on myhost.yoyodyne.com at port 1314'? This means the engine doesn't identify the class of all festival synthesizers. Perhaps the host and port number should be the mode, but then it doesn't map to a unique locale (since one server could have, say, English and Spanish voices). Related to this, how should the location of the server be passed to the JSAPI code. The current method of using the two system properties limits us to one server. Some work needs to be done on making the JSAPI interface query the server, to get a list of modes and voices, and also to get and set properties. festival/src/modules/java/cstr/festival/jsapi/EngineCentral.java0000644000175000017500000000740707262605650023456 0ustar awbawb //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Centre for Speech Technology Research \\ // University of Edinburgh, UK \\ // Copyright (c) 1996,1997 \\ // All Rights Reserved. \\ // Permission to use, copy, modify, distribute this software and its \\ // documentation for research, educational and individual use only, is \\ // hereby granted without fee, subject to the following conditions: \\ // 1. The code must retain the above copyright notice, this list of \\ // conditions and the following disclaimer. \\ // 2. Any modifications must be clearly marked as such. \\ // 3. Original authors' names are not deleted. \\ // This software may not be used for commercial purposes without \\ // specific prior written permission from the authors. \\ // THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK \\ // DISCLAIM ALL WARRANTIES With REGARD TO THIS SOFTWARE, INCLUDING \\ // ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT \\ // SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE \\ // FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES \\ // WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN \\ // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, \\ // ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF \\ // THIS SOFTWARE. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Author: Richard Caley (rjc@cstr.ed.ac.uk) \\ // -------------------------------------------------------------------- \\ // Class which describes what festival voices are available. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ package cstr.festival.jsapi ; import java.lang.*; import java.util.*; import java.awt.*; import javax.speech.*; import javax.speech.synthesis.*; public class EngineCentral implements javax.speech.EngineCentral { private static String server; private static int port; static { server = System.getProperty("festival.server", "localhost"); port = Integer.getInteger("festival.port", 1314).intValue(); } Voice rab = new Voice("Roger Burroughs", Voice.GENDER_MALE, Voice.AGE_MIDDLE_ADULT, "LPC Diphone" ); Voice ked = new Voice("Kurt Dusterhoff", Voice.GENDER_MALE, Voice.AGE_YOUNGER_ADULT, "LPC Diphone" ); SynthesizerModeDesc festival_rab = new FestivalModeDesc(server, port, server+":"+Integer.toString(port), Locale.UK, Boolean.TRUE, new Voice [] {rab} ); SynthesizerModeDesc festival_ked = new FestivalModeDesc(server, port, server+":"+Integer.toString(port), Locale.US, Boolean.TRUE, new Voice [] {ked} ); public EngineList createEngineList(EngineModeDesc require) throws SecurityException { EngineList engines = new EngineList(); engines.addElement(festival_rab); engines.addElement(festival_ked); return engines; } } festival/src/modules/java/cstr/festival/jsapi/FestivalModeDesc.java0000644000175000017500000000617007262605650024115 0ustar awbawb //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Centre for Speech Technology Research \\ // University of Edinburgh, UK \\ // Copyright (c) 1996,1997 \\ // All Rights Reserved. \\ // Permission to use, copy, modify, distribute this software and its \\ // documentation for research, educational and individual use only, is \\ // hereby granted without fee, subject to the following conditions: \\ // 1. The code must retain the above copyright notice, this list of \\ // conditions and the following disclaimer. \\ // 2. Any modifications must be clearly marked as such. \\ // 3. Original authors' names are not deleted. \\ // This software may not be used for commercial purposes without \\ // specific prior written permission from the authors. \\ // THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK \\ // DISCLAIM ALL WARRANTIES With REGARD TO THIS SOFTWARE, INCLUDING \\ // ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT \\ // SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE \\ // FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES \\ // WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN \\ // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, \\ // ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF \\ // THIS SOFTWARE. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Author: Richard Caley (rjc@cstr.ed.ac.uk) \\ // -------------------------------------------------------------------- \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ package cstr.festival.jsapi ; import java.lang.*; import java.util.*; import java.io.*; import java.net.*; import javax.speech.*; import javax.speech.synthesis.*; public class FestivalModeDesc extends SynthesizerModeDesc implements EngineCreate { private String server; private int port; FestivalModeDesc(String s, int p, String mode, Locale l, Boolean running, Voice [] voices) { super("Festival", mode, l, running, voices); server=s; port=p; } public int getPort() { return port; } public String getServer() { return server; } public Engine createEngine() throws IllegalArgumentException, EngineException { return new FestivalSynthesizer(this); } } festival/src/modules/java/cstr/festival/jsapi/FestivalSynthesizer.java0000644000175000017500000002026307262605650024760 0ustar awbawb //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Centre for Speech Technology Research \\ // University of Edinburgh, UK \\ // Copyright (c) 1996,1997 \\ // All Rights Reserved. \\ // Permission to use, copy, modify, distribute this software and its \\ // documentation for research, educational and individual use only, is \\ // hereby granted without fee, subject to the following conditions: \\ // 1. The code must retain the above copyright notice, this list of \\ // conditions and the following disclaimer. \\ // 2. Any modifications must be clearly marked as such. \\ // 3. Original authors' names are not deleted. \\ // This software may not be used for commercial purposes without \\ // specific prior written permission from the authors. \\ // THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK \\ // DISCLAIM ALL WARRANTIES With REGARD TO THIS SOFTWARE, INCLUDING \\ // ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT \\ // SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE \\ // FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES \\ // WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN \\ // AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, \\ // ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF \\ // THIS SOFTWARE. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ // \\ // Author: Richard Caley (rjc@cstr.ed.ac.uk) \\ // -------------------------------------------------------------------- \\ // A JSAPI synthesizer which talks to festivel. \\ // \\ //\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\//\\ package cstr.festival.jsapi ; import java.lang.*; import java.util.*; import java.io.*; import java.net.*; import javax.speech.*; import javax.speech.synthesis.*; import cstr.festival.client.*; public class FestivalSynthesizer extends com.sun.speech.engine.BaseEngine implements javax.speech.synthesis.Synthesizer { FestivalModeDesc desc; private String server; private int port; private Session session; private Vector speakableListeners = new Vector(0); private JobQueue queue; private String voice; public FestivalSynthesizer(FestivalModeDesc d) { desc=d; server=d.getServer(); port=d.getPort(); session = null; queue = new JobQueue(); } public void allocate() throws EngineException, EngineStateError { super.allocate(); try { session = new Session(server, port); } catch (IOException ex) { super.deallocate(); throw new EngineException("IO Exception: "+ex.getMessage()); } session.initialise(); Object res = session.synchronousRequest("(tts_return_to_client)"); // System.out.println("(tts_return_to_client)="+res); Object res2 = session.synchronousRequest("(Parameter.set 'Wavefiletype 'snd)"); // System.out.println("(Parameter.set 'Wavefiletype 'snd)="+res2); setVoice("rab"); } public void setVoice(String v) { voice=v; Object res = session.synchronousRequest("(voice_"+ voice +"_diphone)"); } public void deallocate() throws EngineException, EngineStateError { super.deallocate(); session.terminate(false); } public Session getSession() { return session; } public EngineModeDesc getEngineModeDesc() { return desc; } public void addSpeakableListener(SpeakableListener listener) { speakableListeners.addElement(listener); } public void removeSpeakableListener(SpeakableListener listener) { speakableListeners.removeElement(listener); } public void cancel() { cancelJob((FestivalQueueItem)queue.top()); } public void cancel(Object it) { FestivalQueueItem item = findJobFor(it); if (item != null) cancelJob(item); } public void cancelAll() { Enumeration jobs = queue.elements(); while (jobs.hasMoreElements()) { FestivalQueueItem item = (FestivalQueueItem)jobs.nextElement(); cancelJob(item); } } public Enumeration enumerateQueue() { return queue.elements(); } public SynthesizerProperties getSynthesizerProperties() { return null; } public String phoneme(String p) { return ""; } public void speak(String text, SpeakableListener listener) { FestivalQueueItem item = new FestivalQueueItem(this, text, "jsml", listener); addJob(item); } public void speak(URL url, SpeakableListener listener) { String text = "URL not yet supported"; FestivalQueueItem item = new FestivalQueueItem(this, text, "text", listener); addJob(item); } public void speak(Speakable text, SpeakableListener listener) { FestivalQueueItem item = new FestivalQueueItem(this, text, listener); addJob(item); } public void speakPlainText(String text, SpeakableListener listener) { FestivalQueueItem item = new FestivalQueueItem(this, text, "text", listener); addJob(item); } protected void checkQueue() { if (queue.isEmpty()) { engineState &= ~Synthesizer.QUEUE_NOT_EMPTY; engineState |= Synthesizer.QUEUE_EMPTY; } else { engineState &= ~Synthesizer.QUEUE_NOT_EMPTY; engineState |= Synthesizer.QUEUE_EMPTY; toTop((FestivalQueueItem)queue.top()); } } protected void addJob(FestivalQueueItem item) { queue.add(item); item.startSynthesis(); checkQueue(); } protected void cancelJob(FestivalQueueItem item) { queue.remove(item); item.cancel(); SpeakableEvent e = new SpeakableEvent(item.getSource(), SpeakableEvent.SPEAKABLE_CANCELLED); if (item.getSpeakableListener() != null) item.getSpeakableListener().speakableCancelled(e); for(int i=0; i",stdin); else { ent = NewExternalEntity(0,0,strdup8(inname),0,0); if (ent) source = EntityOpen(ent); } if (!source) { cerr << "xml: unable to open input file \"" << inname << "\"" << endl; festival_error(); } element_defs = siod_get_lval("xxml_elements",NULL); p = NewParser(); ParserSetEntityOpener(p, entity_open); ParserSetFlag(p, ReturnDefaultedAttributes, 1); if (ParserPush(p, source) == -1) { cerr << "xml: parser error\n" << endl; festival_error(); } while (1) { XBit bit = ReadXBit(p); if (bit->type == XBIT_eof) break; else if (bit->type == XBIT_start) { Attribute b; LISP att=NIL; for (b=bit->attributes; b; b=b->next) att = cons(cons(rintern(b->definition->name), cons(cons(rintern(b->value),NIL),NIL)),att); utt = xxml_call_element_function( EST_String("(")+bit->element_definition->name,att, element_defs,utt); } else if (bit->type == XBIT_end) { utt = xxml_call_element_function( EST_String(")")+bit->element_definition->name,NIL, element_defs,utt); } else if (bit->type == XBIT_empty) { Attribute b; LISP att=NIL; for (b=bit->attributes; b; b=b->next) att = cons(cons(rintern(b->definition->name), cons(cons(rintern(b->value),NIL),NIL)),att); utt = xxml_call_element_function( EST_String(bit->element_definition->name),att, element_defs,utt); } else if (bit->type == XBIT_pcdata) { utt = xxml_get_tokens(bit->pcdata_chars, siod_get_lval("xxml_word_features",NULL), utt); } else if (bit->type == XBIT_cdsect) { utt = xxml_get_tokens(bit->cdsect_chars, siod_get_lval("xxml_word_features",NULL), utt); } else if (bit->type == XBIT_pi) { cerr << "xml: ignoring pi " << bit->pi_chars << endl; } else if (bit->type == XBIT_error) { ParserPerror(p,bit); festival_error(); } else { // ignore it } FreeXBit(bit); } // Last call (should synthesize trailing tokens) utt = xxml_call_element_function(" ",NIL,element_defs,utt); FreeDtd(p->dtd); FreeParser(p); if (ent) FreeEntity(ent); return NIL; } static LISP xml_register_id(LISP pattern_l, LISP result_l) { EST_String pattern = get_c_string(pattern_l); EST_String result = get_c_string(result_l); utterance_xml_register_id(pattern, result); return NIL; } static LISP xml_registered_ids() { EST_StrList ids; utterance_xml_registered_ids(ids); LISP result= NIL; EST_Litem *p; for(p=ids.head(); p != NULL; p=p->next()) { EST_String pat = ids(p); p=p->next(); EST_String res = ids(p); result = cons( cons(strcons(pat.length(), pat), strcons(res.length(), res)), result); } return result; } void festival_rxp_init() { proclaim_module("rxp"); init_subr_1("tts_file_xml",tts_file_xml, "(tts_file_xml FILE)\n\ Low level tts processor for XML files. This assumes that element\n\ instructions are set up in the variable xxml_elements."); init_subr_2("xml_register_id", xml_register_id, "(xml_register_id PATTERN RESULT) \n\ Add a rule for where to find XML entities such as DTDs.\n\ The pattern is a regular expression, the result is a string\n\ with substitutions. If the PATTERN matches the a PUBLIC\n\ or SYSTEM identifier of an XML entity, the RESULT is expanded\n\ and then used as a filename."); init_subr_0("xml_registered_ids", xml_registered_ids, "(xml_registered_ids) \n\ Return the current list of places to look for XML entities."); } static InputSource entity_open(Entity ent, void *arg) { (void)arg; return utterance_xml_try_and_open(ent); } festival/src/modules/UniSyn_phonology/Makefile0000644000175000017500000000566107327271540020240 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## Metrical Trees ## ########################################################################### TOP=../../.. DIRNAME=src/modules/UniSyn_phonology H = us_duration.h TSRCS = us_aux.cc CPPSRCS = UniSyn_phonology.cc mettree.cc syllabify.cc subword.cc \ UniSyn_build.cc us_duration.cc unisyn_tilt.cc $(TSRCS) SRCS = $(CPPSRCS) OBJS = $(CPPSRCS:.cc=.o) FILES=Makefile $(SRCS) $(H) unisyn_phonology.mak LOCAL_INCLUDES = -I../include INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/UniSyn_phonology/UniSyn_phonology.cc0000644000175000017500000004477111465010356022431 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Paul Taylor */ /* Date : June 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* Metrical Tree based Phonology system */ /* */ /*=======================================================================*/ #include #include "festival.h" //#include "development/EST_FeatureData.h" #include "../UniSyn/us_features.h" /**** FUNCTIONS FOR UNISYN BUILD ****/ LISP FT_add_trans_seg(LISP l_utt, LISP lf_seg); LISP FT_add_trans_duration(LISP l_utt, LISP lf_seg); LISP FT_add_trans_metrical_tree(LISP l_utt, LISP lf_input, LISP lf_output); LISP FT_add_xml_relation(LISP l_utt, LISP xml_file); LISP FT_add_trans_word(LISP l_utt, LISP lf_word); LISP FT_add_trans_intonation(LISP l_utt, LISP lf_int, LISP l_add_words); LISP FT_add_f0_points(LISP l_utt, LISP lf_f0); LISP FT_add_coefs(LISP l_utt, LISP lf_f0); LISP FT_syl_to_word_intonation(LISP l_utt); LISP FT_intonation_diagnostics(LISP l_ref, LISP l_test, LISP l_rel_name, LISP l_valid); /**** END FUNCTIONS FOR UNISYN BUILD ****/ void extend_tree(EST_Item *m, EST_Item *p, const EST_String &terminal, const EST_String &second_tree); void parse_words(EST_Utterance &utt); LISP FT_add_trans_word(LISP l_utt, LISP lf_word, LISP keep_times); LISP FT_US_add_intonation(LISP utt); LISP FT_focus_nth_item(LISP utt, LISP lrel, LISP w); LISP FT_focus_nth_tree_item(LISP utt, LISP lrel, LISP w); LISP FT_foot_nth_item(LISP utt, LISP w); void make_prosodic_tree(EST_Item *m, EST_Item *p); void phrase_factor(EST_Utterance &u, const EST_String &base_stream, const EST_String &mettree); void stress_factor1(EST_Utterance &u, const EST_String &base_stream, const EST_String &m); void stress_factor2(EST_Utterance &u, const EST_String &base_stream, const EST_String &m); void main_stress(EST_Item *s); void footing(EST_Item *n1); void add_monotone_targets(EST_Utterance &u, float start_f0, float end_f0); void parse_words(EST_Utterance &utt); void syntax_to_metrical_words(EST_Utterance &utt); void binaryize_tree(EST_Utterance &utt, const EST_String &base_tree, const EST_String &new_tree); /*void assign_phone_z_scores(EST_Utterance &u, const EST_String &seg_name); void promote_vowel_z_score(EST_Utterance &u, const EST_String &st_name, const EST_String &syl_name); void promote_mean_z_score(EST_Utterance &u, const EST_String &st_name, const EST_String &syl_name); void find_cvc_words(EST_Utterance &u); */ void tilt_to_f0(EST_Utterance &u); void scale_tilt(EST_Relation &ev, float shift, float scale); void vowel_tilt_to_abs_tilt(EST_Utterance &u); void targets_to_f0(EST_Relation &targ, EST_Track &f0, const float shift); void syntax_metrical_tree(EST_Utterance &utt, const EST_String &wordfile); void tilt_to_f0(EST_Relation &intonation, EST_Relation &f0); void legal_metrical_tree(EST_Item *s); LISP FT_focus_nth_item(LISP utt, LISP lrel, LISP w) { EST_Utterance *u = get_c_utt(utt); int f = get_c_int(w); EST_String relname = get_c_string(lrel); EST_Item *n; int i; cout << "Focusing item " << f << " in relation " << relname << endl; for (i = 1, n = u->relation(relname)->head(); n; n = n->next(), ++i) if (i == f) break; if (n == 0) { cerr << "Error: Can't focus node " << f << " in a relation with only " << i << " items\n"; return NIL; } main_stress(n->as_relation("MetricalTree")); if (siod_get_lval("mettree_debug", NULL) != NIL) u->save("focus.utt", "est"); return utt; } LISP FT_focus_nth_tree_item(LISP utt, LISP lrel, LISP w) { EST_Utterance *u = get_c_utt(utt); EST_String dir; EST_String relname = get_c_string(lrel); EST_Item *n; LISP l; n = u->relation(relname)->head(); for (l = w; (l != NIL) && (n); l = cdr(l)) { dir = get_c_string(car(l)); cout << "dir = " << dir << endl; if (dir == "l") n = daughter1(n); else if (dir == "r") n = daughter2(n); else cerr << "Bad instruction: " << dir << endl; } main_stress(n->as_relation("MetricalTree")); if (siod_get_lval("mettree_debug", NULL) != NIL) u->save("focus.utt", "est"); return utt; } LISP FT_foot_nth_item(LISP utt, LISP w) { EST_Utterance *u = get_c_utt(utt); int f = get_c_int(w); (void) f; (void)u; EST_Item *n; int i; cout << "Footing item " << f << endl; for (i = 1, n = u->relation("Syllable")->head(); n; n = n->next(), ++i) if (i == f) break; if (n == 0) { cerr << "Error: Can't foot node " << f << " in a relation with only " << i << " items\n"; return NIL; } footing(n->as_relation("MetricalTree")); if (siod_get_lval("mettree_debug", NULL) != NIL) u->save("foot.utt", "est"); return utt; } LISP FT_US_add_intonation(LISP utt) { (void) utt; EST_Utterance *u = get_c_utt(utt); EST_String base_int; /* lt = siod_get_lval("us_acc_thresh", NULL); t = (lt == NIL) ? DEF_THRESH : get_c_float(lt); lt = siod_get_lval("us_base_int", NULL); if (lt == NIL) base_int = "Syllable"; else { char *x = get_c_string(lt); base_int = x; } add_intonation(*u, base_int, t); */ float start_f0 = get_c_float(siod_get_lval("us_start_f0", NULL)); float end_f0 = get_c_float(siod_get_lval("us_end_f0", NULL)); u->create_relation("f0"); EST_Track *f0 = new EST_Track; EST_Item *a = u->relation("f0")->append(); a->set_val("f0",est_val(f0)); add_monotone_targets(*u, start_f0, end_f0); targets_to_f0(*u->relation("Target"), *f0, 0.01); return utt; } LISP FT_extend_tree(LISP l_utt, LISP largs) { EST_Item *p, *m; EST_Utterance *u = get_c_utt(l_utt); EST_String new_tree = get_c_string(car(largs)); EST_String first_tree = get_c_string(car(cdr(largs))); EST_String second_tree = get_c_string(car(cdr(cdr(largs)))); EST_String terminal = get_c_string(car(cdr(cdr(cdr(largs))))); u->create_relation(new_tree); for (m = u->relation(first_tree)->head(); m; m = m->next()) { p = u->relation(new_tree)->append(m); extend_tree(m, p, terminal, second_tree); } return l_utt; } static void add_keep_nodes(EST_Item *n, EST_String keep) { if (n == 0) return; n->set(keep, 1); for (EST_Item *p = daughter1(n); p; p = p->next()) add_keep_nodes(p, keep); } static void remove_sisters(EST_Item *n) { EST_Item *m; EST_Item *p = parent(n); if (p == 0) return; for (EST_Item *s = daughter1(p); s; s = m) { m = s->next(); if (s != n) s->unref_all(); } move_sub_tree(n, p); remove_sisters(p); } LISP FT_copy_sub_tree(LISP l_utt, LISP l_id, LISP l_relation) { EST_Utterance *u = get_c_utt(l_utt); EST_Utterance *new_utt = new EST_Utterance; new_utt = u; EST_Item *n = new_utt->id(get_c_string(l_id))-> as_relation(get_c_string(l_relation)); remove_sisters(n); // reset n - should now be a root node. n = new_utt->id(get_c_string(l_id))-> as_relation(get_c_string(l_relation)); // remove other root nodes. EST_Item *s, *m; for (s = new_utt->relation(get_c_string(l_relation))->head(); s; s = m) { m = s->next(); if (s != n) s->unref_all(); } n = new_utt->id(get_c_string(l_id))-> as_relation(get_c_string(l_relation)); add_keep_nodes(n, "keep"); for (s = new_utt->relation("Segment")->head(); s; s = m) { m = s->next(); if (!s->f_present("keep")) s->unref_all(); } for (s = new_utt->relation("Syllable")->head(); s; s = m) { m = s->next(); if (!s->f_present("keep")) s->unref_all(); } for (s = new_utt->relation("Word")->head(); s; s = m) { m = s->next(); if (!s->f_present("keep")) s->unref_all(); } /* for (s = new_utt->relation("Segment")->head(); s->S("id") != first_leaf(n)->S("id"); s = m) { m = s->next(); cout << "deleting segment :" << s->S("name") << endl; s->unref_all(); } for (s = next(last_leaf(n)->as_relation("Segment")); s; s = m) { m = s->next(); cout << "deleting segment :" << s->S("name") << endl; s->unref_all(); } */ cout << "h1\n"; LISP n_utt; cout << "h1\n"; n_utt = siod(new_utt); cout << "h1\n"; return n_utt; } void add_syllable_name(EST_Item *syl, const EST_String &fname); void add_non_terminal_features(EST_Item *s, EST_Features &f); LISP FT_add_match_features(LISP l_utt) { EST_Item *p; EST_Utterance *u = get_c_utt(l_utt); for (p = u->relation("Word")->head(); p; p = p->next()) p->set("match", p->S("name")); for (p = u->relation("Segment")->head(); p; p = p->next()) p->set("match", p->S("name")); for (p = u->relation("Syllable")->head(); p; p = p->next()) add_syllable_name(p, "match"); EST_Features tf; tf.set_function("end", "standard+unisyn_leaf_end"); tf.set_function("start", "standard+unisyn_leaf_start"); tf.set_function("dur", "standard+duration"); tf.set("time_path", "ProsodicTree"); tf.set("time_path", "ProsodicTree"); add_non_terminal_features(u->relation("ProsodicTree")->head(), tf); return l_utt; } LISP FT_tilt_to_f0(LISP l_utt, LISP l_f0_name) { EST_String f0_name = get_c_string(l_f0_name); EST_Utterance *u = get_c_utt(l_utt); EST_Relation *f0 = u->create_relation(f0_name); tilt_to_f0(*u->relation("Intonation"), *f0); return l_utt; } LISP FT_scale_tilt(LISP l_utt, LISP l_shift, LISP l_scale) { scale_tilt(*(get_c_utt(l_utt)->relation("Intonation")), get_c_float(l_shift), get_c_float(l_scale)); return l_utt; } LISP FT_vowel_tilt_to_abs_tilt(LISP l_utt) { vowel_tilt_to_abs_tilt(*get_c_utt(l_utt)); return l_utt; } LISP FT_phrase_factor(LISP l_utt, LISP l_base_name, LISP l_met_name) { phrase_factor(*get_c_utt(l_utt), get_c_string(l_base_name), get_c_string(l_met_name)); return l_utt; } LISP FT_stress_factor(LISP l_utt, LISP l_rel_name, LISP l_num) { if (get_c_int(l_num) == 1) stress_factor1(*get_c_utt(l_utt), get_c_string(l_rel_name), "LexicalMetricalTree"); else stress_factor2(*get_c_utt(l_utt), get_c_string(l_rel_name), "LexicalMetricalTree"); return l_utt; } LISP FT_smooth_f0(LISP l_utt, LISP l_rel_name, LISP l_num) { if (get_c_int(l_num) == 1) stress_factor1(*get_c_utt(l_utt), get_c_string(l_rel_name), "LexicalMetricalTree"); else stress_factor2(*get_c_utt(l_utt), get_c_string(l_rel_name), "LexicalMetricalTree"); return l_utt; } LISP FT_legal_metrical_tree(LISP l_utt) { EST_Utterance *u = get_c_utt(l_utt); EST_Item *s; for (s = u->relation("MetricalTree")->head(); s; s= s->next()) legal_metrical_tree(s); return l_utt; } void auto_metrical_lex(EST_Utterance &utt); LISP FT_unisyn_lex(LISP l_utt) { EST_Utterance *u = get_c_utt(l_utt); auto_metrical_lex(*u); return l_utt; } /*LISP FT_MetricalTree_Utt(LISP l_utt) { EST_Utterance *u = get_c_utt(l_utt); auto_metrical_tree(*u); auto_metrical_lex(*u); return l_utt; } */ LISP FT_syntax_to_metrical_words(LISP l_utt) { syntax_to_metrical_words(*get_c_utt(l_utt)); return l_utt; } LISP FT_binaryize_tree(LISP l_utt, LISP old_tree, LISP new_tree) { binaryize_tree(*get_c_utt(l_utt), get_c_string(old_tree), get_c_string(new_tree)); return l_utt; } LISP FT_parse_words(LISP l_utt) { parse_words(*get_c_utt(l_utt)); return l_utt; } void festival_UniSyn_phonology_init(void) { init_subr_1("US_add_intonation", FT_US_add_intonation, "(US_add_intonation UTT)"); init_subr_1("vowel_tilt_to_abs_tilt", FT_vowel_tilt_to_abs_tilt, "."); init_subr_2("tilt_to_f0", FT_tilt_to_f0, "."); init_subr_3("scale_tilt", FT_scale_tilt, "(scale_tilt UTT shift scale)\n" "Add shift Hz to each event and increase range by a factor\n" "of scale. (scale UTT 0.0 1.0) leaves the tilt parameters\n" "unaffected\n"); /* init_subr_3("promote_vowel_z_score", FT_promote_vowel_z_score, "(promote_vowel_z_score UTT RELATION)\n\ Focus nth item in relation."); init_subr_3("promote_mean_z_score", FT_promote_mean_z_score, "(promote_mean_z_score UTT RELATION)\n\ Focus nth item in relation."); init_subr_2("assign_phone_z_scores", FT_assign_phone_z_scores, "(assign_phone_z_scores UTT RELATION RELATION)\n\ Focus nth item in relation."); */ init_subr_3("focus_nth_item", FT_focus_nth_item, "(focus_nth_item UTT RELATION WordNumber)\n\ Focus nth item in relation."); init_subr_3("focus_nth_tree_item", FT_focus_nth_tree_item, "(focus_tree_nth_item UTT RELATION (l r path)\n\ Focus nth item in relation."); init_subr_3("copy_sub_tree", FT_copy_sub_tree, "(foot_nth_item UTT WordNumber)\n\ Foot nth item in relation."); init_subr_2("foot_nth_item", FT_foot_nth_item, "(foot_nth_item UTT WordNumber)\n\ Foot nth item in relation."); init_subr_1("legal_metrical_tree", FT_legal_metrical_tree, "(legal_metrical_tree UTT))\n\ load data."); init_subr_1("add_match_features", FT_add_match_features, "(legal_metrical_tree UTT))\n\ load data."); init_subr_2("extend_tree", FT_extend_tree, "(extend_tree utt (new_tree first_tree second_tree terminal))"); init_subr_3("phrase_factor", FT_phrase_factor, "(phrase_factor utt)\n\ Apply phrase factor algorithm."); init_subr_3("stress_factor", FT_stress_factor, "(stress_factor utt RELATION VERSION)\n\ Apply version VERSION of stress factor algorithm to RELATION."); init_subr_1("syntax_to_metrical_words", FT_syntax_to_metrical_words, "(add_trans_intonation UTT Tilt file)\n\ Foot nth item in relation."); init_subr_3("binaryize_tree", FT_binaryize_tree, "(add_trans_intonation UTT Tilt file)\n\ Foot nth item in relation."); init_subr_1("parse_words", FT_parse_words, "(add_trans_intonation UTT Tilt file)\n\ Foot nth item in relation."); init_subr_1("unisyn_lex", FT_unisyn_lex, "(unisyn_lex UTT Tilt file)\n\ Foot nth item in relation."); /**** FUNCTIONS FOR UNISYN BUILD ****/ init_subr_3("add_trans_intonation", FT_add_trans_intonation, "(add_trans_intonation UTT Tilt_file ADD_WORDS)\n\ Foot nth item in relation."); init_subr_2("add_f0_points", FT_add_f0_points, "(add_f0_points UTT f0 file)\n\ Foot nth item in relation."); init_subr_2("add_coefs", FT_add_coefs, "(add_coefs UTT f0 file)\n\ Foot nth item in relation."); init_subr_2("add_trans_segment", FT_add_trans_seg, "(add_trans_segment UTT label_file)\n\ Add segment information\n\ Foot nth item in relation."); init_subr_2("add_trans_duration", FT_add_trans_duration, "(add_trans_duration UTT label_file)\n\ Foot nth item in relation."); init_subr_3("add_trans_word", FT_add_trans_word, "(add_trans_word UTT word_label_file)\n\ Foot nth item in relation."); init_subr_2("add_xml_relation", FT_add_xml_relation, "(add_xml_relation UTT xml_file))\n\ load data."); init_subr_1("syl_to_word_intonation", FT_syl_to_word_intonation, "(add_xml_relation UTT xml_file))\n\ load data."); init_subr_4("intonation_diagnostics", FT_intonation_diagnostics, "(intonation_diagnostics UTT Foot nth item in relation."); // semi redundant init_subr_3("add_trans_metrical_tree", FT_add_trans_metrical_tree, "(add_trans_intonation UTT Tilt file)\n\ Foot nth item in relation."); /**** END FUNCTIONS FOR UNISYN BUILD ****/ } /* LISP FT_assign_phone_z_scores(LISP l_utt, LISP l_rel_name) { assign_phone_z_scores(*get_c_utt(l_utt), get_c_string(l_rel_name)); return l_utt; } LISP FT_promote_mean_z_score(LISP l_utt, LISP l_st_name, LISP l_syl_name) { promote_mean_z_score(*get_c_utt(l_utt), get_c_string(l_st_name), get_c_string(l_syl_name)); return l_utt; } LISP FT_promote_vowel_z_score(LISP l_utt, LISP l_st_name, LISP l_syl_name) { promote_vowel_z_score(*get_c_utt(l_utt), get_c_string(l_st_name), get_c_string(l_syl_name)); return l_utt; } */ festival/src/modules/UniSyn_phonology/mettree.cc0000644000175000017500000010644711465010356020552 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black and Paul Taylor */ /* Date : February 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* An implementation of Metrical Tree Phonology */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "lexicon.h" #include "../UniSyn/us_features.h" EST_Features phone_def; float local_cost(const EST_Item *s1, const EST_Item *s2); static void mettree_add_words(EST_Utterance &u); void construct_metrical_tree(EST_Utterance &word); void add_end_silences(EST_Relation &segment); void StrListtoString(EST_StrList &l, EST_String &s, EST_String sep=" "); void parse_wsj_syntax(void); static void apply_nsr(EST_Utterance &u, const EST_String &tree); #if 0 static void remove_punctuation(EST_Utterance &u); static void add_intonation(EST_Utterance &u, const EST_String &base_int, float threshold); #endif void subword_list(EST_Item *w, EST_Relation &syllable, EST_Relation &metricaltree); void add_non_terminal_features(EST_Relation &r, EST_Features &f); void stress_factor1(EST_Utterance &u, const EST_String &base_stream, const EST_String &m); void stress_factor2(EST_Utterance &u, const EST_String &base_stream, const EST_String &m); void phrase_factor(EST_Utterance &u); void main_stress(EST_Item *s); void end_to_dur(EST_Relation &r); void footing(EST_Item *n1); LISP FT_Classic_Phrasify_Utt(LISP args); LISP FT_Classic_POS_Utt(LISP args); LISP FT_PParse_Utt(LISP args); LISP FT_MultiParse_Utt(LISP utt); void MultiParse(EST_Utterance &u); void add_feature_string(EST_Relation &r, const EST_String &fname, const EST_String &f); void add_monotone_targets(EST_Utterance &u, float start_f0, float end_f0); void clear_feature(EST_Relation &r, const EST_String &name); typedef float (*local_cost_function)(const EST_Item *item1, const EST_Item *item2); void subword_phonology(EST_Utterance &word); void lex_to_phones(EST_Utterance &u, const EST_String &relname); bool dp_match(const EST_Relation &lexical, const EST_Relation &surface, EST_Relation &match, local_cost_function lcf, EST_Item *null_syl); //bool dp_match(EST_Relation &a, EST_Relation &b, EST_Relation &c, // local_cost_function lcf, const EST_String &null_sym); void lex_to_phones(const EST_String &name, const EST_String &pos, EST_Relation &phone); void subword_metrical_tree(EST_Relation &syllable, EST_Relation &metricaltree); int syllabify_word(EST_Item *nw, EST_Relation &phone, EST_Relation &sylstructure, EST_Relation &syl, int flat); void subword_metrical_tree(EST_Item *w, EST_Relation &syllable, EST_Relation &metricaltree); /*void phonemic_trans(EST_Relation &trans) { EST_Item *s, *n; EST_String a; // cout << "trans: " << trans << endl; for (s = trans.head(); s; s = s->next()) { n = s->next(); // cout << *s; if (s->S("name").contains("cl")) { a = s->S("name").before("cl"); if ((next(s) != 0) && (next(s)->S("name") == a)) trans.remove_item(s); else if ((next(s) != 0) && (a == "dcl" ) && (next(s)->S("name") == "jh")) trans.remove_item(s); else if ((next(s) != 0) && (a == "tcl" ) && (next(s)->S("name") == "ch")) trans.remove_item(s); else s->set("name", a); // cout << "here1: " << a << "\n"; // s->set("name", s->S("name").before("cl")); } } } */ EST_Item *prev_match(EST_Item *n) { EST_Item *p = n->prev(); if (p == 0) return 0; if (daughter1(p->as_relation("Match")) == 0) prev_match(p); return daughter1(p->as_relation("Match")); } void insert_schwa(EST_Item *n) { EST_Item *p, *s; float pp_end = 0; float schwa_length = 0.01; if ((p = prev_match(n)) == 0) { cout << "Couldn't insert dummy schwa after " << *n << endl; return; } p = p->as_relation("SurfacePhone"); pp_end = (prev(p) != 0) ? prev(p)->F("end",0.0) : 0.0; s = p->insert_after(); s->set("name", "ax"); s->set("stress_num", "0"); if ((p->F("end",0) - pp_end) < schwa_length) schwa_length = p->F("dur") / 2.0; s->set("end", p->F("end",0)); p->set("end", p->F("end",0) - schwa_length); s->set("start", p->F("end",0)); s->set("df", phone_def.A("ax")); // cout << "end 1:" << s->f("end") << endl; // cout << "end 2:" << p->f("end") << endl; } void add_initial_silence(EST_Relation &lexical, EST_Relation &surface, EST_Relation &match) { EST_Item *s, *p, *n, *m; s = lexical.head(); if ((s->f("name") != "pau") && (s->f("name") != "sil")) { p = s->insert_before(); p->set("name", "pau"); p->set("df", phone_def.A("pau")); n = surface.head(); if ((n->f("name") == "pau") || (n->f("name") == "sil")) { m = match.head()->insert_before(p); m->append_daughter(n); } } } void add_even_segment_times(EST_Item *w, EST_Relation &phone) { EST_Item *s; int i; float start,dur=0,n,div; start = w->F("start"); dur = w->F("end") - start; n = (float)phone.length(); div = dur/n; for (i = 0, s = phone.head(); s; s = s->next(), ++i) { s->set("start", start + div * (float) i); s->set("end", start + div * (float) (i + 1)); } } #if 0 static void add_trans_phrase_phrase(EST_Utterance &utt) { EST_Item *s, *t, *a, *r; EST_Item *first_accent = 0, *last_accent = 0; bool exist; // This looks insanely comlicated, but all it really does is // add phrase_start and phrase_end items to the root node of // each metrical tree and then places these in the right position // in the intonation relation. utt.create_relation("IntonationPhrase"); for (s = utt.relation("MetricalTree", 1)->head(); s; s = s->next()) { for (r = first_leaf_in_tree(s); r != next_leaf(last_leaf_in_tree(s)); r = next_leaf(r)) if (r->in_relation("IntonationSyllable")) { if (first_accent == 0) first_accent = parent(r->as_relation("IntonationSyllable")) ->as_relation("Intonation"); last_accent = parent(r->as_relation("IntonationSyllable")) ->as_relation("Intonation"); } exist = false; // cout << "\nroot node: " <<*s << endl; if (first_accent) { cout << "first accent: " << *first_accent << endl; a = first_accent->prev(); if (a->S("name","") != "phrase_start") a = first_accent->insert_before(); } else { if (a == 0) a = utt.relation("Intonation")->prepend(); else a = a->insert_after(); } if (a->S("name","") != "phrase_start" ) // i.e. its a new one { a->set("name", "phrase_start"); a->set("ev:f0", 100); } // re-write position as relative to metrical tree a->set("position", usf_int_start); // add this as daughter to root node t = utt.relation("IntonationPhrase")->append(s); t->append_daughter(a); exist = false; //cout << "appended phrase end\n"; if (last_accent) { cout << "last accent: " << *last_accent << endl; a = last_accent->next(); if (a->S("name","") != "phrase_end") a = last_accent->insert_after(); } else a = a->insert_after(); if (a->S("name","") != "phrase_end") { a->set("name", "phrase_end"); a->set("ev:f0", 100); } // re-write position as relative to metrical tree a->set("position", usf_int_end); // add this as daughter to root node t->append_daughter(a); //cout << "appended phrase start\n"; first_accent = 0; // trigger for first time operation of loop } // now join any other marked phrase_start/ends to intermediate // nodes in metrical tree. /* for (s = u.relation("Intonation", 1)->head(); s; s = s->next()) { if (!s->in_relation("IntonationPhrase") && !s->in_relation("IntonationSyllable")) { pos = s->F("position"); } */ } #endif void add_single_phrase(EST_Utterance &utt, EST_Item *t) { EST_Item *s=0, *a, *p; float pos, max, d = 0, start = 0.0; pos = t->F("time"); max = 100000.0; for (p = utt.relation("Syllable")->head(); p; p = p->next()) { if (t->S("name") == "phrase_end") d = fabs(pos - p->F("end")); else d = fabs(pos - start); if (d < max) { max = d; s = p; } start = p->F("end"); } /* if (s) cout << "joining syllable " << *s << endl; else cout << "No legal syllable " << endl; cout << "to " << *t << endl; cout << "d = " << d << endl; */ if (!s->in_relation("IntonationSyllable")) a = utt.relation("IntonationSyllable")->append(s); else a = s->as_relation("IntonationSyllable"); a->append_daughter(t); t->set("time_path", "IntonationSyllable"); t->set_function("position", "standard+unisyn_tilt_phrase_position"); } void add_times(EST_Relation &lexical, EST_Relation &surface, EST_Relation &match) { (void) surface; (void) match; EST_Item *s, *t, *p; float prev_end, inc, first_end, last_end; int i; // first pass, copy times as appropriate, and find first // and last defined ends // This is hacky and certainly won't work for many cases first_end = -1.0; prev_end = 0.0; last_end = 0.0; // cout << "surface: " << surface << endl; for (s = lexical.head(); s; s = s->next()) { if ((t = daughter1(s->as_relation("Match"))) != 0) { s->set("end", t->F("end")); s->set("start", t->F("start")); last_end = t->F("end"); if (first_end < 0.0) first_end = t->F("end"); } } if (!lexical.head()->f_present("end")) { lexical.head()->set("end", first_end / 2.0); lexical.head()->set("start", 0.0); } if (!lexical.tail()->f_present("end")) { lexical.tail()->set("end", last_end + 0.01); lexical.tail()->set("start", last_end); } for (s = lexical.head(); s; s = s->next()) { if (!s->f_present("end")) { // cout << "missing end feature for " << *s << endl; for (i = 1, p = s; p; p = p->next(), ++i) if (p->f_present("end")) break; inc = (p->F("end") - prev_end) / ((float) i); // cout << "inc is : " << inc << endl; // cout << "stop phone is " << *p << endl; for (i = 1; s !=p ; s = s->next(), ++i) { s->set("end", (prev_end + ((float) i * inc))); s->set("start", (prev_end + ((float) (i - 1 )* inc))); } } prev_end = s->F("end"); } } static void met_error(EST_Item *s) { cerr << "Illegally named daughters of metrical node\n" << "daughter1 : " << *daughter1(s) << endl << "daughter2 : " << *daughter2(s) << endl; EST_error(""); } void legal_metrical_tree(EST_Item *s) { if (s == 0) return; if ((daughter1(s) == 0) || (daughter2(s) == 0)) return; if ((daughter1(s)->S("MetricalValue") == "s") && (daughter2(s)->S("MetricalValue") != "w")) met_error(s); else if ((daughter1(s)->S("MetricalValue") == "w") && (daughter2(s)->S("MetricalValue") != "s")) met_error(s); else if ((daughter1(s)->S("MetricalValue") != "w") && (daughter1(s)->S("MetricalValue") != "s")) met_error(s); legal_metrical_tree(daughter1(s)); legal_metrical_tree(daughter2(s)); } void parse_words(EST_Utterance &utt) { utt.create_relation("Token"); FT_Classic_POS_Utt(siod(&utt)); FT_Classic_Phrasify_Utt(siod(&utt)); MultiParse(utt); utt.relation("Syntax")->remove_item_feature("pos_index"); utt.relation("Syntax")->remove_item_feature("pos_index_score"); utt.relation("Syntax")->remove_item_feature("phr_pos"); utt.relation("Syntax")->remove_item_feature("pbreak_index"); utt.relation("Syntax")->remove_item_feature("pbreak_index_score"); utt.relation("Syntax")->remove_item_feature("pbreak"); utt.relation("Syntax")->remove_item_feature("blevel"); utt.relation("Syntax")->remove_item_feature("prob"); } void binaryize_tree(EST_Item *t) { // terminating condition if (daughter1(t) == 0) return; // nodes with single children should be merged if (daughter2(t) == 0) { // cout << "Single daughter: " << *t << endl; EST_Item *d = daughter1(t); move_sub_tree(d, t); } for (EST_Item *p = daughter1(t); p; p = p->next()) binaryize_tree(p); } void binaryize_tree(EST_Utterance &utt, const EST_String &base_tree, const EST_String &new_tree) { utt.create_relation(new_tree); copy_relation(*utt.relation(base_tree), *utt.relation(new_tree)); for (EST_Item *p = utt.relation(new_tree)->head(); p; p = p->next()) binaryize_tree(p); } void syntax_to_metrical_words(EST_Utterance &utt) { utt.create_relation("MetricalWord"); // copy syntax tree while merging single daughter nodes binaryize_tree(utt, "Syntax", "MetricalWord"); // add strong and weak values apply_nsr(utt, "MetricalWord"); } void add_metrical_functions(EST_Utterance &utt) { // Note that we don't add "start" functions here as this depends on // pause behaviour add_feature_function(*utt.relation("Syllable"), "vowel_start", "unisyn_vowel_start"); add_feature_function(*utt.relation("Syllable"), "end", "standard+unisyn_leaf_end"); add_feature_function(*utt.relation("Syllable"), "start", "standard+unisyn_leaf_start"); for (EST_Item *s = utt.relation("Syllable")->head(); s; s = s->next()) s->set("time_path", "SylStructure"); EST_Features tf; tf.set_function("end", "standard+unisyn_leaf_end"); tf.set_function("start","standard+unisyn_leaf_start"); tf.set_function("dur","standard+unisyn_duration"); tf.set("time_path", "MetricalTree"); tf.set("time_path", "MetricalTree"); // add_non_terminal_features(*utt.relation("MetricalTree"), tf); tf.set("time_path", "SylStructure"); add_non_terminal_features(*utt.relation("SylStructure"), tf); add_feature_function(*utt.relation("Segment"), "dur", "standard+duration"); } void auto_metrical_lex(EST_Utterance &utt) { LISP l_pdef; utt.create_relation("Syllable"); utt.create_relation("Segment"); l_pdef = siod_get_lval("darpa_fs", NULL); lisp_to_features(l_pdef, phone_def); mettree_add_words(utt); LISP lt = siod_get_lval("us_base_int", NULL); EST_String base_int; if (lt == NIL) base_int = "Syllable"; else { const char *x = get_c_string(lt); base_int = x; } // add_end_silences(*utt.relation("Segment")); add_metrical_functions(utt); } void extend_tree(EST_Item *m, EST_Item *p, const EST_String &terminal, const EST_String &second_tree) { EST_Item *d, *e; if (!daughter1(m)) { if (m->in_relation(terminal)) // ie. really hit the bottom return; m = m->as_relation(second_tree); // swap to a new tree } for (d = daughter1(m); d; d = d->next()) { e = p->append_daughter(d); extend_tree(d, e, terminal, second_tree); } } static void nsr(EST_Item *n) { EST_Item *left, *right; left = daughter1(n); right = daughter2(n); if (left == 0) return; else { nsr(left); left->set("MetricalValue","w"); } if (right == 0) return; else { nsr(right); right->set("MetricalValue","s"); } } static void apply_nsr(EST_Utterance &u, const EST_String &tree) { EST_Item *n; for (n = u.relation(tree)->head(); n; n = n->next()) nsr(n); } EST_Item *other_daughter(EST_Item *parent, EST_Item *daughter) { return (daughter1(parent) == daughter) ? daughter2(parent) : daughter1(parent); } static void stress_factor1(EST_Item *s, int max_depth) { EST_Item *a; EST_String val, pad; char *str; long n, i; float max; val = ""; for (a = s; parent(a); a = parent(a)) if (a->f("MetricalValue") == "s") val += "2"; else val += "0"; // cout << "\nSyllable " << s << " has value " << val << endl; if (val.length() < max_depth) for (pad = "", i = 0; i < (max_depth - val.length()); ++i) pad += "2"; val += pad; // cout << "Syllable " << s << " has padded value " << val << endl; str = strdup(val); max = pow(3.0, (float)max_depth) - 1.0; n = strtol(str, (char **)NULL, 3); // cout << "decimal value: " << n; // cout << " normalised: " << (float)n/max << endl; s->set("StressFactor1", ((float)n/max)); } EST_Item * find_apex(EST_Item *n, int &num_nodes) { EST_Item *p; p = parent(n); if (p == 0) return n; if (daughter2(p) == n) return find_apex(p, ++num_nodes); return p; } void find_leaf(EST_Item *n, int &num_nodes) { if (n == 0) return; find_leaf(daughter1(n), ++num_nodes); } static void phrase_factor(EST_Item &syl, const EST_String &met_name) { EST_Item *p; EST_String val, pad; int num_nodes = 1; // cout << "Terminal Syl = " << syl << " f:" << syl.f << endl; p = find_apex(syl.as_relation(met_name), num_nodes); // cout << "up nodes: " << num_nodes; // cout << "Apex = " << *p << endl; find_leaf(daughter2(p), num_nodes); // cout << " downp nodes: " << num_nodes << endl; syl.set("PhraseIndex", num_nodes); } static int max_tree_depth(EST_Utterance &u, const EST_String &base_stream, const EST_String &mettree) { EST_Item *s, *a; int depth; int max_depth = 0; for (s = u.relation(base_stream)->head(); s; s = s->next()) { depth = 0; for (a = s->as_relation(mettree); parent(a); a = parent(a)) ++depth; if (depth > max_depth) max_depth = depth; } return max_depth; } void stress_factor1(EST_Utterance &u, const EST_String &base_stream, const EST_String &mettree) { EST_Item *s; int max_depth = max_tree_depth(u, base_stream, mettree); for (s = u.relation(base_stream)->head(); s; s = s->next()) stress_factor1(s->as_relation(mettree), max_depth); } EST_Item *strong_daughter(EST_Item *n) { if (daughter1(n) == 0) return 0; return (daughter1(n)->f("MetricalValue") == "s") ? daughter1(n) : daughter2(n); } EST_Item *weak_daughter(EST_Item *n) { if (daughter1(n) == 0) return 0; return (daughter1(n)->f("MetricalValue") == "w") ? daughter1(n) : daughter2(n); } static void fill_mini_tree(EST_Item *s, int val) { if (s->f("MetricalValue") == "s") s->set("StressVal", val); else s->set("StressVal", 0); if (strong_daughter(s)) fill_mini_tree(strong_daughter(s), val); if (weak_daughter(s)) fill_mini_tree(weak_daughter(s), val - 1); } void stress_factor2(EST_Utterance &u, const EST_String &base_stream, const EST_String &mettree) { EST_Item *s; int sv = -1; float b; (void) base_stream; s = u.relation(mettree)->head(); fill_mini_tree(s, sv); // normalise values sv = 0; for (s = u.relation(base_stream)->head(); s; s = s->next()) sv = Lof(s->I("StressVal"), sv); cout << "Max Stress: " << sv << endl; for (s = u.relation(base_stream)->head(); s; s = s->next()) { b = (float)(s->I("StressVal") - sv + 1); if (s->f("MetricalValue") == "s") s->set("StressFactor2", (b / float(sv)) * -1.0); else s->set("StressFactor2", 0); } } void phrase_factor(EST_Utterance &u, const EST_String &base_stream, const EST_String &mettree) { EST_Item *s; float max_pf = 0; for (s = u.relation(base_stream)->head(); s; s = s->next()) phrase_factor(*s, mettree); for (s = u.relation(base_stream)->head(); s; s = s->next()) if (s->I("PhraseIndex") > max_pf) max_pf = s->I("PhraseIndex"); for (s = u.relation(base_stream)->head(); s; s = s->next()) { s->set("PhraseFactor", (float)s->I("PhraseIndex")/max_pf); // cout << *s << " pf = " << // s->F("PhraseFactor") << endl; } } #if 0 static void remove_punctuation(EST_Utterance &u) { // The syntactic grammar has unary rules for the preterminals // these would make the mtettrical tree have an extra layer // at the word level. So here we remove that extra layer EST_Item *w; EST_Item *a, *b, *c, *od; for (w = u.relation("Word")->head(); w != 0; w = w->next()) { if (w->f("pos") == "punc") { a = w->as_relation("Syntax"); b = parent(a); c = parent(b); od = other_daughter(c, b); remove_item(b, "Syntax"); move_sub_tree(od, c); remove_item(w, "Word"); } } } static void add_intonation(EST_Utterance &u, const EST_String &base_stream, float threshold) { EST_Item *e, *s; cout << "Threshold = " << threshold << endl; for (s = u.relation(base_stream)->head(); s; s = s->next()) { if (s->F("StressFactor") > threshold) { // cout << *s <<" **stress factor:" << s->F("StressFactor") << endl; e = u.relation("IntSyl")->append(); e->insert_below(s); e->set_name("Accent"); e->set("prominence", s->F("StressFactor")); u.relation("Intonation")->append(e); } } } #endif void add_monotone_targets(EST_Utterance &u, float start_f0, float end_f0) { EST_Item *t; float end; end = u.relation("Segment")->tail()->f("end"); cout << "Phone ends\n"; cout << *u.relation("Segment"); cout << "last position is :" << end << endl; u.create_relation("Target"); t = u.relation("Target")->append(); t->set("f0", start_f0); t->set("pos", 0.0); // temporary - should disappear when awb changes code // t->set("name", ftoString(start_f0)); // t->set("end", 0.0); t = u.relation("Target")->append(); t->set("f0", end_f0); t->set("pos", end); // temporary - should disappear when awb changes code // t->set("name", ftoString(end_f0)); // t->set("end", end); } static void mettree_add_words(EST_Utterance &u) { EST_Utterance word; EST_Item *w; word.create_relation("Word"); word.create_relation("Segment"); word.create_relation("SylStructure"); word.create_relation("Syllable"); word.create_relation("WordStructure"); for (w = u.relation("Word")->head(); w != 0; w = w->next()) { word.clear_relations(); cout << "N:"; cout << w->f("name") << " " << w->f("pos", "") << endl; lex_to_phones(w->f("name"), w->f("pos", "0"), *word.relation("Segment")); EST_Item *nw = word.relation("Word")->append(); nw->set("name", w->S("name")); syllabify_word(nw, *word.relation("Segment"), *word.relation("SylStructure"), *word.relation("Syllable"), 0); subword_metrical_tree(nw, *word.relation("Syllable"), *word.relation("WordStructure")); utterance_merge(u, word, w, word.relation("Word")->head()); } } void add_metrical_nodes(EST_Utterance &u, EST_Item *n, LISP lpos); EST_String strip_vowel_num(EST_String p) { if (p.contains(RXint)) p = p.before(RXint); return p; } static void percolate(EST_Item *start) { EST_Item *n; for (n = start; n; n = parent(n)) { // cout << "altering sister\n"; if (prev(n) != 0) prev(n)->set("MetricalValue", "w"); else if (next(n) != 0) next(n)->set("MetricalValue", "w"); } } void main_stress(EST_Item *s) { EST_Item *n; for (n = s; parent(n); n = parent(n)) n->set("MetricalValue", "s"); n = s; percolate(n); } void footing(EST_Item *n1) { EST_Item *n2, *n3, *n4, *p1, *p3, *r; r = parent(n1); // root node p1 = daughter2(r); n2 = daughter1(p1); n3 = daughter2(p1); if (p1 == 0) { cerr << "Error: Empty 3rd node after " << *n1 << " in footing\n"; return; } if (n2 == 0) { cerr << "Error: Empty 3rd node after " << *n1 << " in footing\n"; return; } if (n3 == 0) { cerr << "Error: Empty 3rd node after " << *n1 << " in footing\n"; return; } cout << "n1: " << *n1 << endl << endl; cout << "n2: " << *n2 << endl << endl; cout << "n3: " << *n3 << endl << endl; cout << "p1: " << *p1 << endl << endl; p3 = n1->insert_parent(); n1 = daughter1(p3); n4 = p3->append_daughter(); move_sub_tree(n2, n4); move_sub_tree(n3, p1); p3->set("MetricalValue", "w"); p3->set("Altered_a", "DONE_W"); n1->set("MetricalValue", "s"); n1->set("Altered_b", "DONE_S"); } #if 0 LISP FT_metrical_data(LISP lf_word, LISP lf_seg, LISP lf_int) { EST_Utterance word, *u = new EST_Utterance; EST_Relation phone; EST_Item *s, *p, *w, *nw, *n; EST_StrList plist; float phone_start, mid; LISP lutt; EST_Track fz; int i; u->create_relation("Word"); u->create_relation("Segment"); u->create_relation("Syllable"); u->create_relation("MetricalTree"); u->create_relation("LexicalMetricalTree"); u->create_relation("SurfacePhone"); u->create_relation("Surface"); u->create_relation("Intonation"); u->create_relation("IntonationSyllable"); EST_String segfile = get_c_string(lf_seg); EST_String wordfile = get_c_string(lf_word); if (u->relation("Word")->load(wordfile) != format_ok) { cerr << "Couldn't load file " << get_c_string(lf_word) << endl; festival_error(); } if ((segfile != "dummy") &&(u->relation("Segment")-> load(get_c_string(lf_seg)) != format_ok)) { cerr << "Couldn't load file " << get_c_string(lf_seg) << endl; festival_error(); } if (lf_int != NIL) if (u->relation("Intonation")->load(get_c_string(lf_int)) != format_ok) { cerr << "Couldn't load file " << get_c_string(lf_int) << endl; festival_error(); } u->f.set("fileroot", basename(wordfile, "*")); // cout << "Words: " << *u->relation("Word"); if (segfile != "dummy") phonemic_trans(*u->relation("Segment")); // u->relation("Intonation")->load(get_c_string(lf_int)); // tmp hack float prev_end = 0.0; for (w = u->relation("Word")->head(); w != 0; w = n) { n = w->next(); // w->set("start", prev_end); w->f_remove("end"); // prev_end = w->F("end"); if ((w->f("name") == "sil") || (w->f("name") == "pau")) u->relation("Word")->remove_item(w); } gc_protect(&lutt); lutt = siod_make_utt(u); cout << *u->relation("Word") << endl; FT_POS_Utt(lutt); FT_Phrasify_Utt(lutt); MultiParse(*u); // remove_punctuation(*u); // Copy Syntax tree into a new Metrical Tree copy_relation(*u->relations.val("Syntax"), *u->relations.val("MetricalTree")); // flatten preterminal unary rules flatten_preterminals(*u); apply_nsr(*u); copy_relation(*u->relations.val("MetricalTree"), *u->relations.val("LexicalMetricalTree")); word.create_relation("Word"); word.create_relation("Match"); word.create_relation("NewMatch"); word.create_relation("Segment"); word.create_relation("SurfacePhone"); word.create_relation("LexicalSylStructure"); word.create_relation("SurfaceSylStructure"); word.create_relation("LexicalSyllable"); word.create_relation("SurfaceSyllable"); word.create_relation("LexicalMetricalTree"); word.create_relation("SurfaceMetricalTree"); phone_start = 0.0; // Note starts are hardwired here because feature function thing // isn't fully operational and because deleting silence messes // it up. // u->save("zz_parse.utt", "est"); if (segfile != "dummy") { for (s = u->relation("Segment")->head(); s; s = s->next()) { s->set("start", phone_start); phone_start = s->F("end"); } phone_start = 0.0; s = u->relation("Segment")->head(); if ((s->f("name") == "pau") || (s->f("name") == "sil")) { w = u->relation("SurfacePhone")->append(); w->set("name", "pau"); w->set("end", s->F("end")); w->set("start", s->F("start")); } } // cout <<"Surface 1:" << *u->relation("SurfacePhone") << endl; for (i = 0, w = u->relation("Word")->head(); w != 0; w = w->next(), ++i) { word.clear_relations(); lex_to_phones(w->f("name"), w->f("pos"), *word.relation("Segment")); if (segfile == "dummy") *word.relation("SurfacePhone") = *word.relation("Segment"); else trans_to_phones(w, *u->relation("Segment"), *word.relation("SurfacePhone")); // cout << "lex phones: " << *word.relation("LexicalPhone") << endl; // cout << "sur phones: " << *word.relation("SurfacePhone") << endl; if (siod_get_lval("mettree_phones_debug", NULL) != NIL) { cout << "phones for word" << *w << endl; cout << *word.relation("SurfacePhone") << endl; } nw = word.relation("Word")->append(); nw->set("name", w->S("name")); syllabify_word(nw, *word.relation("LexicalPhone"), *word.relation("LexicalSylStructure"), *word.relation("LexicalSyllable")); subword_metrical_tree(nw, *word.relation("LexicalSyllable"), *word.relation("LexicalMetricalTree")); if (siod_get_lval("mettree_debug", NULL) != NIL) word.save("word_lex.utt", "est"); // copy_relation(*word.relation("LexicalMetricalTree"), // *word.relation("HackMT")); EST_Item xx; dp_match(*word.relation("LexicalPhone"), *word.relation("SurfacePhone"), *word.relation("Match"), local_cost, &xx); if (syllabify_word(nw, *word.relation("SurfacePhone"), *word.relation("SurfaceSylStructure"), *word.relation("SurfaceSyllable")) < 1) { cerr << "Pronuciation for \"" << w->S("name") << "\" doesn't contain a vowel: " << *word.relation("SurfacePhone") << endl; // festival_error(); } fix_syllables(nw, word); subword_metrical_tree(nw, *word.relation("SurfaceSyllable"), *word.relation("SurfaceMetricalTree")); if (siod_get_lval("mettree_debug_word", NULL) != NIL) word.save("word_dp.utt", "est"); if (siod_get_lval("mettree_debug_word", NULL) != NIL) if (get_c_int(siod_get_lval("mettree_debug_word", NULL)) == i) word.save("word_nth.utt", "est"); utterance_merge(*u, word, w, "LexicalMetricalTree"); } // u->save("zz_parse2.utt", "est"); // u->save("test.utt"); // cout <<"Surface 2:" << *u->relation("SurfacePhone") << endl; add_initial_silence(*u->relation("LexicalPhone"), *u->relation("SurfacePhone"), *u->relation("Match")); // cout <<"Surface 3:" << *u->relation("SurfacePhone") << endl; add_times(*u->relation("LexicalPhone"), *u->relation("SurfacePhone"), *u->relation("Match")); u->relation("LexicalPhone")->f.set("timing_style", "segment"); u->relation("SurfacePhone")->f.set("timing_style", "segment"); // u->relation("Word")->f.set("timing_style", "segment"); u->relation("LexicalSyllable")->f.set("timing_style", "segment"); u->relation("LexicalSyllable")->f.set("time_path", "LexicalSylStructure"); // u->relation("LexicalSylStructure")->f.set("timing_style", "segment"); // u->relation("LexicalSylStructure")->f.set("time_relation", // "LexicalPhone"); u->relation("LexicalMetricalTree")->f.set("timing_style", "segment"); // if (lf_int != NIL) // add_feature_function(*u->relation("LexicalSyllable"),"vowel_start", // vowel_start_time); // add_feature_function(*u->relation("LexicalPhone"), "start", // ff_start_time); // add_feature_function(*u->relation("SurfacePhone"), "start", // ff_start_time); // add_feature_function(*u->relation("SurfacePhone"), "dur", // duration_time); // add_feature_function(*u->relation("LexicalSyllable"),"end", leaf_end_time); EST_Features tf; tf.set("time_path", "LexicalMetricalTree"); tf.set("end", leaf_end_time); // add_feature_string(*u->relation("LexicalMetricalTree"), "time_path", // "LexicalMetricalTree"); // add_feature_string(*u->relation("LexicalSylStructure"), "time_path", // "LexicalSylStructure"); // add_non_terminal_features(*u->relation("LexicalMetricalTree"), tf); tf.set("time_path", "LexicalSylStructure"); add_non_terminal_features(*u->relation("LexicalSylStructure"), tf); // add_feature_function(*u->relation("LexicalSyllable"),"start", // ff_start_time); // add_feature_function(*u->relation("LexicalSyllable"),"dur", duration_time); // cout << "ADDED Features to phone\n\n"; // cout << *(u->relation("LexicalPhone")) << endl << endl; // cout << "ADDED Features\n\n"; // cout << *(u->relation("LexicalSyllable")); // cout << "\nfinished\n\n"; // if (lf_int != NIL) // add_trans_intonation(*u); // cout <<"Lexical 3:" << *u->relation("LexicalPhone") << endl; // end_to_dur(*u->relation("SurfacePhone")); // end_to_dur(*u->relation("LexicalPhone")); // cout <<"Lexical 3:" << *u->relation("LexicalPhone") << endl; // clear_feature(*u->relation("SurfacePhone"), "end"); // clear_feature(*u->relation("LexicalPhone"), "end"); if (siod_get_lval("mettree_debug", NULL) != NIL) u->save("met_data.utt", "est"); gc_unprotect(&lutt); // u->save("zz_parse3.utt", "est"); return lutt; } #endif festival/src/modules/UniSyn_phonology/syllabify.cc0000644000175000017500000004272111465010356021075 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black and Paul Taylor */ /* Date : February 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* An implementation of Metrical Tree Phonology */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "lexicon.h" extern EST_Features phone_def; static int nucleus_count(EST_Relation &phone) { int v = 0; for (EST_Item *l = phone.head(); l; l = l->next()) if (l->S("df.syllabic") == "+") ++v; return v; } static bool legal_c1(EST_Item *x) { return (x->S("df.syllabic") == "-"); } static bool legal_c2(EST_Item *c1, EST_Item *x) { if (x->S("df.syllabic") == "+") return false; if ((x->name() == "s") && ((c1->name() == "l") || (c1->name() == "w") || (c1->name() == "p") || (c1->name() == "t") || (c1->name() == "k") || (c1->name() == "m") || (c1->name() == "n") || (c1->name() == "r") || (c1->name() == "f"))) return true; if ((c1->S("df.manner") == "approximant") && (x->S("df.manner") == "stop") || (x->name() == "th") || (x->name() == "f")) // if (ph_is_semivowel(c1->name()) && (ph_is_stop(x->name()) // || (x->name() == "th") // || (x->name() == "f"))) { if (x->name() == "y") return false; if ((c1->name() == "l") && ((x->name() == "t") || (x->name() == "d") || (x->name() == "th"))) return false; if ((c1->name() == "w") && ((x->name() == "p") || (x->name() == "b") || (x->name() == "f"))) return false; return true; } // for "vroom" if ((c1->name() == "r") && (x->name() == "v")) return true; return false; } static bool legal_c3(EST_Item * c1, EST_Item * c2, EST_Item *pos) { (void) c1; if (pos->name() != "s") return false; if ((c2->S("df.manner") == "stop") && (c2->S("df.voicing") == "-")) return true; return false; } EST_Item *make_onset(EST_Item *syl_struct_root, EST_Item *nucleus, int flat) { EST_Item *c1, *c2, *c3; EST_Item *p, *onset; // if first syllable in word, put all prevocalic segments in onset if ((prev(syl_struct_root) == 0) && prev(nucleus)) { if (flat) onset = syl_struct_root; else { onset = daughter1(syl_struct_root)->insert_before(); onset->set("sylval", "Onset"); } // tmpeorary hack because of lack of prepend daughter fn. EST_Item *s; for (s = nucleus->prev(); prev(s); s = s->prev()); for (c1 = s; c1 != nucleus; c1 = next(c1)) onset->append_daughter(c1); return 0; } c1 = prev(nucleus); if (c1 == 0) return c1; // if (ph_is_vowel(c1->name())) // return next(c1); if (c1->S("df.syllabic") == "+") return next(c1); // if (c1->S("df.type") == "vowel") // return next(c1); if (flat) onset = syl_struct_root; else { onset = daughter1(syl_struct_root)->insert_before(); onset->set("sylval", "Onset"); } // add first consonant if (legal_c1(c1)) p = onset->append_daughter(c1); else return nucleus; // add second consonant c2 = prev(c1); if (c2 == 0) return 0; if (legal_c2(c1, c2)) p = p->insert_before(c2); else return c1; // add third consonant (s) c3 = prev(c2); if (c3 == 0) return 0; // add third consonant // if (legal_c3(c1->name(), c2->name(), c3->name())) // p = p->insert_before(c3); if (legal_c3(c1, c2, c3)) p = p->insert_before(c3); else return c2; return c3; } static void make_nucleus(EST_Item *syl_struct_root, EST_Item *nucleus, int flat) { EST_Item *m; if (flat) m = syl_struct_root; else { // add rhyme m = syl_struct_root->append_daughter(); m->set("sylval", "Rhyme"); // add nucleus m = m->append_daughter(); m->set("sylval", "Nucleus"); } m->append_daughter(nucleus); } static void make_coda(EST_Item *first_coda, EST_Item *first_onset, EST_Item *syl_struct_root, int flat) { EST_Item *m=0;; if ((first_coda != 0) && (first_coda != first_onset)) { if (flat) m = syl_struct_root; else { m = daughter1(syl_struct_root); if (m->f("sylval") != "Rhyme") m = daughter2(syl_struct_root); m = m->append_daughter(); m->set("sylval", "Coda"); } } for (; (first_coda != 0) && (first_coda != first_onset); first_coda = first_coda->next()) m->append_daughter(first_coda); } // if "flat" is set to 1 a tree syllable structure is built, otherwise // a flat list like structure is built. int syllabify_word(EST_Item *w, EST_Relation &phone, EST_Relation &sylstructure, EST_Relation &syllable, int flat) { EST_Item *prev_syl, *this_syl, *l, *this_struct, *prev_struct; EST_Item *first_onset, *first_coda = 0; EST_String v; // cout << "phones: " << phone << endl; int count = nucleus_count(phone); prev_struct =first_onset = 0; if (count < 1) return 0; for (prev_syl = 0, l = phone.head(); l; l = l->next()) { // cout << "type " << l->S("name") << ": " << l->S("df.type") << endl; if (l->S("df.syllabic") == "+") // if (ph_is_vowel(l->name())) { if (count == 1) this_syl = syllable.append(w); else this_syl = syllable.append(); this_struct = sylstructure.append(this_syl); this_syl->set("stress_num", l->I("stress_num")); // note: it must be in this order // cout << "this struct: " << *this_syl << endl; // cout << "this struct: " << *this_struct << endl; make_nucleus(this_struct, l, flat); first_onset = make_onset(this_struct, l, flat); make_coda(first_coda, first_onset, prev_struct, flat); prev_syl = this_syl; prev_struct = this_struct; first_coda = l->next(); } } make_coda(first_coda, first_onset, prev_struct, flat); return count; } void fix_lex_string(LISP lpos, EST_StrList &s) { LISP a, b, c; EST_String p; for (a = car(lpos); a != NIL; a = cdr(a)) { // cout << "1:\n"; // lprint(a); // for (b = car(a); b != NIL; b = cdr(b)) // { b = car(a); // cout << "0:\n"; // lprint(b); for (c = car(b); c != NIL; c = cdr(c)) { p = get_c_string(car(c)); if (ph_is_vowel(p)) p += "0"; s.append(p); } } // cout << "def list: " << s << endl; } //Adds phoneme name to syllable as a string void add_syllable_name(EST_Item *syl, const EST_String &fname) { EST_Item *s, *p; s = syl->as_relation("SylStructure"); for (p = first_leaf_in_tree(s); p != next_leaf(last_leaf_in_tree(s)); p = next_leaf(p)) if (p == first_leaf_in_tree(s)) s->set(fname, p->S("name")); else s->set(fname, s->S(fname) + " " + p->S("name")); } void lex_to_phones(const EST_String &name, const EST_String &pos, EST_Relation &phone) { LISP entry, lpos; EST_Item *p; EST_StrList lex_def; EST_String lex_phone; if (pos != "0") lpos = rintern(pos); else lpos = NIL; entry = lex_lookup_word(name, lpos); lpos = cdr(cdr(entry)); if (!siod_atomic_list(car(lpos))) fix_lex_string(lpos, lex_def); else siod_list_to_strlist(car(lpos), lex_def); for (EST_Litem *sl = lex_def.head(); sl; sl = sl->next()) { p = phone.append(); lex_phone = lex_def(sl); if (lex_phone.contains(RXint)) { p->set("name", lex_phone.before(RXint)); p->set("stress_num", lex_phone.at(RXint)); } else p->set("name", lex_phone); // df = "distinctive features" if (phone_def.present(p->S("name"))) p->set("df", phone_def.A(p->S("name"))); else EST_error("Word \"%s\" has illegal phoneme \"%s\" in lexicon\n", (const char *)name, (const char *)p->S("name")); } } /*static bool legal_c3(EST_String c1, EST_String c2, EST_String pos) { (void) c1; if (pos != "s") return false; if (ph_is_stop(c2) && (!ph_is_voiced(c2))) return true; return false; } */ /*static int vowel_count(EST_StrList &full) { int v = 0; EST_Litem *l; for (l = full.head(); l; l = l->next()) if (ph_is_stress_vowel(full(l))) ++v; return v; } static bool ph_is_s(const EST_String &p) { return (p == "s") ? true : false; } */ /*void subword_phonology(EST_Utterance &word) { word.create_relation("MetricalTree"); word.create_relation("SylStructure"); word.create_relation("Syllable"); syllabify_word(word); subword_metrical_tree(word); if (siod_get_lval("mettree_debug", NULL) != NIL) word.save("word.utt", "est"); } */ /*bool vowel(EST_String p) { if (p.contains(RXint)) p = p.before(RXint); return ph_is_vowel(p); } */ // Add phones, and sylstructure for a single syllable /*bool met_node_is_leaf(EST_Item *met_node) { return met_node->in_relation("Syllable"); } */ /*static int sonority(EST_String p) { if (p.contains(RXint)) p = p.before(RXint); if (ph_is_vowel(p)) return 6; if (ph_is_liquid(p) || ph_is_approximant(p)) return 5; if (ph_is_nasal(p)) return 4; if (ph_is_fricative(p) && (!ph_is_s(p))) return 3; if (ph_is_stop(p)) return 2; return 1; } // Parse arbitrary phone string with numbered vowels into // syllable, phone and sylstructure relations static bool ph_is_semivowel(EST_String c1) { return (ph_is_liquid(c1) || (ph_is_approximant(c1))); } static bool ph_is_s(EST_String c1) { return (c1 == "s"); } */ /*static bool ph_is_stress_vowel(EST_String p) { if (p.contains(RXint)) p = p.before(RXint); // cout << "p = " << p << endl; return (ph_is_vowel(p)); } static int vowel_count(EST_Relation &phone) { int v = 0; for (EST_Item *l = phone.head(); l; l = l->next()) if (ph_is_vowel(l->name())) ++v; return v; } */ /*static void syllabify_word(EST_Utterance &word) { EST_Item *prev_syl, *this_syl, *l; EST_Item *first_onset, *first_coda = 0; EST_String v; first_onset = 0; if (nucleus_count(*word.relation("Phone")) < 1) { cerr << "Error: Pronunciation for " << *(word.relation("Word")->head()) << " does not contain vowel\n"; festival_error(); } for (prev_syl = 0, l = word.relation("Phone")->head(); l; l = l->next()) { cout << "syl: " << l->S("name") << ": " << l->S("df.syllabic", 1) << endl; if (l->S("df.syllabic") == "+") { this_syl = word.relation("Syllable")->append(); word.relation("SylStructure")->append(this_syl); this_syl->set("stress_num", l->I("stress_num")); // note: it must be in this order make_nucleus(this_syl->as_relation("SylStructure"), l); first_onset = make_onset(this_syl->as_relation("SylStructure"),l); make_coda(first_coda, first_onset, prev_syl->as_relation("SylStructure")); prev_syl = this_syl; first_coda = l->next(); } } make_coda(first_coda, first_onset, prev_syl->as_relation("SylStructure")); } */ /*void convert_cmu_lex_to_utt(EST_Utterance &u, EST_Item *w) { LISP entry, lpos; EST_String pos; EST_Utterance word; EST_Item *nw, *p; EST_StrList lex_def; pos = w->f("pos"); if (pos != "0") lpos = rintern(pos); else lpos = NIL; entry = lex_lookup_word(w->name(), lpos); lprint(entry); lpos = cdr(cdr(entry)); printf("lpos\n"); lprint(lpos); word.create_relation("Word"); word.create_relation("MetricalTree"); word.create_relation("SylStructure"); word.create_relation("Syllable"); word.create_relation("Phone"); nw = word.relation("Word")->append(); nw->set("name", w->S("name")); // parse_lex_string(word, car(lpos)); EST_String def, lex_phone; // cout << "DEF\n"; lprint(lpos); // make phone relation siod_list_to_strlist(car(lpos), lex_def); for (EST_Litem *sl = lex_def.head(); sl; sl = sl->next()) { p = word.relation("Phone")->append(); lex_phone = lex_def(sl); if (lex_phone.contains(RXint)) { p->set("name", lex_phone.before(RXint)); p->set("stress_num", lex_phone.at(RXint)); } else p->set("name", lex_phone); } syllabify_word(word); // cout << "before MT: " << *w << " F:" << w->f << endl; // cout << "before MT: " << *(word.relation("Syllable")->head()->Info()) // << " F:" << word.relation("Syllable")->head()->Info()->f << endl; subword_metrical_tree(word); // temporary hack because single syllable words don't get a metrical // node // if ((word.relation("MetricalTree")->head() == 0) && // (word.relation("Syllable")->head() != 0)) // word.relation("MetricalTree")->append(word.relation("Syllable")->head()->Info()); if (siod_get_lval("mettree_debug", NULL) != NIL) word.save("word.utt", "est"); // u.save("before.utt", "est"); // cout << "before merge: " << *w << " F:" << w->f << endl; // cout << "before syl: " << *(word.relation("Syllable")->head()->Info()) // << " F:" << word.relation("Syllable")->head()->Info()->f << endl; utterance_merge(u, word, w, "MetricalTree"); // cout << "after merge: " << *w << " F:" << w->f << endl; // cout << "after syl: " << *(word.relation("Syllable")->head()->Info()) // << " F:" << word.relation("Syllable")->head()->Info()->f << endl; // u.save("after.utt", "est"); } */ /*static void subword_metrical_tree(EST_Utterance &word) { EST_Item *s; EST_Item *new_leaf; // cout << endl<< endl << *(word.relation("Word")->head()->Info()) << endl; // cout << "head: " << word.relation("Syllable")->head() << endl; // cout << "head: " << *(word.relation("Syllable")->head()->Info()) << endl; // cout << "pre foot iteration:" << *(s->Info()) << " stress: " << s->Info()->f("stress_num") << endl << endl; // absorb initial unstressed syllables for (s = word.relation("Syllable")->head(); s && (s->f("stress_num") == 0); s = s->next()) { // cout << "**1 syl:" << *s << endl; new_leaf = word.relation("MetricalTree")->append(s); new_leaf->set("MetricalValue", "w"); } // cout << "utt to now 1c: " << word << endl; // In a multi-syllable word if (next(word.relation("Syllable")->head())) { //s = word.relation("Syllable")->head(); // cout << "**2 syl:" << *s << endl; for (; s;) { cout << "**3 syl:" << *s << endl; cout << "foot iteration\n" << *s << endl << endl; new_leaf = word.relation("MetricalTree")->append(s); new_leaf->set("MetricalValue", "s"); // s = make_foot(new_leaf, next(s)); } } else if (s) // For single syllable words { // cout << "adding single node\n" << *s << endl << endl; new_leaf = word.relation("MetricalTree")->append(s); // cout << "added node\n" << *s << endl << endl; } // cout << "utt to now 2: " << word << endl; if (siod_get_lval("mettree_debug", NULL) != NIL) word.save("sub_word.utt", "est"); s = word.relation("MetricalTree")->head(); // make_super_foot(s, next(s)); if (siod_get_lval("mettree_debug", NULL) != NIL) word.save("super_foot.utt", "est"); // all_stress(word); } */ festival/src/modules/UniSyn_phonology/subword.cc0000644000175000017500000001627511465010356020571 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black and Paul Taylor */ /* Date : February 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* An implementation of Metrical Tree Phonology */ /* */ /*=======================================================================*/ #include "festival.h" void main_stress(EST_Item *s); void subword_metrical_tree(EST_Relation &syllable, EST_Relation &metricaltree); EST_Item *make_foot(EST_Item *w, EST_Item *met_node, EST_Item *next_syl_node); void subword_metrical_tree(EST_Item *w, EST_Relation &syllable, EST_Relation &metricaltree); static void all_stress(EST_Relation &syllable, EST_Relation &mettree); static void make_super_foot(EST_Item *w, EST_Item *met_node, EST_Item *next_syl_node); // Note: this function doesn't work as it should: currently a // monosyllabic word ends up being the same item as the syllable, when // a single daughter item would be better. void subword_list(EST_Item *w, EST_Relation &syllable, EST_Relation &metricaltree) { EST_Item *s, *n; n = metricaltree.append(w); if (next(syllable.head()) == 0) return; for (s = syllable.head(); s ; s = s->next()) { cout << "appending syl\n"; n->append_daughter(s); } } void subword_metrical_tree(EST_Item *w, EST_Relation &syllable, EST_Relation &metricaltree) { EST_Item *s; EST_Item *new_leaf; // single syllable if (next(syllable.head()) == 0) { new_leaf = metricaltree.append(w); return; } // absorb initial unstressed syllables for (s = syllable.head(); s && (s->f("stress_num") == 0); s = s->next()) { new_leaf = metricaltree.append(s); new_leaf->set("MetricalValue", "w"); } while (s) { new_leaf = metricaltree.append(s); new_leaf->set("MetricalValue", "s"); s = make_foot(w, new_leaf, next(s)); } if (siod_get_lval("mettree_debug", NULL) != NIL) metricaltree.utt()->save("foot.utt", "est"); s = metricaltree.head(); make_super_foot(w, s, next(s)); if (siod_get_lval("mettree_debug", NULL) != NIL) metricaltree.utt()->save("super_foot.utt", "est"); all_stress(syllable, metricaltree); } EST_Item *make_foot(EST_Item *w, EST_Item *met_node, EST_Item *next_syl_node) { EST_Item *new_parent; EST_Item *fl; if (next_syl_node == 0) return 0; if (next_syl_node->f("stress_num") == 0) { met_node->set("MetricalValue", "s"); next_syl_node->set("MetricalValue", "w"); fl = first_leaf(met_node); if (next(next_syl_node)) new_parent = met_node->insert_parent(); else { if (prev(fl)) new_parent = met_node->insert_parent(); else { // cout << "making met node word node in foot\n"; // cout << "foot root:" << *w << endl; new_parent = met_node->insert_parent(); // new_parent = met_node->insert_above(w); merge_item(new_parent, w); // cout << "foot root:" << *w << endl; // cout << "foot root:" << *new_parent << endl; } } new_parent->append_daughter(next_syl_node); next_syl_node = make_foot(w, new_parent, next(next_syl_node)); } return next_syl_node; } // construct left branching unlabelled tree using feet roots as terminals static void make_super_foot(EST_Item *w, EST_Item *met_node, EST_Item *next_syl_node) { EST_Item *new_parent; if (next_syl_node == 0) return; // make sure root node is w, i.e. word if (next(next_syl_node)) new_parent = met_node->insert_parent(); else { // cout << "inserted word as root in super foot:" << *w << endl; new_parent = met_node->insert_parent(); // KTH this crashes in linux merge_item(new_parent, w); // cout << "after inserted word as root in super foot:" << *w << endl; // cout << "after inserted word as root in super foot:" << // *new_parent << endl; // w = new_parent->as_relation("Word"); } new_parent->append_daughter(next_syl_node); make_super_foot(w, new_parent, next(new_parent)); } static void all_stress(EST_Relation &syllable, EST_Relation &mettree) { EST_Item *n, *s; int stress_num = -1; for (s = syllable.head(); s; s = s->next()) if (s->I("stress_num",0) > stress_num) stress_num = s->I("stress_num"); // cout << "max stress num:" << stress_num << endl; for (; stress_num > 0; --stress_num) { for (s = syllable.head(); s; s = s->next()) if (s->I("stress_num",0) == stress_num) break; if (s == 0) { cerr << "No main stress found in definition of lexical entry\n"; festival_error(); } n = s->as_relation(mettree.name()); main_stress(n); } } festival/src/modules/UniSyn_phonology/UniSyn_build.cc0000644000175000017500000006723211465010356021507 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Paul Taylor */ /* Date : June 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* Metrical Tree based Phonology system */ /* */ /*=======================================================================*/ #include #include #include #include #include "festival.h" #include "../UniSyn/us_features.h" void merge_features(EST_Item *from, EST_Item *to, int keep_id); void insert_schwa(EST_Item *n); extern EST_Features phone_def; void subword_metrical_tree(EST_Item *w, EST_Relation &syllable, EST_Relation &metricaltree); void lex_to_phones(const EST_String &name, const EST_String &pos, EST_Relation &phone); void trans_to_phones(EST_Item *w, EST_Relation &trans, EST_Relation &phone); void fix_syllables(EST_Item *nw, EST_Utterance &word); typedef float (*local_cost_function)(const EST_Item *item1, const EST_Item *item2); void add_metrical_functions(EST_Utterance &utt); bool dp_match(const EST_Relation &lexical, const EST_Relation &surface, EST_Relation &match, local_cost_function lcf, EST_Item *null_syl); float local_cost(const EST_Item *s1, const EST_Item *s2); void add_times(EST_Relation &lexical, EST_Relation &surface, EST_Relation &match); void add_initial_silence(EST_Relation &lexical, EST_Relation &surface, EST_Relation &match); void subword_metrical_tree(EST_Relation &syllable, EST_Relation &metricaltree); void add_metrical_functions(EST_Utterance &utt); int syllabify_word(EST_Item *nw, EST_Relation &phone, EST_Relation &sylstructure, EST_Relation &syl, int flat); void add_even_segment_times(EST_Item *w, EST_Relation &phone); void lex_to_phones(EST_Utterance &u, const EST_String &relname); void phonemic_trans(EST_Relation &trans); void add_single_phrase(EST_Utterance &utt, EST_Item *t); LISP FT_add_trans_metrical_tree(LISP l_utt, LISP lf_input, LISP lf_output); #if 0 static void add_trans_phrase(EST_Utterance &utt, const EST_String &i_name, const EST_String &s_name); #endif float local_cost(const EST_Item *s1, const EST_Item *s2) { float insertion_cost = get_c_int(siod_get_lval("met_insertion", NULL)); float deletion_cost = get_c_int(siod_get_lval("met_deletion", NULL)); float substitution_cost = get_c_int(siod_get_lval("met_substitution", NULL)); EST_String null_sym = "nil"; // otherwise cost is either insertion cost, or cost_matrix value if (s1->name() == s2->name()) return 0; else { if (s1->name() == null_sym) return insertion_cost; else if (s2->name() == null_sym) return deletion_cost; else return substitution_cost; } } void trans_to_phones(EST_Item *w, EST_Relation &trans, EST_Relation &phone) { int prev_phone; EST_Item *t, *p; int r; prev_phone = w->prev() ? prev(w)->I("phon_ref") : -1; r = w->I("phon_ref"); for (t = trans.head(); t; t = t->next()) { if ((t->f("name") == "sil") || (t->f("name") == "pau")) continue; if ((t->I("ref") > prev_phone) && (t->I("ref") <= r)) { p = phone.append(); p->set("name", t->S("name")); p->set_val("end", t->f("end")); p->set_val("start", t->f("start")); p->set("df", phone_def.A(p->S("name"))); } } } void add_trans_intonation(EST_Utterance &utt, const EST_String &i_name, const EST_String &s_name, int add_words) { EST_Item *s, *w, *t, *a, *b; EST_String wref; int s_num; EST_String w_num; EST_String is_name = i_name + s_name; utt.relation(i_name)->f.set("intonation_style", "tilt"); utt.create_relation(is_name); cout << "created : " << is_name << endl; // optional feature to add intonation events to words rather than syllables if (add_words) utt.create_relation("IntonationWord"); for (t = utt.relation(i_name, 1)->head(); t; t = t->next()) { t->f_remove("end"); if (!t->f_present("word_ref")) add_single_phrase(utt, t); else { w_num = t->S("word_ref"); s_num = t->I("syl_num"); for (w = utt.relation("Word", 1)->head(); w; w = w->next()) { if (w->S("id") == w_num) break; } if (w == 0) { cerr << "Error: couldn't find word ref " << endl; cerr << "For intonation event " << *t << endl; festival_error(); } if (add_words) { if (!w->in_relation("IntonationWord")) b = utt.relation("IntonationWord")->append(w); else b = w->as_relation("IntonationWord"); b->append_daughter(t); } // cout << "matching word: " << w->name() << endl; if ((b = w->as_relation("WordStructure")) == 0) EST_error("Item is not in WordStructure\n"); if ((s = nth_leaf(b, s_num)) == 0) { cerr << "Intonation element " << *t << "\nis linked to syllable " << s_num << " but word \"" << w->S("name") << "\"" " has only " << num_leaves(b) << " syllables\n"; } // cout << "here is s\n"; // cout << "matching syllable: " << *s << endl; if (!s->in_relation(is_name)) a = utt.relation(is_name)->append(s); else a = s->as_relation(is_name); a->append_daughter(t); // cout << "s1: " << s->S("id", "XX") << endl; s = s->as_relation(s_name); if (s == 0) cerr << "Syllable with id " << nth_leaf(b, s_num)->S("id") << "exists " "but is not in syllable relation. Suspect corrupted " "lexical conversion\n"; // cout << "s2: " << s->S("id", "XX") << endl; // change to relative positions if not already specified if (!t->f_present("rel_pos")) t->set("rel_pos", t->F("time") - s->F("vowel_start")); t->set("time_path", is_name); t->set_function("time", "standard+unisyn_tilt_event_position"); // cout << "end syl:" << endl; t->f_remove("word_ref"); t->f_remove("syl_num"); } } // add_trans_phrase(utt, i_name, s_name); } void syl_to_word_intonation(EST_Utterance &utt) { EST_Item *s, *w, *t=0, *b; utt.create_relation("IntonationWord"); for (s = utt.relation("Syllable", 1)->head(); s; s = s->next()) { if (!s->in_relation("IntonationSyllable")) continue; w = root(s, "WordStructure"); if (w == 0) { cerr << "Error: couldn't find word ref " << endl; cerr << "For intonation event " << *t << endl; festival_error(); } if (!w->in_relation("IntonationWord")) b = utt.relation("IntonationWord")->append(w); else b = w->as_relation("IntonationWord"); for (t = daughter1(s->as_relation("IntonationSyllable")); t; t = t->next()) b->append_daughter(t); } } static bool legal_daughter(EST_Item *r, const EST_String &iname, const EST_StrList &valid) { if (!r->in_relation(iname)) return false; if (strlist_member(valid, daughter1(r->as_relation(iname))->S("name", ""))) return true; return false; } void intonation_diagnostics(EST_Utterance &ref, EST_Utterance &test, const EST_String &rel, const EST_StrList &valid) { EST_Item *r, *t; EST_String iname = "Intonation" + rel; for (r = ref.relation(rel, 1)->head(), t = test.relation(rel, 1)->head(); r && t; r = r->next(), t = t->next()) { if (legal_daughter(r, iname, valid) && legal_daughter(t, iname, valid)) t->set("i_status", "COR"); else if (legal_daughter(r, iname, valid) && (!legal_daughter(t, iname, valid))) t->set("i_status", "DEL"); else if (!legal_daughter(r, iname, valid) && legal_daughter(t, iname, valid)) t->set("i_status", "INS"); // else // t->set("i_status", "0"); } } #if 0 static void add_trans_phrase(EST_Utterance &utt, const EST_String &i_name, const EST_String &s_name) { EST_Item *s, *t, *a, *p; float pos, max, d; EST_String is_name = i_name + s_name; for (t = utt.relation(i_name, 1)->head(); t; t = t->next()) { if (t->in_relation(is_name)) continue; pos = t->F("time"); max = 100000.0; cout << "here 1\n"; for (p = utt.relation(s_name)->head(); p; p = p->next()) { if (t->S("name","0") == "phrase_end") d = fabs(pos - p->end()); else d = fabs(pos - p->start()); if (d < max) { max = d; s = p; } } a = utt.relation(is_name)->append(t); } } #endif LISP FT_add_trans_intonation(LISP l_utt, LISP lf_int, LISP l_add_words) { EST_String int_file = get_c_string(lf_int); EST_Utterance *u = get_c_utt(l_utt); EST_Relation lab; EST_Item *s, *n; int add_words = (l_add_words == NIL) ? 0 : 1; u->create_relation("Intonation"); if (lab.load(int_file) != format_ok) EST_error("Couldn't load file %s\n", (const char *) int_file); for (s = lab.head(); s; s = s->next()) { n = u->relation("Intonation")->append(); merge_features(n, s, 1); if (n->S("name") =="afb") n->set("name", "a"); else if (n->S("name") == "m") { n->set("name", "a"); n->set("minor", 1); } else if ((n->S("name") == "a") || (n->S("name") == "arb") || (n->S("name") == "rb") || (n->S("name") == "phrase_end") || (n->S("name") == "phrase_start") || (n->S("name") == "fb") ) // tmp check (awb) continue; else EST_error("Illegal intonation name \"%s\"\n", (const char *) n->S("name")); } add_trans_intonation(*u, "Intonation", "Syllable", add_words); return l_utt; } LISP FT_add_trans_word(LISP l_utt, LISP lf_word, LISP keep_times) { EST_String word_file = get_c_string(lf_word); EST_Utterance *u = get_c_utt(l_utt); EST_Relation lab; EST_Item *s, *n; float p_end = 0; u->create_relation("Word"); if (lab.load(word_file) != format_ok) EST_error("Couldn't load file %s\n", (const char *) word_file); for (s = lab.head(); s; s = s->next()) { s->set("start", p_end); p_end = s->F("end"); if ((s->S("name") == "pau") || (s->S("name") == "sil")) continue; n = u->relation("Word")->append(); merge_features(n, s, 0); if (keep_times == NIL) { n->f_remove("end"); n->f_remove("start"); } } return l_utt; } LISP FT_add_f0_points(LISP l_utt, LISP lf_f0) { EST_String f0_file = get_c_string(lf_f0); EST_Utterance *u = get_c_utt(l_utt); EST_Track f0; EST_Item *s; float prev_mid, next_mid; if (f0.load(f0_file) != format_ok) EST_error("Couldn't load file %s\n", (const char *) f0_file); for (s = u->relation("Segment")->head(); s; s = s->next()) { prev_mid = s->prev() ? (prev(s)->F("end") + prev(s)->F("start"))/2.0 : 0.0; next_mid = s->next() ? (next(s)->F("end") + next(s)->F("start"))/2.0 : 0.0; s->set("prev_mid_f0", f0.a(f0.index(prev_mid))); s->set("start_f0", f0.a(f0.index(s->F("start")))); s->set("mid_f0", f0.a(f0.index((s->F("end") + s->F("start"))/2.0))); s->set("end_f0", f0.a(f0.index(s->F("end")))); s->set("next_mid_f0", f0.a(f0.index(next_mid))); } return l_utt; } LISP FT_add_coefs(LISP l_utt, LISP lf_coef) { EST_String coef_file = get_c_string(lf_coef); EST_Utterance *u = get_c_utt(l_utt); EST_Track coef; EST_Item *s; float prev_mid, next_mid; EST_FVector *frame; cout << "loading\n"; if (coef.load(coef_file) != format_ok) EST_error("Couldn't load file %s\n", (const char *) coef_file); cout << "done\n"; frame = new EST_FVector; frame->fill(0.0); // special case for first frame. for (s = u->relation("Segment")->head(); s; s = s->next()) { prev_mid = s->prev() ? (prev(s)->F("end") + prev(s)->F("start"))/2.0 : 0.0; next_mid = s->next() ? (next(s)->F("end") + next(s)->F("start"))/2.0 : 0.0; frame = new EST_FVector; coef.copy_frame_out(coef.index((s->F("end") + s->F("start"))/2.0), *frame); s->set_val("mid_coef", est_val(frame)); frame = new EST_FVector; coef.copy_frame_out(coef.index(s->F("end")), *frame); s->set_val("end_coef", est_val(frame)); frame = new EST_FVector; coef.copy_frame_out(coef.index(s->F("start")), *frame); s->set_val("start_coef", est_val(frame)); frame = new EST_FVector; coef.copy_frame_out(coef.index(prev_mid), *frame); s->set_val("prev_mid_coef", est_val(frame)); frame = new EST_FVector; coef.copy_frame_out(coef.index(next_mid), *frame); s->set_val("next_mid_coef", est_val(frame)); } return l_utt; // imid = coef.index((s->F("end") + s->F("start"))/2.0); // iend = coef.index(s->F("end")); } LISP FT_add_xml_relation(LISP l_utt, LISP xml_file) { EST_Utterance *u, tmp; u = get_c_utt(l_utt); tmp.clear(); tmp.load(get_c_string(xml_file)); EST_Features::Entries p; for (p.begin(tmp.relations); p; ++p) { relation(p->v)->remove_item_feature("actuate"); relation(p->v)->remove_item_feature("estExpansion"); relation(p->v)->remove_item_feature("xml:link"); relation(p->v)->remove_item_feature("href"); relation(p->v)->remove_item_feature("show"); } utterance_merge(*u, tmp, "id"); return l_utt; } void fix_syllables(EST_Item *nw, EST_Utterance &word) { EST_Item *t, *n, *s, *m; if (word.relation("Syllable")->length() == word.relation("SurfaceSyllable")->length()) return; cout << "Word \"" << word.relation("Word")->head()->name() << "\" has " << word.relation("Syllable")->length() << " lexical syllables and " << word.relation("SurfaceSyllable")->length() << " surface syllables\n"; for (s = word.relation("Syllable")->head(); s; s = s->next()) { t = s->as_relation("SylStructure"); n = syl_nucleus(t); m = daughter1(n->as_relation("Match")); if (m == 0) insert_schwa(n->as_relation("Segment")); } word.relation("SylStructure")->clear(); word.relation("Syllable")->clear(); word.relation("Match")->clear(); syllabify_word(nw, *word.relation("Segment"), *word.relation("SylStructure"), *word.relation("Syllable"), 0); // syllabify_word(nw, *word.relation("SurfacePhone"), // *word.relation("SurfaceSylStructure"), // *word.relation("SurfaceSyllable")); EST_Item xx; dp_match(*word.relation("Segment"), *word.relation("SurfacePhone"), *word.relation("Match"), local_cost, &xx); } /* Add segment durations from file */ void add_trans_duration(EST_Utterance &utt, const EST_String &segfile) { EST_Utterance word; EST_Relation phone, lab; EST_Item *s, *n; EST_StrList plist; float phone_start; EST_Item xx; if (lab.load(segfile) != format_ok) EST_error("Couldn't load file %s\n", (const char *) segfile); phone_start = 0.0; utt.create_relation("LabelSegment"); utt.create_relation("Match"); for (s = lab.head(); s; s = s->next()) { if (!phone_def.present(s->S("name"))) EST_error("Phone %s is not defined in phone set\n", (const char *) s->S("name")); n = utt.relation("LabelSegment")->append(); merge_features(n, s, 1); n->set("start", phone_start); phone_start = s->F("end"); n->set("dur", n->F("end") - n->F("start")); } dp_match(*utt.relation("Segment"), *utt.relation("LabelSegment"), *utt.relation("Match"), local_cost, &xx); add_times(*utt.relation("Segment"), *utt.relation("LabelSegment"), *utt.relation("Match")); for (s = utt.relation("Segment")->head(); s; s = s->next()) { s->set("target_dur", (s->F("end") - s->F("start"))); s->f_remove("end"); s->f_remove("dur"); s->f_remove("start"); } } static void add_silences(EST_Utterance &utt,EST_Item *w) { EST_Item *s; int r; if (w == 0) // insert initial silence { s = utt.relation("LabelSegment")->head(); if (s->name() == "pau") { EST_Item *sil = utt.relation("Segment")->append(); sil->set("name","pau"); sil->set("start",s->F("start")); sil->set("end",s->F("end")); } return; } cout << "Looking at inserting\n"; // Intermeditate silences r = w->I("phon_ref"); for (s=utt.relation("LabelSegment")->head(); s; s=s->next()) { if (r == s->I("ref")) { if (next(s)->name() == "pau") { cout << "actually inserting\n"; EST_Item *sil = utt.relation("Segment")->append(); sil->set("name","pau"); sil->set("start",s->F("end")); sil->set("end",next(s)->F("end")); } return; } } } void add_trans_seg(EST_Utterance &utt, const EST_String &segfile) { EST_Utterance word; EST_Relation phone, lab; EST_Item *s, *w, *nw, *n; EST_StrList plist; float phone_start; LISP lutt; int i; LISP l_pdef; l_pdef = siod_get_lval("darpa_fs", NULL); lisp_to_features(l_pdef, phone_def); utt.create_relation("LabelSegment"); utt.create_relation("tmpSegment"); utt.create_relation("Syllable"); utt.create_relation("Segment"); utt.create_relation("WordStructure"); if (lab.load(segfile) != format_ok) EST_error("Couldn't load file %s\n", (const char *) segfile); phone_start = 0.0; for (s = lab.head(); s; s = s->next()) { if (!phone_def.present(s->S("name"))) EST_error("Phone %s is not defined in phone set\n", (const char *) s->S("name")); n = utt.relation("LabelSegment")->append(); // cout << "append ls id " << n->S("id") << endl; merge_features(n, s, 1); // cout << "keep ls id " << n->S("id") << endl; n->set("start", phone_start); phone_start = s->F("end"); } // phonemic_trans(*utt.relation("LabelSegment")); /* for (w = utt.relation("Word")->head(); w != 0; w = n) { n = w->next(); w->f_remove("end"); if ((w->f("name") == "sil") || (w->f("name") == "pau")) utt.relation("Word")->remove_item(w); } */ gc_protect(&lutt); word.create_relation("Word"); word.create_relation("Match"); word.create_relation("NewMatch"); word.create_relation("Segment"); word.create_relation("tmpSegment"); word.create_relation("SylStructure"); word.create_relation("Syllable"); word.create_relation("WordStructure"); // Note starts are hardwired here because feature function thing // isn't fully operational and because deleting silence messes // it up. /* s = utt.relation("LabelSegment")->head(); if ((s->f("name") == "pau") || (s->f("name") == "sil")) { w = utt.relation("SurfacePhone")->append(); w->set("name", "pau"); w->set("end", s->F("end")); w->set("start", s->F("start")); w->set("df", phone_def.A("pau")); } */ add_silences(utt,0); for (i = 0, w = utt.relation("Word")->head(); w != 0; w = w->next(), ++i) { word.clear_relations(); word.f.set("max_id", 0); cout << "word: " << *w << endl; lex_to_phones(w->f("name"), w->f("pos", ""), *word.relation("Segment")); trans_to_phones(w, *utt.relation("LabelSegment"), *word.relation("tmpSegment")); nw = word.relation("Word")->append(); nw->set("name", w->S("name")); syllabify_word(nw, *word.relation("Segment"), *word.relation("SylStructure"), *word.relation("Syllable"), 0); // subword_list(nw, *word.relation("Syllable"), // *word.relation("MetricalTree")); if (siod_get_lval("mettree_debug", NULL) != NIL) word.save("word_lex.utt", "est"); EST_Item xx; dp_match(*word.relation("Segment"), *word.relation("tmpSegment"), *word.relation("Match"), local_cost, &xx); // fix_syllables(nw, word); subword_metrical_tree(nw, *word.relation("Syllable"), *word.relation("WordStructure")); // cout << "C2\n"; if (siod_get_lval("mettree_debug_word", NULL) != NIL) word.save("word_dp.utt", "est"); if (siod_get_lval("mettree_debug_word", NULL) != NIL) if (get_c_int(siod_get_lval("mettree_debug_word", NULL)) == i) word.save("word_nth.utt", "est"); word.remove_relation("SurfaceSylStructure"); word.remove_relation("SurfaceMetrcialTree"); word.remove_relation("SurfaceSyllable"); //cout << "32\n"; EST_String wid = w->S("id"); utterance_merge(utt, word, w, word.relation("Word")->head()); w->set("id", wid); add_silences(utt,w); } cout << "time2\n"; // utt.save("test.utt"); /* add_initial_silence(*utt.relation("Segment"), *utt.relation("SurfacePhone"), *utt.relation("Match")); */ add_times(*utt.relation("Segment"), *utt.relation("LabelSegment"), *utt.relation("Match")); // utt.relation("Word")->f.set("timing_style", "segment"); // cout << "here d\n"; // add_feature_function(*utt.relation("SurfacePhone"), "dur", // usf_duration); add_metrical_functions(utt); // if silences aren't wanted we still have to build with them so that // start times before pauses are done properly. if (!siod_get_lval("unisyn_build_with_silences",NULL)) for (s = next(utt.relation("Segment")->head());s;s = s->next()) if ((prev(s)->S("name") != "pau") && (prev(s)->S("name") != "sil")) s->set_function("start", "standard+unisyn_start"); else utt.relation("Segment")->remove_item(prev(s)); utt.relation("Segment")->remove_item_feature("stress_num"); utt.relation("Word")->remove_item_feature("phon_ref"); utt.remove_relation("tmpSegment"); if (siod_get_lval("mettree_debug", NULL) != NIL) utt.save("met_data.utt", "est"); gc_unprotect(&lutt); cout << "here c\n"; } LISP FT_add_trans_seg(LISP l_utt, LISP lf_seg) { add_trans_seg(*get_c_utt(l_utt), get_c_string(lf_seg)); return l_utt; } LISP FT_add_trans_duration(LISP l_utt, LISP lf_seg) { add_trans_duration(*get_c_utt(l_utt), get_c_string(lf_seg)); return l_utt; } LISP FT_syl_to_word_intonation(LISP l_utt) { syl_to_word_intonation(*get_c_utt(l_utt)); return l_utt; } LISP FT_intonation_diagnostics(LISP l_ref, LISP l_test, LISP l_rel_name, LISP l_valid) { EST_StrList valid; siod_list_to_strlist(l_valid, valid); intonation_diagnostics(*get_c_utt(l_ref), *get_c_utt(l_test), get_c_string(l_rel_name), valid); return NIL; } /*LISP FT_metrical_data(LISP lf_word, LISP lf_seg, LISP lf_int, LISP lf_met) { EST_Utterance *u = new EST_Utterance; LISP l_utt = siod_make_utt(u); EST_String word_file = get_c_string(lf_word); u->f.set("fileroot", basename(word_file, "*")); if (lf_met) { if (siod_get_lval("us_xml_metrical_trees", NULL) != NIL) data_metrical_tree(*u, get_c_string(lf_met), "xml"); else data_metrical_tree(*u, get_c_string(lf_met), ""); } else syntax_metrical_tree(*u, get_c_string(lf_word)); if (lf_int) data_metrical_lex(*get_c_utt(l_utt), get_c_string(lf_seg), get_c_string(lf_int)); else data_metrical_lex(*get_c_utt(l_utt), get_c_string(lf_seg), ""); return l_utt; } */ /* Specific to s-expression weather. Should be replaced when metrical trees go into XML. void data_metrical_tree(LISP l, EST_Item *met_parent, EST_Relation &word) { EST_String mv, name; int id, phon_ref; LISP a; EST_Item *m; //cout << "full entry\n"; // lprint(l); //cout << "now parsing\n"; mv = get_c_string(car(l)); //cout << "adding node strength: " << mv << endl; // root nodes are added in calling routine. if (mv != "r") m = met_parent->append_daughter(); else m = met_parent; m->set("MetricalValue", mv); if (siod_atomic_list(cdr(l))) { // cout << "atomic cdr is: "; // lprint(cdr(l)); a = cdr(l); name = get_c_string(car(a)); phon_ref = get_c_int(car(cdr(a))); id = get_c_int(car(cdr(cdr(a)))); m->set("name", name); m->set("phon_ref", phon_ref); m->set("id", id); word.append(m); //cout << "adding " << name << " on id: " << id << endl; return; } //cout << "\ndoing left branch\n"; data_metrical_tree(car(cdr(l)), m, word); //cout << "\ndoing right branch\n"; data_metrical_tree(car(cdr(cdr(l))), m, word); } */ int lisp_tree_to_xml(ofstream &outf, LISP l) { int id; LISP a; EST_String mv; //cout << "full entry\n"; // lprint(l); //cout << "now parsing\n"; mv = get_c_string(car(l)); outf << "\n"; return 0; } else outf << ">\n"; if (lisp_tree_to_xml(outf, car(cdr(l)))) outf << "\n"; if (lisp_tree_to_xml(outf, car(cdr(cdr(l))))) outf << "\n"; return 1; } LISP FT_add_trans_metrical_tree(LISP l_utt, LISP lf_input, LISP lf_output) { EST_Utterance *utt; LISP lmet, l; EST_Item *m; utt = get_c_utt(l_utt); utt->create_relation("Word"); utt->create_relation("Token"); utt->create_relation("MetricalTree"); lmet = vload(get_c_string(lf_input), 1); ofstream outf; outf.open(get_c_string(lf_output)); outf << "\n"; outf << "]>\n"; outf << "n"; // lprint(lmet); // have to ensure that next id of build nodes is greater than // any in the file. This should be done properly sometime. utt->set_highest_id(10000); for (l = lmet; l ; l = cdr(l)) { m = utt->relation("MetricalTree")->append(); // cout << "\nNew Tree\n"; if (lisp_tree_to_xml(outf, car(l))) outf << "" << endl; } outf << "\n"; return l_utt; } festival/src/modules/UniSyn_phonology/us_duration.cc0000644000175000017500000001573311465010356021436 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black and Paul Taylor */ /* Date : February 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* An implementation of Metrical Tree Phonology */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "us_duration.h" float phone_z_score(const EST_String &p, float dur) { float mean, sd; mean = met_duration.val(p).val("mean"); sd = met_duration.val(p).val("sd"); return ((dur - mean) / sd); } void clear_feature(EST_Relation &r, const EST_String &name) { for (EST_Item *p = r.head(); p ; p = p->next()) p->f_remove(name); } // void dur_to_end(EST_Relation &r) // moved to UniSyn us_diphone.cc void end_to_dur(EST_Relation &r) { float prev_end = 0; for (EST_Item *p = r.head(); p ; p = p->next()) { p->set("dur", p->F("end") - prev_end); prev_end = p->F("end"); } } void assign_phone_z_scores(EST_Utterance &u, const EST_String &seg_name) { EST_Item *s; end_to_dur(*u.relation(seg_name)); for (s = u.relation(seg_name)->head(); s; s = s->next()) s->set("z_score", phone_z_score(s->f("name"), s->F("dur"))); } void promote_mean_z_score(EST_Utterance &u, const EST_String &st_name, const EST_String &syl_name) { EST_Item *p, *s, *l; float z, n; for (s = u.relation(syl_name)->head(); s; s = s->next()) { p = s->as_relation(st_name); z = 0.0; for (n = 1, l = first_leaf_in_tree(p); l!= last_leaf_in_tree(p); l = next_leaf(l), n += 1.0) z += l->F("z_score"); z += l->F("z_score"); z = z / n; s->set("m_z_score", z); // n = named_daughter(s->as_relation(st_name), "sylval", "Rhyme"); // n = daughter1(named_daughter(n, "sylval", "Nucleus")); // s->set("z_score", n->F("z_score")); } } void promote_vowel_z_score(EST_Utterance &u, const EST_String &st_name, const EST_String &syl_name) { EST_Item *n, *s; for (s = u.relation(syl_name)->head(); s; s = s->next()) { n = named_daughter(s->as_relation(st_name), "sylval", "Rhyme"); n = daughter1(named_daughter(n, "sylval", "Nucleus")); s->set("z_score", n->F("z_score")); } } // set everything to its phone's mean duration LISP FT_met_dur_predict_1(LISP lutt, LISP lrel) { EST_Utterance *utt = get_c_utt(lutt); EST_String rel = get_c_string(lrel); EST_Item *p; clear_feature(*utt->relation(rel), "dur"); clear_feature(*utt->relation(rel), "end"); for (p = utt->relation(rel)->head(); p ; p = p->next()) p->set("dur", met_duration.val(p->f("name")).F("mean")); cout << "dur end\n"; dur_to_end(*utt->relation(rel)); return lutt; } LISP FT_met_dur_predict_2(LISP lutt, LISP lrel) { EST_Utterance *utt = get_c_utt(lutt); EST_String rel = get_c_string(lrel); clear_feature(*utt->relation(rel), "dur"); clear_feature(*utt->relation(rel), "end"); for (EST_Item *p = utt->relation(rel)->head(); p ; p = p->next()) p->set("dur", 0.2); dur_to_end(*utt->relation(rel)); return lutt; } typedef float (*local_cost_function)(const EST_Item *item1, const EST_Item *item2); float local_cost(const EST_Item *s1, const EST_Item *s2); bool dp_match(const EST_Relation &lexical, const EST_Relation &surface, EST_Relation &match, local_cost_function lcf, EST_Item *null_syl); void add_times(EST_Relation &lexical, EST_Relation &surface, EST_Relation &match); LISP FT_nat_dur_predict(LISP lutt, LISP lrel_name, LISP llab_file) { EST_Utterance *utt = get_c_utt(lutt); EST_String rel_name = get_c_string(lrel_name); EST_String lab_file = get_c_string(llab_file); EST_Relation lab, *segment, match, *ulab, *umatch; utt->create_relation("Match"); utt->create_relation("Lab"); if (utt->relation("Lab")->load(lab_file) != format_ok) festival_error(); // if (lab.load(lab_file) != format_ok) // festival_error(); EST_Item xx; segment = utt->relation(rel_name); ulab = utt->relation("Lab"); umatch = utt->relation("Match"); clear_feature(*segment, "dur"); clear_feature(*segment, "end"); dp_match(*segment, *ulab, *umatch, local_cost, &xx); add_times(*segment, *ulab, match); utt->remove_relation("Match"); utt->remove_relation("Lab"); // dp_match(*segment, lab, match, local_cost, &xx); // add_times(*segment, lab, match); return lutt; } festival/src/modules/UniSyn_phonology/unisyn_tilt.cc0000644000175000017500000000742211465010356021457 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Paul Taylor */ /* Date : July 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* Tilt synthesis code */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "EST_tilt.h" #include "../UniSyn/us_features.h" void vowel_tilt_to_abs_tilt(EST_Utterance &u) { EST_Item *s, *t; float pos; for (t = u.relation("Intonation")->head(); t; t = t->next()) { if (t->as_relation("IntonationSyllable")) { s = daughter1(t->as_relation("IntonationSyllable")); s = s->as_relation("Syllable"); pos = t->F("tilt:position"); t->set("position", pos); // cout << "pos: " << t->fF("position") << " rel:" << //t->fF("rel_pos") << " vowel start:" << s->fF("vowel_start") //<< endl; } } } void tilt_to_f0(EST_Relation &intonation, EST_Relation &f0) { EST_Track *fz = new EST_Track; tilt_synthesis(*fz, intonation, 0.01, 0); fz->save("tilt.f0", "est"); EST_Item *t = f0.append(); t->set("name", "f0"); t->set_val("f0", est_val(fz)); } festival/src/modules/UniSyn_phonology/us_aux.cc0000644000175000017500000001220011465010356020370 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black and Paul Taylor */ /* Date : February 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* An implementation of Metrical Tree Phonology */ /* */ /*=======================================================================*/ #include #include "festival.h" #include "us_duration.h" int num_daughters(EST_Item *s) { EST_Item *d1, *dn, *d; d1 = daughter1(s); if (d1 == 0) return 0; dn = daughtern(s); int n = 1; for (d = d1; d != dn; d = d->next()) ++n; return n; } void phones_in_word(EST_Item *w, const EST_String &met_name, const EST_String &ss_name, const EST_String &seg_name, EST_Item **first_p, EST_Item **last_p) { EST_Item *s, *t; cout << "&word: " << w << endl; cout << "in relation: " << w->in_relation(met_name) << endl; w = w->as_relation(met_name); cout << "&word: " << w << endl; if (w == 0) EST_error("Word isn't in metrical tree\n"); cout << "word: " << *w << endl; cout << "d1: " << daughter1(w) << endl; if (daughter1(w)) cout << "*d1: " << daughter1(w) << endl; cout << "d2: " << daughter2(w) << endl; if (daughter2(w)) cout << "*d2: " << daughter2(w) << endl; *first_p = 0; for (s = first_leaf_in_tree(w); s != next_leaf(last_leaf_in_tree(w)); s = next_leaf(s)) { cout << "leaf: " << *s << endl; cout << "in ss relation: " << s->in_relation(ss_name) << endl; // cout << "relations: " << s->relations() << endl; cout << "first leaf: " << *first_leaf_in_tree(s->as_relation(ss_name)) << endl; cout << "last leaf: " << *last_leaf_in_tree(s->as_relation(ss_name)) << endl; for (t = first_leaf_in_tree(s->as_relation(ss_name)); t !=next_leaf(last_leaf_in_tree(s->as_relation(ss_name))); t = next_leaf(t)) { *last_p = t->as_relation(seg_name); cout << "phone: " << *t << endl; if (*first_p == 0) *first_p = t->as_relation(seg_name); } } cout << "word: " << *w << endl; cout << "first: " << **first_p << " last " << **last_p << endl; cout << "\n\n"; } #if 0 static EST_Item *onset(EST_Item * ss_root) { return (daughter1(ss_root)->f("sylval") == "Onset")? daughter1(ss_root) : 0; } static EST_Item *rhyme(EST_Item * ss_root) { return (daughter2(ss_root) == 0) ? daughter1(ss_root) : daughter2(ss_root); } static EST_Item *nucleus(EST_Item * ss_root) { return daughter1(rhyme(ss_root)); } static EST_Item *coda(EST_Item * ss_root) { return daughter2(rhyme(ss_root)); } #endif festival/src/modules/UniSyn_phonology/us_duration.h0000644000175000017500000000661507327271540021305 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1998 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black and Paul Taylor */ /* Date : February 1998 */ /*-----------------------------------------------------------------------*/ /* */ /* An implementation of Metrical Tree Phonology */ /* */ /*=======================================================================*/ #ifndef __UNISYN_H__ #define __UNISYN_H__ #include "festival.h" bool operator == (const EST_Features &a, const EST_Features &b); typedef EST_TList EST_UttList; typedef EST_TKVL EST_FeatureList; EST_Item *named_daughter(EST_Item *syl, const EST_String &fname, const EST_String &fval); void dur_to_end(EST_Relation &r); void end_to_dur(EST_Relation &r); float phone_z_score(const EST_String &p, float dur); extern EST_FeatureList met_duration; #endif festival/src/modules/UniSyn_phonology/unisyn_phonology.mak0000644000175000017500000000605007327271540022706 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1998 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## -------------------------------------------------------------------- ## ## Make definitions to include unisyn_phonology. ## ## ## ########################################################################### INCLUDE_UNISYN_PHONOLOGY=1 MOD_DESC_Unisyn_phonology=Phonology for Unisyn ifeq ($(DIRNAME),src/modules) EXTRA_LIB_BUILD_DIRS := $(EXTRA_LIB_BUILD_DIRS) UniSyn_phonology endif festival/src/modules/MultiSyn/Makefile0000644000175000017500000000751310452241416016467 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## (University of Edinburgh, UK) and ## ## Korin Richmond ## ## Copyright (c) 2002 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### # # # Contains implementation of Generalis[ed/able] synthesis # # (like unit selection) # # # # Author: Korin Richmond (korin@cstr.ed.ac.uk) Aug 2002 # ########################################################################### TOP=../../.. DIRNAME=src/modules/MultiSyn LIB_BUILD_DIRS = inst_tmpl BUILD_DIRS = $(LIB_BUILD_DIRS) ALL_DIRS = $(BUILD_DIRS) H = UnitSelection.h \ VoiceBase.h DiphoneUnitVoice.h \ VoiceModuleBase.h DiphoneVoiceModule.h \ EST_TargetCost.h TargetCostRescoring.h \ EST_JoinCost.h EST_JoinCostCache.h \ DiphoneBackoff.h safety.h EST_DiphoneCoverage.h \ EST_FlatTargetCost.h SRCS = UnitSelection.cc \ VoiceBase.cc DiphoneUnitVoice.cc \ VoiceModuleBase.cc DiphoneVoiceModule.cc \ EST_TargetCost.cc TargetCostRescoring.cc \ EST_JoinCost.cc EST_JoinCostCache.cc \ DiphoneBackoff.cc EST_DiphoneCoverage.cc \ EST_FlatTargetCost.cc OBJS = $(SRCS:.cc=.o) FILES = Makefile $(SRCS) $(H) LOCAL_INCLUDES = -I../include INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/MultiSyn/UnitSelection.cc0000644000175000017500000006327112442333465020135 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: August 2002 */ /* --------------------------------------------------------------------- */ /* Generalis[ed/able] Unit Selection synthesis */ /* */ /*************************************************************************/ #include "siod.h" #include "EST.h" #include "UnitSelection.h" #include "DiphoneUnitVoice.h" #include "DiphoneVoiceModule.h" #include "EST_JoinCost.h" #include "EST_TargetCost.h" #include "EST_FlatTargetCost.h" #include "safety.h" static LISP FT_voice_get_units(LISP l_voice, LISP l_utt) { EST_Utterance *u = get_c_utt(l_utt); VoiceBase *v = voice( l_voice ); // Find units and put in utterance v->getUnitSequence( u ); return l_utt; } /////////////////////////////////////////////////////////////////////////////// // experimental candidate omission stuff ////////////////////////////////////// static LISP FT_voice_reget_units(LISP l_duv, LISP l_utt) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_duv)) ){ EST_Utterance *u = get_c_utt(l_utt); duv->regetUnitSequence( u ); } else EST_error( "du_voice_reget_units: expects DiphoneUnitVoice" ); return l_utt; } static LISP FT_utt_tag_unit( LISP l_utt, LISP l_unitnum ) { EST_Utterance *u = get_c_utt(l_utt); const int n = get_c_int( l_unitnum ); if( n<1 ) EST_error( "unit number must be greater than 1" ); EST_Item *it = u->relation("Unit")->first(); int i; for( i=1; i<=n && it!= 0; i++ ) it=it->next(); if( i<=n ) EST_error( "unit number greater than number of items in unit relation") ; ItemList* omitlist=0; if( !it->f_present("omitlist") ){ omitlist = new ItemList; CHECK_PTR(omitlist); it->set_val( "omitlist", est_val(omitlist) ); } else omitlist = itemlist( it->f("omitlist")); EST_Item *currentCandidateUsed = item(it->f("source_ph1")); fprintf(stderr, "setting omit flag on unit %d (item %p)\n", i-1, currentCandidateUsed ); omitlist->append( currentCandidateUsed ); return l_utt; } /////////////////////////////////////////////////////////////////////////////// static LISP FT_voice_get_name(LISP l_voice) { VoiceBase *v = voice( l_voice ); EST_String n = v->name(); return strintern(n); } static LISP FT_voice_set_name(LISP l_voice, LISP l_name) { EST_String n = get_c_string(l_name); VoiceBase *v = voice( l_voice ); v->set_name( n ); return NIL; } static LISP FT_voice_init( LISP l_voice, LISP l_ignore_bad_tag ) { VoiceBase *v = voice( l_voice ); bool ignore_bad_tag = false; if( l_ignore_bad_tag != NIL ) ignore_bad_tag = true; v->initialise( ignore_bad_tag ); return NIL; } static LISP FT_voice_debugLevel( LISP l_voice, LISP l_level ) { VoiceBase *v = voice( l_voice ); if( l_level != NIL ) v->setVerbosity( get_c_int(l_level) ); return flocons( v->verbosity() ); } static void parseVoiceDataParams( LISP l_dataparams, EST_String *uttDir, EST_String *wavDir, EST_String *pmDir, EST_String *coefDir, EST_String *uttExt, EST_String *wavExt, EST_String *pmExt, EST_String *coefExt ) { int listlen = siod_llength( l_dataparams ); if( listlen == 8 ){ *uttExt = get_c_string( CAR1(CDR4(l_dataparams)) ); *wavExt = get_c_string( CAR2(CDR4(l_dataparams)) ); *pmExt = get_c_string( CAR3(CDR4(l_dataparams)) ); *coefExt = get_c_string( CAR4(CDR4(l_dataparams)) ); } else if( listlen == 4 ){ //set some defaults *uttExt = ".utt"; *wavExt = ".wav"; *pmExt = ".pm"; *coefExt = ".coef"; } else EST_error( "Incorrect number of voice data parameters" ); *uttDir = get_c_string( CAR1(l_dataparams) ); *wavDir = get_c_string( CAR2(l_dataparams) ); *pmDir = get_c_string( CAR3(l_dataparams) ); *coefDir = get_c_string( CAR4(l_dataparams) ); } static LISP FT_make_du_voice( LISP l_bnames, LISP l_datadirs, LISP l_srate ) { EST_String uttDir, wavDir, pmDir, coefDir; EST_String uttExt, wavExt, pmExt, coefExt; int wav_srate = get_c_int( l_srate ); if( wav_srate <= 0 ) EST_error( "Waveform sample rate set to %d", wav_srate ); parseVoiceDataParams( l_datadirs, &uttDir, &wavDir, &pmDir, &coefDir, &uttExt, &wavExt, &pmExt, &coefExt ); EST_StrList bnames; siod_list_to_strlist( l_bnames, bnames ); DiphoneUnitVoice *v; v = new DiphoneUnitVoice( bnames, uttDir, wavDir, pmDir, coefDir, static_cast(wav_srate), uttExt, wavExt, pmExt, coefExt ); CHECK_PTR(v); return siod( static_cast(v) ); } static LISP FT_make_du_voice_module( LISP l_bnames, LISP l_datadirs, LISP l_srate ) { EST_String uttDir, wavDir, pmDir, coefDir; EST_String uttExt, wavExt, pmExt, coefExt; int wav_srate = get_c_int( l_srate ); if( wav_srate <= 0 ) EST_error( "Waveform sample rate set to %d", wav_srate ); parseVoiceDataParams( l_datadirs, &uttDir, &wavDir, &pmDir, &coefDir, &uttExt, &wavExt, &pmExt, &coefExt ); EST_StrList bnames; siod_list_to_strlist( l_bnames, bnames ); DiphoneVoiceModule *vm; vm = new DiphoneVoiceModule( bnames, uttDir, wavDir, pmDir, coefDir, static_cast(wav_srate), uttExt, wavExt, pmExt, coefExt ); CHECK_PTR(vm); return siod( vm ); } static LISP FT_voice_add_module( LISP l_duv, LISP l_bnames, LISP l_datadirs, LISP l_srate ) { EST_String uttDir, wavDir, pmDir, coefDir; EST_String uttExt, wavExt, pmExt, coefExt; int wav_srate = get_c_int( l_srate ); if( wav_srate <= 0 ) EST_error( "Waveform sample rate set to %d", wav_srate ); parseVoiceDataParams( l_datadirs, &uttDir, &wavDir, &pmDir, &coefDir, &uttExt, &wavExt, &pmExt, &coefExt ); EST_StrList bnames; siod_list_to_strlist( l_bnames, bnames ); if( DiphoneUnitVoice* duv = dynamic_cast(voice(l_duv)) ){ if( ! duv->addVoiceModule(bnames, uttDir, wavDir, pmDir, coefDir, static_cast(wav_srate), uttExt, wavExt, pmExt, coefExt ) ) EST_error( "voice.addModule failed" ); } else EST_error( "voice_add_module: expects DiphoneUnitVoice for now" ); return NIL; } static LISP FT_du_voice_function( LISP x ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(x)) ){ // do things you can only do to a DiphoneUnitVoice (void) duv; } else EST_error( "du_voice_function: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_precomputeJoinCosts( LISP l_voice, LISP l_phones ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ EST_StrList phones; siod_list_to_strlist( l_phones, phones ); duv->precomputeJoinCosts( phones ); } else EST_error( "du_voice_function: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_set_pruning_beam( LISP l_voice, LISP l_width ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ duv->set_pruning_beam( get_c_float( l_width ) ); } else EST_error( "du_voice_set_pruning: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_set_ob_pruning_beam( LISP l_voice, LISP l_width ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ duv->set_ob_pruning_beam( get_c_float( l_width ) ); } else EST_error( "du_voice_set_pruning: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_set_tc_rescoring_beam( LISP l_voice, LISP l_width ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ duv->set_tc_rescoring_beam( get_c_float( l_width ) ); } else EST_error( "du_voice_set_tc_scoring_beam: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_set_tc_rescoring_weight( LISP l_voice, LISP l_weight ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ duv->set_tc_rescoring_weight( get_c_float( l_weight ) ); } else EST_error( "du_voice_set_tc_rescoring_weight: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_set_target_cost_weight( LISP l_voice, LISP l_weight ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ duv->set_target_cost_weight( get_c_float( l_weight ) ); } else EST_error( "du_voice_set_target_cost_weight: expects DiphoneUnitVoice" ); return NIL; } // static LISP FT_du_voice_set_join_cost_weight( LISP l_voice, LISP l_weight ) // { // if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ // duv->set_join_cost_weight( get_c_float( l_weight ) ); // } // else // EST_error( "du_voice_set_target_cost_weight: expects DiphoneUnitVoice" ); // return NIL; // } static LISP FT_du_voice_set_prosodic_modification( LISP l_voice, LISP l_mod ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ duv->set_prosodic_modification( get_c_int( l_mod ) ); } else EST_error( "du_voice_set_prosodic_modification: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_prosodic_modification(LISP l_voice) { int pm; if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ) { pm = duv->get_prosodic_modification(); if (pm == 0) return NIL; else return truth; } else { EST_error( "du_voice_prosodic_modification: expects DiphoneUnitVoice" ); return NIL; } } static LISP FT_du_voice_set_diphonebackoff(LISP l_voice, LISP l_list) { DiphoneBackoff *dbo; if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ) { dbo = new DiphoneBackoff(l_list); CHECK_PTR(dbo); duv->set_diphone_backoff(dbo); } else EST_error( "du_voice_set_diphone_backoff: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_setTargetCost(LISP l_voice, LISP l_tc) { EST_TargetCost *tc=0; if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ) { if( l_tc == NIL ){ tc = new EST_TargetCost(); CHECK_PTR(tc); } else if(l_tc == truth){ tc = new EST_DefaultTargetCost(); CHECK_PTR(tc); } else if(TYPE(l_tc) == tc_closure) { tc = new EST_SchemeTargetCost(l_tc); CHECK_PTR(tc); } else if(streq(get_c_string(l_tc),"flat")){ tc = new EST_FlatTargetCost(); CHECK_PTR(tc); } else if(streq(get_c_string(l_tc),"apml")){ tc = new EST_APMLTargetCost(); CHECK_PTR(tc); } else if(streq(get_c_string(l_tc),"singing")){ tc = new EST_SingingTargetCost(); CHECK_PTR(tc); } else EST_error( "du_voice_setTargetcost: Unknown targetcost type." ); duv->setTargetCost(tc,true); } else EST_error( "du_voice_setTargetcost: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_setJoinCost(LISP l_voice, LISP l_tc) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ) { EST_JoinCost *jc=0; if( l_tc == truth ){ jc = new EST_JoinCost(); CHECK_PTR(jc); } else EST_error( "du_voice_setJoinCost: currently t is the only supported second arguement" ); duv->setJoinCost(jc,true); } else EST_error( "du_voice_setJoinCost: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_voicemodule_getUtterance( LISP l_vm, LISP l_n ) { EST_Utterance *utt = 0; if( DiphoneVoiceModule *dvm = dynamic_cast(voice(l_vm)) ){ dvm->getUtterance( &utt, get_c_int(l_n) ); } else EST_error( "du_voicemodule_function: expects DiphoneVoiceModule" ); EST_warning( "EST_Utterance = %x\n", utt ); return siod( utt ); } static LISP FT_voice_getUtteranceByFileID( LISP l_vm, LISP l_fileid ) { EST_Utterance *utt = 0; if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_vm)) ){ duv->getCopyUnitUtterance( get_c_string(l_fileid), &utt ); } else EST_error( "du_voicemodule_function: expects DiphoneVoiceModule" ); EST_warning( "EST_Utterance = %x\n", utt ); return siod( utt ); } static LISP FT_voice_unit_count( LISP l_voice ) { VoiceBase *v = voice( l_voice ); return flocons( static_cast(v->numDatabaseUnits()) ); } static LISP FT_voice_unit_type_count( LISP l_voice ) { VoiceBase *v = voice( l_voice ); return flocons( static_cast(v->numUnitTypes()) ); } static LISP FT_voice_unit_available( LISP l_voice, LISP l_unit ) { VoiceBase *v = voice( l_voice ); if( v->unitAvailable( get_c_string( l_unit ) ) ) return truth; return NIL; } static LISP FT_voice_num_available_candidates( LISP l_voice, LISP l_unit ) { VoiceBase *v = voice( l_voice ); unsigned int number = v->numAvailableCandidates( get_c_string(l_unit) ); return flocons( number ); } static LISP FT_du_voice_diphone_coverage( LISP l_voice, LISP l_filename) { DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)); EST_String filename = get_c_string(l_filename); duv->diphoneCoverage(filename); return NIL; } static LISP FT_du_voice_set_jc_f0_weight( LISP l_voice, LISP l_val ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ duv->set_jc_f0_weight( get_c_float( l_val ) ); if (duv->get_jc()) duv->get_jc()->set_f0_weight( get_c_float( l_val ) ); } else EST_error( "du_voice_set_jc_f0_weight: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_get_jc_f0_weight( LISP l_voice ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ return flocons(duv->get_jc_f0_weight()); } else EST_error( "du_voice_get_jc_f0_weight: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_set_jc_power_weight( LISP l_voice, LISP l_val ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ duv->set_jc_power_weight( get_c_float( l_val ) ); if (duv->get_jc()) duv->get_jc()->set_power_weight( get_c_float( l_val ) ); } else EST_error( "du_voice_set_jc_power_weight: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_get_jc_power_weight( LISP l_voice ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ return flocons(duv->get_jc_power_weight()); } else EST_error( "du_voice_get_jc_power_weight: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_set_jc_spectral_weight( LISP l_voice, LISP l_val ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ duv->set_jc_spectral_weight( get_c_float( l_val ) ); if (duv->get_jc()) duv->get_jc()->set_spectral_weight( get_c_float( l_val ) ); } else EST_error( "du_voice_set_jc_spectral_weight: expects DiphoneUnitVoice" ); return NIL; } static LISP FT_du_voice_get_jc_spectral_weight( LISP l_voice ) { if( DiphoneUnitVoice *duv = dynamic_cast(voice(l_voice)) ){ return flocons(duv->get_jc_spectral_weight()); } else EST_error( "du_voice_get_jc_spectral_weight: expects DiphoneUnitVoice" ); return NIL; } void festival_MultiSyn_init(void) { proclaim_module("MultiSyn"); init_subr_2("voice.getUnits", FT_voice_get_units, "(voice.getUnits VOICE UTT)\n\ Voice object VOICE looks at the segment relation in utterance UTT\n\ and adds a suitable unit sequence in the Unit relation."); init_subr_2("utt.tag_unit", FT_utt_tag_unit, "(utt.tag_unit UTT INT)\n\ Tags the candidate used in Unit INT in the Unit relation for omission in\n\ subsequent reruns of viterbi search for the unit sequence."); init_subr_2("du_voice.regetUnits", FT_voice_reget_units, "(du_voice.regetUnits DU_VOICE UTT)\n\ Voice object DU_VOICE looks at the unit relation in utterance UTT\n\ redoes the viterbi, respecting candidates flagged for omission"); init_subr_1("voice.getName", FT_voice_get_name, "(voice.getName VOICE)\n\ Gets the name of a voice."); init_subr_2("voice.setName", FT_voice_set_name, "(voice.setName VOICE NAME)\n\ Sets the name of a voice."); init_subr_2("voice.debugLevel", FT_voice_debugLevel, "(voice.debugLevel VOICE LEVEL)\n\ Query and/or set the level of debugging for VOICE to LEVEL (positive int).\n\ A level of 0 switches off all debugging messages in the voice. Leaving\n\ level unspecified simply returns the current level."); init_subr_3( "make_du_voice", FT_make_du_voice, "(make_du_voice BASENAMES DATADIRS SAMPLERATE)\n\ Creates a Diphone UnitSelection Voice, using the list of file basenames\n\ in LISP list BASENAMES, and the four directory strings in the DATADIRS list.\n\ The voice waveform data files are sampled at SAMPLERATE." ); init_subr_3( "make_du_voice_module", FT_make_du_voice_module, "(make_du_voice_module BASENAMES DATADIRS SAMPLERATE)\n\ Creates a Diphone UnitSelection Voice Module, using the list of file basenames\n\ in LISP list BASENAMES, and the four directory strings in the DATADIRS list.\n\ The voice waveform data files are sampled at SAMPLERATE." ); init_subr_4( "voice.addModule", FT_voice_add_module, "(voice.addModule VOICE BASENAMES DATADIRS SAMPLERATE)\n\ Creates a Diphone UnitSelection Voice Module, using the list of file basenames\n\ in LISP list BASENAMES, and the three directory strings in the remaining\n\ argument DATADIRS and adds it to the current voice. The voice waveform data\n\ files are sampled at SAMPLERATE." ); init_subr_2("voice.init", FT_voice_init, "(voice.init VOICE IGNORE_BAD)\n\ Perform any necessary initialisation for the UnitSelection Voice object VOICE.\n\ If optional IGNORE_BAD is not nil, then phones marked with a \"bad\" feature\n\ in the segment relation will not be added to the diphone inventory" ); init_subr_2("voice.getUtteranceByFileID", FT_voice_getUtteranceByFileID, "(voice.getUtteranceByFileID VOICE FILEIDSTRING)\n\ Returns copy of the Utterance in the voice module database, with\n\ all the Unit relation filled in, ready for synthesis."); init_subr_2("voicemodule.getUtterance", FT_voicemodule_getUtterance, "(voicemodule.getUtterance VOICEMODULE UTTNUMBER)\n\ Returns copy of UTTNUMBER Utterance in the voice module database." ); init_subr_1("voice.numUnitTypes", FT_voice_unit_type_count, "(voice.numUnitTypes VOICE)\n\ Number of different unit types available in Voice object VOICE."); init_subr_1("voice.numUnits", FT_voice_unit_count, "(voice.numUnits VOICE)\n\ Total units available in Voice object VOICE."); init_subr_2("voice.unitAvailable", FT_voice_unit_available, "(voice.unitAvailable VOICE UNIT)\n\ Returns true or false whether speech fragment UNIT (string) is\n\ present in the VOICE"); init_subr_2("voice.numAvailableCandidates", FT_voice_num_available_candidates, "(voice.numAvailableCandidates VOICE UNIT)\n\ Returns the number of instances of speech fragment UNIT (string)\n\ present in the VOICE"); init_subr_1("du_voice_function", FT_du_voice_function, "(du_voice_function DU_VOICE)\n\ Does something to a DU_VOICE only"); init_subr_2("du_voice.precomputeJoinCosts", FT_du_voice_precomputeJoinCosts, "(du_voice.precomputeJoinCosts DU_VOICE PHONELIST)\n\ Calculate and store the join costs for all instances of phones present\n\ in the phone list."); init_subr_2("du_voice.set_pruning_beam", FT_du_voice_set_pruning_beam, "(du_voice.set_pruning_beam DU_VOICE BEAMFLOAT)\n\ Sets the beam pruning parameter for Viterbi search"); init_subr_2("du_voice.set_ob_pruning_beam", FT_du_voice_set_ob_pruning_beam, "(du_voice.set_ob_pruning_beam DU_VOICE BEAMFLOAT)\n\ Sets the observation beam pruning parameter for Viterbi search"); init_subr_2("du_voice.set_tc_rescoring_beam", FT_du_voice_set_tc_rescoring_beam, "(du_voice.set_tc_rescoring_beam DU_VOICE BEAMFLOAT)\n\ Sets the target cost rescoring beam width for Viterbi search (set to -1.0 to disable)"); init_subr_2("du_voice.set_tc_rescoring_weight", FT_du_voice_set_tc_rescoring_weight, "(du_voice.set_tc_rescoring_weight DU_VOICE WEIGHTFLOAT)\n\ Sets the target cost rescoring weight for Viterbi search (set to 0.0 to disable)"); init_subr_2("du_voice.set_target_cost_weight", FT_du_voice_set_target_cost_weight, "(du_voice.set_target_cost_weight DU_VOICE FLOAT)\n\ Sets the target cost weight (default is 1)"); /* * This is currently not implemented, due to problems of passing * such a parameter to the viterbi extend path function. * * init_subr_2("du_voice.set_join_cost_weight", FT_du_voice_set_join_cost_weight, * "(du_voice.set_join_cost_weight DU_VOICE FLOAT)\n \ * Sets the join cost weight (default is 1)"); */ init_subr_2("du_voice.set_jc_f0_weight", FT_du_voice_set_jc_f0_weight, "(du_voice.set_jc_f0_weight DU_VOICE FLOAT)\n\ Sets the joincost f0 weight (default 1)"); init_subr_1("du_voice.get_jc_f0_weight", FT_du_voice_get_jc_f0_weight, "(du_voice.get_jc_f0_weight DU_VOICE)\n\ Gets the joincost f0 weight"); init_subr_2("du_voice.set_jc_power_weight", FT_du_voice_set_jc_power_weight, "(du_voice.set_jc_power_weight DU_VOICE FLOAT)\n\ Sets the joincost power weight (default 1)"); init_subr_1("du_voice.get_jc_power_weight", FT_du_voice_get_jc_power_weight, "(du_voice.get_jc_f0_weight DU_VOICE)\n\ Gets the joincost f0 weight"); init_subr_2("du_voice.set_jc_spectral_weight", FT_du_voice_set_jc_spectral_weight, "(du_voice.set_jc_spectral_weight DU_VOICE FLOAT)\n\ Sets the joincost spectral weight (default 1)"); init_subr_1("du_voice.get_jc_spectral_weight", FT_du_voice_get_jc_spectral_weight, "(du_voice.get_jc_f0_weight DU_VOICE)\n\ Gets the joincost f0 weight"); init_subr_2("du_voice.set_prosodic_modification", FT_du_voice_set_prosodic_modification, "(du_voice.set_prosodic_modification DU_VOICE INT)\n\ Turns prosodic modification on or off (default is 0 [off])\n\ This will only work if durations and f0 targets are provided"); init_subr_1("du_voice.prosodic_modification", FT_du_voice_prosodic_modification, "(du_voice.prosodic_modification DU_VOICE)\n\ Status of prosodic modification on or off."); init_subr_2("du_voice.setDiphoneBackoff", FT_du_voice_set_diphonebackoff, "(du_voice.setDiphoneBackoff DU_VOICE LIST)\n\ Adds diphone backoff rules to the voice."); init_subr_2("du_voice.setJoinCost", FT_du_voice_setJoinCost, "(du_voice.setJoinCost DU_VOICE JOINCOST)\n\ Sets the voice joincost function.\n\ If t is specified then the default joincost is used."); init_subr_2("du_voice.setTargetCost", FT_du_voice_setTargetCost, "(du_voice.setTargetCost DU_VOICE TARGETCOST)\n\ Sets the voice targetcost function.\n\ If t is specified then the default targetcost is used.\n\ If nil is specified then a null targetcost is used.\n\ If a closure is specified, this is called as the target cost.\n\ If 'apml is specified and apml targetcost is uses."); init_subr_2("du_voice.getDiphoneCoverage", FT_du_voice_diphone_coverage, "(du_voice.getDiphoneCoverage DU_VOICE FILENAME)\n\ prints diphone coverage information for this voice\n\ use filename '-' for stdout."); } festival/src/modules/MultiSyn/VoiceBase.cc0000644000175000017500000000636310055371305017200 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: Aug 2002 */ /* --------------------------------------------------------------------- */ /* Abstract base class for "voices" - a top level interface to any code */ /* which knows how to take a preprocessed utterance object and fill in */ /* the information for subsequent synthesis by later festival modules */ /*************************************************************************/ #include "VoiceBase.h" #include "siod_est.h" SIOD_REGISTER_CLASS(voice,VoiceBase) VAL_REGISTER_CLASS(voice,VoiceBase) festival/src/modules/MultiSyn/DiphoneUnitVoice.cc0000644000175000017500000006626512442333465020572 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: Aug 2002 */ /* --------------------------------------------------------------------- */ /* first stab at a diphone unit selection "voice" - using a list of */ /* utterance objects */ /*************************************************************************/ #include "festival.h" #include "DiphoneUnitVoice.h" #include "DiphoneVoiceModule.h" #include "EST_DiphoneCoverage.h" #include "EST_rw_status.h" #include "EST_viterbi.h" #include "EST_Track.h" #include "EST_track_aux.h" #include "EST_Wave.h" #include "EST_THash.h" #include "EST_TList.h" #include "EST_types.h" #include "ling_class/EST_Utterance.h" #include "siod.h" #include "siod_est.h" #include "safety.h" #include #include "EST_TargetCost.h" #include "TargetCostRescoring.h" #include "EST_JoinCost.h" #include "EST_JoinCostCache.h" #include "EST_Val.h" SIOD_REGISTER_TYPE(itemlist,ItemList) VAL_REGISTER_TYPE(itemlist,ItemList) // from src/modules/UniSyn_diphone/us_diphone.h // this won't be staying here long... void parse_diphone_times(EST_Relation &diphone_stream, EST_Relation &source_lab); SIOD_REGISTER_CLASS(du_voice,DiphoneUnitVoice) VAL_REGISTER_CLASS(du_voice,DiphoneUnitVoice) static void my_parse_diphone_times(EST_Relation &diphone_stream, EST_Relation &source_lab) { EST_Item *s, *u; float dur1, dur_u, p_time=0.0; // NOTE: because of the extendLeft/extendRight phone join hack for missing diphones, // the unit linked list *may be* shorter that the segment list. //(admittedly could cause confusion) for( s=source_lab.head(), u=diphone_stream.head(); (u!=0)&&(s!=0); u=u->next(), s=s->next()){ EST_Track *pm = track(u->f("coefs")); int end_frame = pm->num_frames() - 1; int mid_frame = u->I("middle_frame"); dur1 = pm->t(mid_frame); dur_u = pm->t(end_frame); s->set("end", (p_time+dur1) ); p_time += dur_u; u->set("end", p_time); if( u->f_present("extendRight") ){//because diphone squeezed out (see above) s = s->next(); s->set("end", p_time ); } } if(s) s->set("end", (p_time)); } // temporary hack necessary because decoder can only take a // function pointer (would be better to relax this restriction in // the EST_Viterbi_Decoder class, or in a replacement class, rather // than using this hack) static DiphoneUnitVoice *globalTempVoicePtr = 0; DiphoneUnitVoice::DiphoneUnitVoice( const EST_StrList& basenames, const EST_String& uttDir, const EST_String& wavDir, const EST_String& pmDir, const EST_String& coefDir, unsigned int sr, const EST_String& uttExt, const EST_String& wavExt, const EST_String& pmExt, const EST_String& coefExt ) : pruning_beam( -1 ), ob_pruning_beam( -1 ), tc_rescoring_beam( -1 ), tc_rescoring_weight( 0.0 ), tc_weight( 1.0 ), jc_weight( 1.0 ), jc_f0_weight( 1.0 ), jc_power_weight( 1.0 ), jc_spectral_weight( 1.0 ), prosodic_modification( 0 ), wav_srate( sr ), jc( 0 ), jc_delete( false ), tc( 0 ), tc_delete( false ), tcdh( 0 ) { // make the default voice module with the supplied parameters addVoiceModule( basenames, uttDir, wavDir, pmDir, coefDir, wav_srate, uttExt, wavExt, pmExt, coefExt ); diphone_backoff_rules = 0; } void DiphoneUnitVoice::initialise( bool ignore_bad_tag ) { if( jc == 0 ) EST_error( "Need to set join cost calculator for voice" ); if( tc == 0 ) EST_error( "Need to set target cost calculator for voice" ); EST_TList::Entries it; for( it.begin(voiceModules); it; it++ ) (*it)->initialise( tc, ignore_bad_tag ); } bool DiphoneUnitVoice::addVoiceModule( const EST_StrList& basenames, const EST_String& uttDir, const EST_String& wavDir, const EST_String& pmDir, const EST_String& coefDir, unsigned int srate, const EST_String& uttExt, const EST_String& wavExt, const EST_String& pmExt, const EST_String& coefExt ) { DiphoneVoiceModule *vm; if( srate != wav_srate ) EST_error( "Voice samplerate: %d\nmodule samplerate: %d", wav_srate, srate ); vm = new DiphoneVoiceModule( basenames, uttDir, wavDir, pmDir, coefDir, srate, uttExt, wavExt, pmExt, coefExt ); CHECK_PTR(vm); registerVoiceModule( vm ); return true; } void DiphoneUnitVoice::registerVoiceModule( DiphoneVoiceModule *vm ) { voiceModules.append( vm ); } void DiphoneUnitVoice::setJoinCost( EST_JoinCost *jcost, bool del ) { if( jc_delete == true ) if( jc != 0 ) delete jc; jc = jcost; jc_delete = del; } void DiphoneUnitVoice::setTargetCost( EST_TargetCost *tcost, bool del ) { if( tc_delete == true ) if( tc != 0 ) delete tc; tc = tcost; tc_delete = del; } DiphoneUnitVoice::~DiphoneUnitVoice() { EST_TList::Entries it; for( it.begin(voiceModules); it; it++ ) delete( *it ); if(diphone_backoff_rules) delete diphone_backoff_rules; if( jc_delete == true ) if( jc != 0 ) delete jc; if( tc_delete == true ) if( tc != 0 ) delete tc; if(tcdh) delete tcdh; } void DiphoneUnitVoice::addToCatalogue( const EST_Utterance *utt ) { // needed? } void DiphoneUnitVoice::getDiphone( const EST_VTCandidate *cand, EST_Track* coef, EST_Wave* sig, int *midframe, bool extendLeft, bool extendRight ) { // The need for this function in this class is a bit messy, it would be far // nicer just to be able to ask the Candidate itself to hand over the relevant // synthesis parameters. In future, it will work that way ;) // put there by DiphoneVoiceModule::getCandidateList const DiphoneCandidate *diphcand = diphonecandidate( cand->name ); const DiphoneVoiceModule* parentModule = diphcand->dvm; EST_Item *firstPhoneInDiphone = cand->s; // need to call right getDiphone to do the actual work parentModule->getDiphone( firstPhoneInDiphone, coef, sig, midframe, extendLeft, extendRight ); } // REQUIREMENT: the unit relation must have previously been used to initialise the // Viterbi decoder from which the path was produced. void DiphoneUnitVoice::fillUnitRelation( EST_Relation *units, const EST_VTPath *path ) { EST_Item *it=units->tail(); for ( ; path != 0 && it != 0; path=path->from, it=it->prev() ){ EST_Track *coefs = new EST_Track; CHECK_PTR(coefs); EST_Wave *sig = new EST_Wave; CHECK_PTR(sig); int midf; getDiphone( path->c, coefs, sig, &midf, it->f_present("extendLeft"), it->f_present("extendRight")); EST_Item *firstPhoneInDiphone = path->c->s; it->set_val( "sig", est_val( sig ) ); it->set_val( "coefs", est_val( coefs ) ); it->set( "middle_frame", midf ); it->set( "source_utt", firstPhoneInDiphone->relation()->utt()->f.S("fileid")); it->set_val( "source_ph1", est_val( firstPhoneInDiphone )); it->set( "source_end", firstPhoneInDiphone->F("end")); it->set( "target_cost", path->c->score ); //have to recalculate join cost as it's not currently saved anywhere if( path->from == 0 ) it->set( "join_cost", 0.0); else{ // join cost between right edge of left diphone and vice versa const DiphoneCandidate *l_diph = diphonecandidate(path->from->c->name); const DiphoneCandidate *r_diph = diphonecandidate(path->c->name); it->set( "join_cost", (*jc)( l_diph, r_diph ) ); } } } // The use of the globalFunctionPtr in this function is a really just a temporary hack // necessary because the decoder as it stands at present can only take a function pointer // (would be better to relax this restriction in the EST_Viterbi_Decoder class, or in a // replacement class, rather than using this hack) // static EST_VTPath* extendPath( EST_VTPath *p, EST_VTCandidate *c, // EST_Features&) // { // EST_VTPath *np = new EST_VTPath; // CHECK_PTR(np); // if( globalTempVoicePtr ==0 ) // EST_error( "globalTempVoicePtr is not set, can't continue" ); // const EST_JoinCost &jcost = globalTempVoicePtr->getJoinCostCalculator(); // np->c = c; // np->from = p; // np->state = c->pos; // if ((p == 0) || (p->c == 0)) // np->score = c->score; // else{ // // join cost between right edge of left diphone and vice versa // np->score = p->score + c->score + jcost( p->c->s->next(), c->s ); // } // return np; // } static EST_VTPath* extendPath( EST_VTPath *p, EST_VTCandidate *c, EST_Features&) { EST_VTPath *np = new EST_VTPath; CHECK_PTR(np); if( globalTempVoicePtr ==0 ) EST_error( "globalTempVoicePtr is not set, can't continue" ); const EST_JoinCost &jcost = globalTempVoicePtr->getJoinCostCalculator(); np->c = c; np->from = p; np->state = c->pos; if ((p == 0) || (p->c == 0)) np->score = c->score; else{ const DiphoneCandidate *l_diph = diphonecandidate(p->c->name); const DiphoneCandidate *r_diph = diphonecandidate(c->name); // join cost between right edge of left diphone and vice versa np->score = p->score + c->score + jcost( l_diph, r_diph ); } return np; } // This function is a really just a temporary hack necessary because the decoder // as it stands at present can only take a function pointer (would be better to relax // this restriction in the EST_Viterbi_Decoder class, or in a replacement class, rather // than using this hack) static EST_VTCandidate* getCandidatesFunction( EST_Item *s, EST_Features &f) { DiphoneUnitVoice *duv = globalTempVoicePtr; if( duv==0 ) EST_error( "Candidate source voice is unset" ); return duv->getCandidates( s, f ); } // Function which, given an item from the timeline relation that // was originally used to initialise the EST_Viterbi_Decoder // returns a pointer to a linked list of EST_VTCandidates // (this is provided to the viterbi decoder upon its construction // and (in)directly called by it as part of the decoding process...) EST_VTCandidate* DiphoneUnitVoice::getCandidates( EST_Item *s, EST_Features &f) const { EST_VTCandidate *c = 0; EST_VTCandidate *moduleListHead = 0; EST_VTCandidate *moduleListTail = 0; // these objects [c/sh]ould be a parameter visible in the user's script // land, and will be in future... // tc now a member // EST_DefaultTargetCost default_target_cost; // EST_TargetCost *tc = &default_target_cost; // or // EST_SchemeTargetCost scheme_target_cost(rintern( "targetcost")); // EST_TargetCost *tc = &scheme_target_cost; EST_TList::Entries module_iter; int nfound, total=0; //////////////////////////////////////////////////////////////// // join linked list of candidates from each module into one list for( module_iter.begin(voiceModules); module_iter; module_iter++ ){ nfound = (*module_iter)->getCandidateList( *s, tc, tcdh, tc_weight, &moduleListHead, &moduleListTail ); if( nfound>0 ){ moduleListTail->next = c; c = moduleListHead; total += nfound; } } if( total==0 ) EST_error( "Couldn't find diphone %s", (const char*)s->S("name") ); if( verbosity() > 0 ) printf( "Number of candidates found for target \"%s\": %d\n", (const char*)s->S("name"), total ); if( ! ((tc_rescoring_beam == -1.0) || (tc_rescoring_weight <= 0.0)) ) rescoreCandidates( c, tc_rescoring_beam, tc_rescoring_weight ); return c; } void DiphoneUnitVoice::diphoneCoverage(const EST_String filename) const { EST_DiphoneCoverage dc; EST_TList::Entries module_iter; // for each module for( module_iter.begin(voiceModules); module_iter; module_iter++ ) (*module_iter)->getDiphoneCoverageStats(&dc); dc.print_stats(filename); } bool DiphoneUnitVoice::synthesiseWave( EST_Utterance *utt ) { getUnitSequence( utt ); return true; } void DiphoneUnitVoice::getUnitSequence( EST_Utterance *utt ) { EST_Relation *segs = utt->relation( "Segment" ); EST_Relation *units = utt->create_relation( "Unit" ); if(!tcdh) tcdh = new TCDataHash(20); else tcdh->clear(); // Initialise the Unit relation time index for decoder EST_String diphone_name; EST_StrList missing_diphones; EST_Item *it=segs->head(); if( it == 0 ) EST_error( "Segment relation is empty" ); bool extendLeftFlag = false; for( ; it->next(); it=it->next() ) { EST_String l = it->S("name"); EST_String r = it->next()->S("name"); EST_String diphone_name = EST_String::cat(l,"_",r); EST_String orig = diphone_name; if(tc->is_flatpack()) tcdh->add_item( it , ((EST_FlatTargetCost *)tc)->flatpack(it) ); // First attempt back off: // If missing diphone is an interword diphone, insert a silence! // Perceptual results say this is prefered. if ( diphone_name != EST_String::Empty && !this->unitAvailable(diphone_name) ) { EST_Item *s1,*s2; EST_Item *w1=0,*w2=0; cerr << "Missing diphone: "<< diphone_name << endl; if((s1 = parent(it,"SylStructure"))) w1= parent(s1,"SylStructure"); if( (s2 = parent(it->next(),"SylStructure"))) w2= parent(s2,"SylStructure"); if( w1 && w2 && (w1 != w2) ) { EST_Item *sil; cerr << " Interword so inseting silence.\n"; sil = it->insert_after(); sil->set("name",ph_silence()); r = it->next()->S("name"); diphone_name = EST_String::cat(l,"_",r); } } // Simple back off. // Change diphone name for one we actually have. while(diphone_name != EST_String::Empty && !this->unitAvailable(diphone_name) && diphone_backoff_rules) { cerr << " diphone still missing, backing off: " << diphone_name << endl; diphone_name = diphone_backoff_rules->backoff(l,r); l = diphone_name.before("_"); r = diphone_name.after("_"); cerr << " backed off: " << orig << " -> " << diphone_name << endl; if( verbosity() > 0 ){ EST_warning("Backing off requested diphone %s to %s", orig.str(), diphone_name.str() ); } } //// Complex backoff. Changes the segment stream to the right, //// may still leave a discontinuity to the left. This could be //// fixed, but it would requires a better search. Rob's thoughts //// are that the simple method works better, unless it resorts to //// a bad default rule. // while(!this->unitAvailable(diphone_name) && // diphone_backoff_rules && // !diphone_backoff_rules->backoff(it)) // diphone_name = EST_String::cat(it->S("name"),"_",it->next()->S("name")); if( !this->unitAvailable( diphone_name ) ){ missing_diphones.append( diphone_name ); if(units->tail()) units->tail()->set( "extendRight", 1 ); extendLeftFlag = true; // trigger for next unit to make up second half of missing diphone } else{ EST_Item *t = units->append(); t->set( "name", diphone_name ); if(orig != diphone_name) t->set( "missing_diphone",orig); t->set_val( "ph1", est_val(it) ); if( extendLeftFlag == true ){ t->set( "extendLeft", 1 ); extendLeftFlag = false; } } } // stop if necessary units are still missing. if( missing_diphones.length() > 0 ){ for( EST_Litem *it=missing_diphones.head(); it!=0 ; it=it->next() ) printf( "requested diphone missing: %s\n", missing_diphones(it).str() ); EST_warning("Making phone joins to compensate..."); // EST_error("Unable to synthesise utterance due to missing diphones"); } // Make the decoder do its thing // -1 means number of states at each time point not fixed EST_Viterbi_Decoder v( getCandidatesFunction, extendPath, -1 ); // turn on pruning if necessary if( (pruning_beam>0) || (ob_pruning_beam>0) ) v.set_pruning_parameters( pruning_beam, ob_pruning_beam ); // temporary hack necessary because decoder can only take a // function pointer (would be better to relax this restriction in // the EST_Viterbi_Decoder class, or in a replacement class, rather // than using this hack) globalTempVoicePtr = this; v.set_big_is_good(false); if( verbosity() > 0 ) v.turn_on_trace(); v.initialise( units ); v.search(); // take hold of the best path (end thereof) EST_VTPath *bestp=0; if( !v.result( &bestp ) ) EST_error( "No best candidate sequence found" ); // fill in the best path features in the Unit Relation fillUnitRelation( units, bestp ); my_parse_diphone_times( *units, *segs ); } ///////////////////////////////////////////////////////////////////////////////////// // Canned example experimental code (proof of concept rather than intelligently done) static inline bool itemListContainsItem( const ItemList* il, const EST_Item *item ) { ItemList::Entries it; for( it.begin( *il ); it; it++ ) if( (*it) == item ) return true; return false; } static EST_VTCandidate* getCandidatesWithOmissionsFunction( EST_Item *s, EST_Features &f ) { DiphoneUnitVoice *duv = globalTempVoicePtr; if( duv==0 ) EST_error( "Candidate source voice is unset" ); //get candidate list as usual EST_VTCandidate *candlist = duv->getCandidates( s, f ); //filter out candidates on basis of omission list (yes, this is quite dumb) if( s->f_present( "omitlist" ) ){ EST_warning( "omitlist found in unit %s", s->S("name").str() ); ItemList *omitlist = itemlist( s->f("omitlist") ); //until one candidate remains as head (to keep hold of list head) while( candlist != 0 && itemListContainsItem( omitlist, candlist->s ) ){ EST_VTCandidate *del_cand = candlist; candlist = candlist->next; del_cand->next = 0; //so deletion doesn't trigger total list deletion delete del_cand; } //then continue down list EST_VTCandidate *prev = candlist; EST_VTCandidate *cand = candlist->next; while( cand!=0 ){ if( itemListContainsItem( omitlist, cand->s ) ){ //delete cand on true prev->next = cand->next; cand->next = 0; //so deletion doesn't trigger total list deletion delete cand; cand = prev; } cand = cand->next; } if( candlist == 0 ) EST_error( "zero candidates remain after filtering" ); } return candlist; } // For when the utterance already has the unit sequence, with certain candidates // flagged as to be avoided, or mandatory and so on... void DiphoneUnitVoice::regetUnitSequence( EST_Utterance *utt ) { // Unit relation should already be in existence for decoder EST_Relation *units = utt->relation( "Unit" ); EST_Item *it=units->head(); if( it == 0 ) EST_error( "Unit relation is empty" ); // Make the decoder do its thing (again) // -1 means number of states at each time point not fixed EST_Viterbi_Decoder v( getCandidatesWithOmissionsFunction, extendPath, -1 ); // turn on pruning if necessary if( (pruning_beam>0) || (ob_pruning_beam>0) ) v.set_pruning_parameters( pruning_beam, ob_pruning_beam ); // temporary hack necessary because decoder can only take a // function pointer (would be better to relax this restriction in // the EST_Viterbi_Decoder class, or in a replacement class, rather // than using this hack) globalTempVoicePtr = this; v.set_big_is_good(false); if( verbosity() > 0 ) v.turn_on_trace(); v.initialise( units ); v.search(); // take hold of the best path (end thereof) EST_VTPath *bestp=0; if( !v.result( &bestp ) ) EST_error( "No best candidate sequence found" ); // fill in the best path features in the Unit Relation fillUnitRelation( units, bestp ); EST_Relation *segs = utt->relation("Segment"); my_parse_diphone_times( *units, *segs ); } // End canned example experimental code /////////////////////////////////////////// /////////////////////////////////////////////////////////////////////////////////// bool DiphoneUnitVoice::unitAvailable( const EST_String &diphone ) const { EST_TList::Entries it; for( it.begin(voiceModules); it; it++ ) if( (*it)->numAvailableCandidates(diphone) > 0 ) return true; return false; } unsigned int DiphoneUnitVoice::numAvailableCandidates( const EST_String &diphone ) const { unsigned int number = 0; EST_TList::Entries it; for( it.begin(voiceModules); it; it++ ) number += (*it)->numAvailableCandidates(diphone); return number; } //////////////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////////////// // special case of the above for utterances structures that are // actually in the voice database, which doesn't do any search // This is useful for doing copy synthesis of utterances (eg. // to test out resynthesis, prosodic modification and so on) void DiphoneUnitVoice::getCopyUnitUtterance( const EST_String &utt_fname, EST_Utterance **utt_out ) const { // need to find which, if any, voice module has this utterance // in its list EST_TList::Entries module_iter; EST_Utterance *db_utt=0; for( module_iter.begin(voiceModules); module_iter; module_iter++ ) if( (*module_iter)->getUtterance(&db_utt, "fileid", utt_fname) == true ) break; if( db_utt == 0 ) EST_error( "Could not find Utterance %s in any voice module", utt_fname.str() ); else{ // deep copy database utterance and fill in Unit relation *utt_out = new EST_Utterance( *db_utt ); CHECK_PTR(utt_out); EST_Utterance myUtt( *db_utt ); cerr << myUtt.relation_present( "Segment" ) << " " << myUtt.num_relations() <relation_present( "Segment" ) << " " << (*utt_out)->relation_present( "Segment" ) << " " << (*utt_out)->num_relations() <relation( "Segment" ); EST_Relation *units = (*utt_out)->create_relation( "Unit" ); // Initialise the Unit relation + fill in necessary/suitable // synthesis parameters EST_String ph1, ph2; EST_Item *it = segs->tail(); EST_Item *db_utt_seg_it = db_utt->relation( "Segment" )->tail(); if( it == 0 ) EST_error( "Segment relation is empty" ); else{ ph2 = it->S("name"); while( ((it=it->prev())!=0) && ((db_utt_seg_it=db_utt_seg_it->prev())!=0) ){ EST_Track *coefs = new EST_Track; CHECK_PTR(coefs); EST_Wave *sig = new EST_Wave; CHECK_PTR(sig); int midf; (*module_iter)->getDiphone( db_utt_seg_it, coefs, sig, &midf ); ph1 = it->S("name"); EST_Item *t = units->prepend(); t->set( "name", EST_String::cat(ph1,"_",ph2) ); t->set_val( "ph1", est_val(it) ); t->set_val( "sig", est_val( sig ) ); t->set_val( "coefs", est_val( coefs ) ); t->set( "middle_frame", midf ); t->set( "source_utt", db_utt->f.S("fileid")); t->set_val( "source_ph1", est_val( db_utt_seg_it )); t->set( "source_end", db_utt_seg_it->F("end")); t->set( "target_cost", 0.0 ); t->set( "join_cost", 0.0); ph2 = ph1; } } my_parse_diphone_times( *units, *segs ); // this is for copy synthesis, so copy actual timings //for( EST_Item *seg = segs->head(); it!=0; it=it->next() ) //seg->set( "end", seg->F("source_end") ); } } //////////////////////////////////////////////////////////////////////// //////////////////////////////////////////////////////////////////////// unsigned int DiphoneUnitVoice::numUnitTypes() const { //necessary? return 0; } unsigned int DiphoneUnitVoice::numDatabaseUnits() const { unsigned int sum=0; EST_TList::Entries it; for( it.begin( voiceModules ); it; it++ ) sum += (*it)->numModuleUnits(); return sum; } ////////////////////////////////////////////////////////////////////////// void DiphoneUnitVoice::set_diphone_backoff(DiphoneBackoff *dbo) { if (diphone_backoff_rules) delete diphone_backoff_rules; diphone_backoff_rules = dbo; } int DiphoneUnitVoice::getPhoneList( const EST_String &phone, ItemList &list ) { unsigned int n=0; EST_TList::Entries it; for( it.begin( voiceModules ); it; it++ ) n += (*it)->getPhoneList( phone, list ); return n; } void DiphoneUnitVoice::precomputeJoinCosts( const EST_StrList &phones, bool verbose ) { EST_StrList::Entries it; for( it.begin( phones ); it; it++ ){ ItemList *l = new ItemList; CHECK_PTR(l); unsigned int n = getPhoneList( (*it), *l ); if( verbose==true ) cerr << "phone " << (*it) << " " << n << " instances\n"; if( n>0 ){ jc->computeAndCache( *l, true ); //verbose=true } else EST_warning( "Phone %s not listed in voice", (*it).str() ); delete l; } } festival/src/modules/MultiSyn/VoiceModuleBase.cc0000644000175000017500000000630410055371305020341 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: Aug 2002 */ /* --------------------------------------------------------------------- */ /* Abstract base class for "voice modules" - a top level interface to */ /* any collection of speech material in the voice database */ /*************************************************************************/ #include "VoiceModuleBase.h" #include "siod_est.h" SIOD_REGISTER_CLASS(voicemodule,VoiceModuleBase) VAL_REGISTER_CLASS(voicemodule,VoiceModuleBase) festival/src/modules/MultiSyn/DiphoneVoiceModule.cc0000644000175000017500000005100311465010356021053 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: Aug 2002 */ /* --------------------------------------------------------------------- */ /* A diphone unit selection "voice module" */ /* (implemented using a list of utterance objects) */ /*************************************************************************/ #include "DiphoneVoiceModule.h" #include "EST_TargetCost.h" #include "EST_viterbi.h" #include "EST_rw_status.h" #include "EST_Track.h" #include "EST_track_aux.h" #include "EST_Wave.h" #include "EST_THash.h" #include "EST_TList.h" #include "EST_types.h" #include "ling_class/EST_Utterance.h" #include "siod.h" #include "siod_est.h" #include "safety.h" #include #include "EST_Val.h" // from src/modules/UniSyn_diphone/us_diphone.h // this won't be staying here long... void parse_diphone_times(EST_Relation &diphone_stream, EST_Relation &source_lab); SIOD_REGISTER_CLASS(du_voicemodule,DiphoneVoiceModule) VAL_REGISTER_CLASS(du_voicemodule,DiphoneVoiceModule) VAL_REGISTER_CLASS(diphonecandidate,DiphoneCandidate) // defined in a single place to avoid inconsistency. // Given a phone segment item, return the standard cut point // time, calculated in the standard way. float getJoinTime( const EST_Item *seg ) { float midt=0.0; // hack to avoid overhead of string creation and deletion // (EST feature access should really be changed to take // const char* instead of const EST_String& ) static const EST_String cl_end_str( "cl_end" ); static const EST_String dipth_str( "dipth" ); static const EST_String start_str( "start" ); // work out boundary for diphone join if( seg->f_present(cl_end_str) ) // join at cl_end point for stops midt = seg->features().val("cl_end").Float(); else if( seg->f_present(dipth_str) ) // join at 25% through a diphthong midt = 0.75*seg->F(start_str) + 0.25*seg->features().val("end").Float(); else midt = ( seg->F(start_str) + seg->features().val("end").Float() ) / 2.0; return midt; } DiphoneVoiceModule::DiphoneVoiceModule( const EST_StrList& basenames, const EST_String& uttDir, const EST_String& wavDir, const EST_String& pmDir, const EST_String& coefDir, unsigned int sr, const EST_String& uttExt, const EST_String& wavExt, const EST_String& pmExt, const EST_String& coefExt ) : fileList( basenames ), utt_dir ( uttDir ), utt_ext ( uttExt ), pm_dir( pmDir ), pm_ext( pmExt ), coef_dir( coefDir ), coef_ext( coefExt ), wave_dir( wavDir ), wave_ext( wavExt ), wav_srate( sr ), tcdatahash ( 0 ), utt_dbase( 0 ), catalogue( 0 ) { } void DiphoneVoiceModule::addCoefficients( EST_Relation *segs, const EST_Track& coefs ) { float startt, midt, endt; EST_FVector *startf, *midf, *endf; const int num_coefs = coefs.num_channels(); // hack to avoid overhead of string creation and deletion // (EST feature access should really be changed to take // const char* instead of const EST_String& ) static const EST_String startcoef_str("startcoef"); static const EST_String midcoef_str("midcoef"); static const EST_String endcoef_str("endcoef"); static const EST_String start_str("start"); EST_Item *seg=segs->head(); startt = seg->F(start_str); startf = new EST_FVector(num_coefs); CHECK_PTR(startf); coefs.copy_frame_out(coefs.index(startt), *startf); //this one not shared for( ; seg!=0; seg=seg->next() ){ // work out boundary for diphone join midt = getJoinTime( seg ); // copy frames out and set as features seg->features().set_val( startcoef_str, est_val(startf) ); midf = new EST_FVector(num_coefs); CHECK_PTR(midf); coefs.copy_frame_out(coefs.index(midt), *midf); seg->features().set_val( midcoef_str, est_val(midf) ); endt = seg->features().val("end").Float(); endf = new EST_FVector(num_coefs); CHECK_PTR(endf); coefs.copy_frame_out(coefs.index(endt), *endf); seg->features().set_val( endcoef_str, est_val(endf) ); startf = endf; // phones share frame at phone boundary (reference counted in EST_Val) } } void DiphoneVoiceModule::flatPack( EST_Relation *segs, const EST_TargetCost *tc ) const { const EST_FlatTargetCost *ftc = (EST_FlatTargetCost *)tc; for( EST_Item *seg=segs->head(); seg->next() !=0; seg=seg->next() ) tcdatahash->add_item(seg, ftc->flatpack(seg)); } void DiphoneVoiceModule::initialise( const EST_TargetCost *tc, bool ignore_bad_tag ) { EST_Utterance *u=0; EST_Relation *segs=0; tcdatahash = new TCDataHash(500); utt_dbase = new EST_TList; CHECK_PTR(utt_dbase); catalogue = new EST_TStringHash( 2500 ); CHECK_PTR(catalogue); int numIgnoredPhones=0; if(ignore_bad_tag) EST_warning( "Looking for bad flags"); else EST_warning( "Ignoring bad flags"); for( EST_Litem *it=fileList.head(); it!=0 ; it=it->next() ){ u = new EST_Utterance; CHECK_PTR(u); if( (u->load(utt_dir+fileList(it)+utt_ext)) != read_ok ) EST_error( "Couldn't load utterance %s\n", (const char*)fileList(it) ); segs = u->relation( "Segment" ); // add join cost coefficients (at middle of phones) EST_Track coefs; if( (coefs.load((coef_dir+fileList(it)+coef_ext))) != read_ok ) EST_error( "Couldn't load data file %s", (const char*) (coef_dir+fileList(it)+coef_ext) ); addCoefficients( segs, coefs ); if (tc->is_flatpack()) { flatPack(segs,tc); u->remove_relation("Token"); u->remove_relation("Word"); u->remove_relation("Phrase"); u->remove_relation("Syllable"); u->remove_relation("SylStructure"); u->remove_relation("IntEvent"); u->remove_relation("Intonation"); } addToCatalogue( u, &numIgnoredPhones, ignore_bad_tag ); utt_dbase->append( u ); } if(ignore_bad_tag) EST_warning( "Ignored %d phones with bad flag set\n", numIgnoredPhones ); } DiphoneVoiceModule::~DiphoneVoiceModule() { if( utt_dbase != 0 ){ EST_Litem *it = utt_dbase->head(); for( ; it!=0 ; it=it->next() ) delete (*utt_dbase)(it); delete utt_dbase; } delete catalogue; if(tcdatahash) delete tcdatahash; } void DiphoneVoiceModule::addToCatalogue( const EST_Utterance *utt, int *num_ignored, bool ignore_bad ) { EST_Item *item, *next_item; ItemList *diphoneList; const EST_String *ph1, *ph2; int found=0; static const EST_String bad_str( "bad" ); item = (utt->relation( "Segment" ))->tail(); if( item!=0 ){ ph2 = &(item->features().val("name").String()); while( (item=item->prev()) != 0 ){ next_item = item->next(); // You'd think we need to check both item->f_present(bad_str) and // next_item->f_present(bad_str) like this: //if((item->f_present(bad_str) || next_item->f_present(bad_str)) && ignore_bad == true){ // But experiment showed that then each time one diphone too many would be // ignored. This was partly compensated by a bug pesent up to r1.14 // (a iteration within "if(item=item->prev()!=0)" just before the "continue") // which caused the leftmost bad phone in a row of bad phones NOT to be ignored // when the length of the row was even (or when it was odd and ended in the // utterance-final phone, which is never checked for badness). if(item->f_present(bad_str) && ignore_bad == true){ (*num_ignored)++; EST_warning( "Ignoring diphone \"%s_%s\" (LEFT %s in %s at %fs, bad flag \"%s\")", item->S("name").str(), next_item->S("name").str(), item->S("name").str(), utt->f.S("fileid").str(), item->F("end"), item->S("bad").str() ); if(item->prev() != 0){ continue; } else break; //already at start of list, so finish up } ph1 = &(item->features().val("name").String()); // EST_warning( "Adding phone \"%s\" (%s, %f) to diphoneList %s_%s", // item->S("name").str(), // utt->f.S("fileid").str(), // item->F("end"), // item->S("name").str(), // next_item->S("name").str()); diphoneList = catalogue->val(EST_String::cat(*ph1,"_",*ph2), found); if( !found ){ diphoneList = new ItemList; CHECK_PTR(diphoneList); catalogue->add_item(EST_String::cat(*ph1,"_",*ph2), diphoneList, 1); // no_search=1 } diphoneList->append( item ); ph2 = ph1; } } } void DiphoneVoiceModule::getDiphone( const EST_Item *phone1, EST_Track* coef, EST_Wave* sig, int *midframe, bool extendLeft, bool extendRight ) const { EST_Item *phone2 = phone1->next(); // load the relevant parts const EST_String &fname = phone1->relation()->utt()->f.val("fileid").String(); static const EST_String start_str( "start" ); float startt,midt,endt; if( extendLeft==true ) startt = phone1->F(start_str); else startt = getJoinTime( phone1 ); midt = phone1->features().val("end").Float(); if( extendRight==true ) endt = phone2->features().val("end").Float(); else endt = getJoinTime( phone2 ); // get pitchmarks for pitch synchronous synthesis EST_Track *tempcoef = new EST_Track; CHECK_PTR(tempcoef); if( (tempcoef->load((pm_dir+fname+pm_ext))) != read_ok ) EST_error( "Couldn't load data file %s", (const char*) (pm_dir+fname+pm_ext) ); // following few lines effectively moves segment boundaries to // line up with pitch periods. int copy_start = tempcoef->index( startt ); int copy_end = tempcoef->index( endt ); //copy_end -= 1; //so that adjacent units don't start and end with same frame int copy_len = copy_end - copy_start; //int copy_len = copy_end - copy_start + 1; startt = tempcoef->t( copy_start ); endt = tempcoef->t( copy_end ); if( copy_len == 0 ){ EST_warning( "%s(%f->%f): %s_%s diphone length means 1 pitchmark will be duplicated", fname.str(), startt, endt, phone1->S("name").str(), phone2->S("name").str() ); copy_len=1; } else if( copy_len < 0 ){ EST_error( "%s(%f->%f): %s_%s diphone length renders %d pitchmark", fname.str(), startt, endt, phone1->S("name").str(), phone2->S("name").str(), copy_len ); } tempcoef->copy_sub_track( *coef, copy_start, copy_len ); *midframe = coef->index( midt ); // adjust timing, which Festival synthesis code makes assumptions about // SPECIFICALLY, the unisyn module wants all units to start from // the first value above 0.0 (as the first pitch mark) float t_off = (copy_start!=0) ? tempcoef->t(copy_start-1) : 0.0; int nframes = coef->num_frames(); for( int i=0; it(i) -= t_off; //start waveform at previous pitchmark (this is period approximation used) int st_sample = (int)rint( t_off * (float) wav_srate ); //preferably end waveform at following pitchmark (follows convention in UniSyn module) int end_sample; if( copy_end < tempcoef->num_frames() ) end_sample = (int) rint( tempcoef->t(copy_end) * (float) wav_srate ); //if( copy_end+1 < tempcoef->num_frames() ) // end_sample = (int) rint( tempcoef->t(copy_end+1) * (float) wav_srate ); else{ // estimate from previous pitch mark shift int pp_centre_sample = (int) rint( endt * (float) wav_srate ); int pp_first_sample = (int) rint( tempcoef->t(copy_end) * (float) wav_srate ); //int pp_first_sample = (int) rint( tempcoef->t(copy_end-1) * (float) wav_srate ); end_sample = (2*pp_centre_sample)-pp_first_sample; } // (obviously, we would want to load and cache any files // // which haven't been loaded yet, rather than just load // // the parts each and every time) // if( sig->load( wave_dir+fname+wave_ext, // st_sample, end_sample-st_sample+1) != read_ok ) // EST_error( "Couldn't load data file %s", // (const char*) (wave_dir+fname+wave_ext) ); // delete tempcoef; } inline EST_VTCandidate* makeCandidate( const EST_Item *target_ph1, const EST_Item *cand_ph1, const EST_TargetCost *tc, const TCData *tcd, const TCDataHash *tcdatahash, float tc_weight, const DiphoneVoiceModule *dvm_p ) { // hack to avoid overhead of string creation and deletion // (EST feature access should really be changed to take // const char* instead of const EST_String& ) static const EST_String extendLeft_str("extendLeft"); static const EST_String extendRight_str("extendRight"); static const EST_String jccid_str("jccid"); EST_VTCandidate *c = new EST_VTCandidate; CHECK_PTR(c); EST_Item *cand_ph2 = cand_ph1->next(); // set up all the members we can here c->s = const_cast(cand_ph1); EST_FVector *left, *right; if( target_ph1->f_present( extendLeft_str ) ) left = fvector( cand_ph1->features().val( "startcoef" ) ); else left = fvector( cand_ph1->features().val( "midcoef" ) ); if( target_ph1->next()->f_present( extendRight_str ) ) right = fvector( cand_ph2->features().val( "endcoef" ) ); else right = fvector( cand_ph2->features().val( "midcoef" ) ); // an abuse of the "name" EST_Val member to store data we want instead // of what is intended to go there // (will become unnecessary with a more general candidate class) DiphoneCandidate *cand = new DiphoneCandidate( cand_ph1, dvm_p, left, right ); CHECK_PTR(cand); c->name = est_val( cand ); //to get synthesis parameters (deleted by EST_Val c->name) if( cand_ph1->f_present( jccid_str ) ){ cand->ph1_jccid = cand_ph1->features().val( "jccid" ).Int(); cand->ph1_jccindex = cand_ph1->features().val( "jccindex" ).Int(); } if( cand_ph2->f_present( jccid_str ) ){ cand->ph1_jccid = cand_ph2->features().val( "jccid" ).Int(); cand->ph1_jccindex = cand_ph2->features().val( "jccindex" ).Int(); } if(tc->is_flatpack()) c->score = tc_weight* ((const EST_FlatTargetCost *)tc) ->operator()( tcd, tcdatahash->val( const_cast(cand_ph1) ) ); else c->score = tc_weight*tc->operator()( target_ph1, cand_ph1 ); return c; } inline void itemListToCandidateList( ItemList::Entries &it, EST_VTCandidate **head, EST_VTCandidate **tail, const EST_Item *target_ph1, const EST_TargetCost *tc, const TCDataHash *tcdh, const TCDataHash *tcdatahash, float tc_weight, int count, const DiphoneVoiceModule *dvm_p ) { int i=0; if( count > 0 ){ TCData *tcd = tcdh->val( const_cast(target_ph1) ); EST_VTCandidate *nextc = 0; // make last one first EST_VTCandidate *c = makeCandidate( target_ph1, (*it), tc, tcd, tcdatahash, tc_weight, dvm_p ); c->next = nextc; *tail = c; // then iterate back prepending to linked list // (order reversed because using c->next) nextc = c; it++; i++; for( ; (it && inext = nextc; nextc = c; } *head = c; // keep hold of last one set up } return; } int DiphoneVoiceModule::getCandidateList( const EST_Item& target, const EST_TargetCost* tc, const TCDataHash *tcdh, float tc_weight, EST_VTCandidate **head, EST_VTCandidate **tail ) const { int nfound = 0; const EST_Item *target_ph1 = item(target.f("ph1")); int found = 0; const ItemList *candidateItemList = catalogue->val( target.S("name"), found ); if( found != 0 ){ nfound = candidateItemList->length(); ItemList::Entries it = ItemList::Entries(*candidateItemList); itemListToCandidateList( it, head, tail, target_ph1, tc, tcdh, tcdatahash, tc_weight, nfound, this ); } return nfound; } int DiphoneVoiceModule::getPhoneList( const EST_String &phone, ItemList &list ) { unsigned int n=0; if( utt_dbase != 0 ){ for( EST_Litem *it=utt_dbase->head(); it!=0 ; it=it->next() ){ EST_Item *ph=(*utt_dbase)(it)->relation("Segment")->head(); for( ; ph!=0; ph=ph->next() ){ if( ph->S("name") == phone ){ list.append( ph ); n++; } } } } return n; } bool DiphoneVoiceModule::getUtterance( EST_Utterance** utt, int n ) const { if( n<0 || n>(utt_dbase->length()-1) ) EST_error( "Utterance index out of bounds" ); if( utt == 0 ) EST_error( "Invalid utterance" ); // deep copy the utterance in question *utt = new EST_Utterance( *(utt_dbase->nth(n)) ); CHECK_PTR(utt); return true; } bool DiphoneVoiceModule::getUtterance( EST_Utterance **utt, const EST_String &feat_name, const EST_Val &value ) const { //search down list of utterance structures, comparing // fileid feature. If find a match, return pointer to that // utterance. for( EST_Litem *it=utt_dbase->head(); it!=0 ; it=it->next() ) if( (*utt_dbase)(it)->f.val(feat_name) == value ){ *utt = (*utt_dbase)(it); return true; } return false; } void DiphoneVoiceModule::getDiphoneCoverageStats(EST_DiphoneCoverage *dc) const { for( EST_Litem *it=utt_dbase->head(); it!=0 ; it=it->next() ) dc->add_stats((*utt_dbase)(it)); } unsigned int DiphoneVoiceModule::numUnitTypes() const { return catalogue ? catalogue->num_entries() : 0; } unsigned int DiphoneVoiceModule::numModuleUnits() const { unsigned int sum=0; if( catalogue != 0 ){ EST_TStringHash::Entries it; for( it.begin( *catalogue ); it; it++ ) sum += it->v->length(); //EST_UList.length() counts the entries :( } return sum; } unsigned int DiphoneVoiceModule::numAvailableCandidates( const EST_String &unit ) const { int number=0; int found=0; const ItemList *candidateItemList = catalogue->val( unit, found ); if( found > 0 ) number = candidateItemList->length(); return number; } festival/src/modules/MultiSyn/EST_TargetCost.cc0000644000175000017500000005505211465010356020133 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: October 2002 */ /* --------------------------------------------------------------------- */ /* */ /* */ /* */ /* */ /* */ /* */ /*************************************************************************/ #include #include "festival.h" #include "ling_class/EST_Item.h" #include "EST_TargetCost.h" #include "siod.h" static const EST_String simple_pos(const EST_String &s); static const EST_Utterance *tc_get_utt(const EST_Item *seg); static const EST_Item* tc_get_syl(const EST_Item *seg); static const EST_Item* tc_get_word(const EST_Item *seg); static EST_String ff_tobi_accent(const EST_Item *s); static EST_String ff_tobi_endtone(const EST_Item *s); static bool threshold_equal(float a, float b, float threshold); /* * BASE CLASS: EST_TargetCost */ /* Individual cost functions */ // This is really designed only for apml! float EST_TargetCost::apml_accent_cost() const { // Check if target is an apml utterance. If not return 0 as we don't // trust its accent specification. if( !tc_get_utt(targ)->relation_present("SemStructure")) return 0.0; // Check if candidate is an apml utterance. If not return 1 // (as we want to use apml if available) if( !tc_get_utt(cand)->relation_present("SemStructure")) return 1.0; // As they are both apml match accents. const EST_Item *tsyl, *csyl; EST_String targ_accent, cand_accent, targ_boundary, cand_boundary; if( ph_is_vowel(targ->features().val("name").String()) && !ph_is_silence(targ->features().val("name").String()) ) { tsyl = tc_get_syl(targ); csyl = tc_get_syl(cand); // Can't assume candidate and target identities are the same // (because of backoff to a silence for example) if( csyl == 0 ) return 1.0; targ_accent = ff_tobi_accent(tsyl); cand_accent = ff_tobi_accent(csyl); targ_boundary = ff_tobi_endtone(tsyl); cand_boundary = ff_tobi_endtone(csyl); if( (cand_accent != targ_accent) || (cand_boundary != targ_boundary) ) return 1.0; } if( ph_is_vowel(targ->next()->features().val("name").String()) && !ph_is_silence(targ->next()->features().val("name").String()) ) { tsyl = tc_get_syl(targ->next()); csyl = tc_get_syl(cand->next()); // Can't assume candidate and target identities are the same // (because of backoff to a silence for example) if( csyl == 0 ) return 1.0; targ_accent = ff_tobi_accent(tsyl); cand_accent = ff_tobi_accent(csyl); targ_boundary = ff_tobi_endtone(tsyl); cand_boundary = ff_tobi_endtone(csyl); if( (cand_accent != targ_accent) || (cand_boundary != targ_boundary) ) return 1.0; } return 0.0; } float EST_TargetCost::stress_cost() const { int cand_stress; int targ_stress; const EST_Item *tsyl, *csyl; if( ph_is_vowel(targ->features().val("name").String()) && !ph_is_silence(targ->features().val("name").String()) ) { tsyl = tc_get_syl(targ); csyl = tc_get_syl(cand); // Can't assume candidate and target identities are the same // (because of backoff to a silence for example) if( csyl == 0 ) { //cout << "SC: 1 returning 1\n"; return 1; } targ_stress = (tsyl->I("stress") > 0) ? 1 : 0; cand_stress = (csyl->I("stress") > 0) ? 1 : 0; if( cand_stress != targ_stress) { //cout << "SC: 2 returning 1\n"; return 1; } } if( ph_is_vowel(targ->next()->features().val("name").String()) && !ph_is_silence(targ->next()->features().val("name").String()) ) { tsyl = tc_get_syl(targ->next()); csyl = tc_get_syl(cand->next()); // Can't assume candidate and target identities are the same // (because of backoff to a silence for example) if( csyl == 0 ) { //cout << "SC: 3 returning 1\n"; return 1; } targ_stress = (tsyl->I("stress") > 0) ? 1 : 0; cand_stress = (csyl->I("stress") > 0) ? 1 : 0; if( cand_stress != targ_stress) { //cout << "SC: 4 returning 1\n"; return 1; } } //cout << "SC: 5 returning 0\n"; return 0; } float EST_TargetCost::position_in_syllable_cost() const { tcpos_t targ_pos = TCPOS_MEDIAL; tcpos_t cand_pos = TCPOS_MEDIAL; const EST_Item *targ_syl = tc_get_syl(targ); const EST_Item *targ_next_syl = tc_get_syl(targ->next()); const EST_Item *targ_next_next_syl = tc_get_syl(targ->next()->next()); const EST_Item *targ_prev_syl = tc_get_syl(targ->prev()); const EST_Item *cand_syl = tc_get_syl(cand); const EST_Item *cand_next_syl = tc_get_syl(cand->next()); const EST_Item *cand_next_next_syl = tc_get_syl(cand->next()->next()); const EST_Item *cand_prev_syl = tc_get_syl(cand->prev()); if( targ_syl != targ_next_syl ) targ_pos = TCPOS_INTER; else if( targ_syl != targ_prev_syl) targ_pos = TCPOS_INITIAL; else if( targ_next_syl != targ_next_next_syl) targ_pos = TCPOS_FINAL; if( cand_syl != cand_next_syl ) cand_pos = TCPOS_INTER; else if( cand_syl != cand_prev_syl) cand_pos = TCPOS_INITIAL; else if( cand_next_syl != cand_next_next_syl) cand_pos = TCPOS_FINAL; return (targ_pos == cand_pos) ? 0 : 1; } float EST_TargetCost::position_in_word_cost() const { tcpos_t targ_pos = TCPOS_MEDIAL; tcpos_t cand_pos = TCPOS_MEDIAL; const EST_Item *targ_word = tc_get_word(targ); const EST_Item *targ_next_word = tc_get_word(targ->next()); const EST_Item *targ_next_next_word = tc_get_word(targ->next()->next()); const EST_Item *targ_prev_word = tc_get_word(targ->prev()); const EST_Item *cand_word = tc_get_word(cand); const EST_Item *cand_next_word = tc_get_word(cand->next()); const EST_Item *cand_next_next_word = tc_get_word(cand->next()->next()); const EST_Item *cand_prev_word = tc_get_word(cand->prev()); if( targ_word != targ_next_word ) targ_pos = TCPOS_INTER; else if( targ_word != targ_prev_word) targ_pos = TCPOS_INITIAL; else if( targ_next_word != targ_next_next_word) targ_pos = TCPOS_FINAL; if( cand_word != cand_next_word ) cand_pos = TCPOS_INTER; else if( cand_word != cand_prev_word) cand_pos = TCPOS_INITIAL; else if( cand_next_word != cand_next_next_word) cand_pos = TCPOS_FINAL; return (targ_pos == cand_pos) ? 0 : 1; } float EST_TargetCost::position_in_phrase_cost() const { const EST_Item *targ_word = tc_get_word(targ); const EST_Item *cand_word = tc_get_word(cand); if (!targ_word && !cand_word) return 0; if (!targ_word || !cand_word) return 1; return (targ_word->features().val("pbreak").String() == cand_word->features().val("pbreak").String()) ? 0 : 1; } float EST_TargetCost::punctuation_cost() const { const EST_Item *targ_word = tc_get_word(targ); const EST_Item *cand_word = tc_get_word(cand); const EST_Item *next_targ_word = tc_get_word(targ->next()); const EST_Item *next_cand_word = tc_get_word(cand->next()); float score = 0.0; if ( (targ_word && !cand_word) || (!targ_word && cand_word) ) score += 0.5; else if (targ_word && cand_word) if ( parent(targ_word,"Token")->features().val("punc","NONE").String() != parent(cand_word,"Token")->features().val("punc","NONE").String() ) score += 0.5; if ( (next_targ_word && !next_cand_word) || (!next_targ_word && next_cand_word) ) score += 0.5; else if(next_targ_word && next_cand_word) if ( parent(next_targ_word,"Token")->features().val("punc","NONE").String() != parent(next_cand_word,"Token")->features().val("punc","NONE").String() ) score += 0.5; return score; } float EST_TargetCost::partofspeech_cost() const { // Compare left phone half of diphone const EST_Item *targ_left_word = tc_get_word(targ); const EST_Item *cand_left_word = tc_get_word(cand); if(!targ_left_word && !cand_left_word) return 0; if(!targ_left_word || !cand_left_word) return 1; const EST_String targ_left_pos( simple_pos(targ_left_word->features().val("pos").String()) ); const EST_String cand_left_pos( simple_pos(cand_left_word->features().val("pos").String()) ); if( targ_left_pos != cand_left_pos ) return 1; // Compare right phone half of diphone const EST_Item *targ_right_word = tc_get_word(targ->next()); const EST_Item *cand_right_word = tc_get_word(cand->next()); if(!targ_right_word && !cand_right_word) return 0; if(!targ_right_word || !cand_right_word) return 1; const EST_String targ_right_pos( simple_pos(targ_right_word->features().val("pos").String()) ); const EST_String cand_right_pos( simple_pos(cand_right_word->features().val("pos").String()) ); if( targ_right_pos != cand_right_pos ) return 1; return 0; } float EST_TargetCost::left_context_cost() const { EST_Item *targ_context = targ->prev(); EST_Item *cand_context = cand->prev(); if ( !targ_context && !cand_context) return 0; if ( !targ_context || !cand_context) return 1; return (targ_context->features().val("name").String() == cand_context->features().val("name").String()) ? 0 : 1; } float EST_TargetCost::right_context_cost() const { EST_Item *targ_context = targ->next()->next(); EST_Item *cand_context = cand->next()->next(); if ( !targ_context && !cand_context) return 0; if ( !targ_context || !cand_context) return 1; return (targ_context->features().val("name").String() == cand_context->features().val("name").String()) ? 0 : 1; } float EST_TargetCost::out_of_lex_cost() const { static const EST_String ool_feat("bad_lex"); // bad_dur may at some stage be set on a target for resynthesis purposes. if( cand->f_present(ool_feat) != targ->f_present(ool_feat) ) return 1.0; if( cand->next()->f_present(ool_feat) != targ->next()->f_present(ool_feat) ) return 1.0; return 0.0; } float EST_TargetCost::bad_duration_cost() const { static const EST_String bad_dur_feat("bad_dur"); // bad_dur may at some stage be set on a target for resynthesis purposes. if( cand->f_present(bad_dur_feat) != targ->f_present(bad_dur_feat) ) return 1.0; if( cand->next()->f_present(bad_dur_feat) != targ->next()->f_present(bad_dur_feat) ) return 1.0; // If the segments next to these segments are bad, then these ones are probably wrong too! if( cand->prev() && targ->prev() && ( cand->prev()->f_present(bad_dur_feat) != targ->prev()->f_present(bad_dur_feat) ) ) return 1.0; if( cand->next()->next() && targ->next()->next() && ( cand->next()->next()->f_present(bad_dur_feat) != targ->next()->next()->f_present(bad_dur_feat) ) ) return 1.0; return 0.0; } float EST_TargetCost::bad_f0_cost() const { // by default, the last element of join cost coef vector is // the f0 (i.e. fv->a_no_check( fv->n()-1 ) ) const EST_Item *cand_left = cand; const EST_Item *cand_right = cand_left->next(); const EST_String &left_phone( cand_left->features().val("name").String() ); const EST_String &right_phone( cand_right->features().val("name").String() ); EST_FVector *fv = 0; float penalty = 0.0; if( ph_is_vowel( left_phone ) || ph_is_approximant( left_phone ) || ph_is_liquid( left_phone ) || ph_is_nasal( left_phone ) ){ fv = fvector( cand_left->f("midcoef") ); if( fv->a_no_check(fv->n()-1) == -1.0 ) // means unvoiced penalty += 0.5; } if( ph_is_vowel( right_phone ) || ph_is_approximant( right_phone ) || ph_is_liquid( right_phone ) || ph_is_nasal( right_phone ) ){ fv = fvector( cand_right->f("midcoef") ); if( fv->a_no_check(fv->n()-1) == -1.0 ) // means unvoiced penalty += 0.5; } return penalty; } /* * DERIVED CLASS: EST_DefaultTargetCost * * This is CSTR's proposed default target cost. Nothing special, if you think you can * do better derive your own class. */ float EST_DefaultTargetCost::operator()(const EST_Item* targ, const EST_Item* cand) const { set_targ_and_cand(targ,cand); score = 0.0; weight_sum = 0.0; score += add_weight(10.0)*stress_cost(); score += add_weight(5.0)*position_in_syllable_cost(); score += add_weight(5.0)*position_in_word_cost(); score += add_weight(6.0)*partofspeech_cost(); score += add_weight(15.0)*position_in_phrase_cost(); score += add_weight(4.0)*left_context_cost(); score += add_weight(3.0)*right_context_cost(); score /= weight_sum; // These are considered really bad, and will result in a score > 1. score += 10.0*bad_duration_cost(); // see also join cost. score += 10.0*bad_f0_cost(); score += 10.0*punctuation_cost(); score += 10.0*out_of_lex_cost(); return score ; } /* * DERIVED CLASS: EST_APMLTargetCost * */ float EST_APMLTargetCost::operator()(const EST_Item* targ, const EST_Item* cand) const { set_targ_and_cand(targ,cand); score = 0.0; weight_sum = 0.0; score += add_weight(10.0)*stress_cost(); score += add_weight(20.0)*apml_accent_cost(); // APML only! score += add_weight(5.0)*position_in_syllable_cost(); score += add_weight(5.0)*position_in_word_cost(); score += add_weight(6.0)*partofspeech_cost(); score += add_weight(4.0)*position_in_phrase_cost(); score += add_weight(4.0)*left_context_cost(); score += add_weight(3.0)*right_context_cost(); score /= weight_sum; score += 10.0*bad_duration_cost(); // see also join cost. score += 10.0*bad_f0_cost(); score += 10.0*punctuation_cost(); score += 10.0*out_of_lex_cost(); return score; } /* * DERIVED CLASS: EST_SingingTargetCost * * Mostly default stuff, but tries to match pitch and duration * specified on Tokens from the xxml * */ float EST_SingingTargetCost::pitch_cost() const { const EST_Item *targ_word = tc_get_word(targ); const EST_Item *cand_word = tc_get_word(cand); const EST_Item *next_targ_word = tc_get_word(targ->next()); const EST_Item *next_cand_word = tc_get_word(cand->next()); const float threshold = 0.1; float targ_pitch, cand_pitch; LISP l_tmp; float score = 0.0; if ( (targ_word && !cand_word) || (!targ_word && cand_word) ) { cout << "PITCH PENALTY WORD NON-WORD MISMATCH\n"; score += 0.5; } else if (targ_word && cand_word) { l_tmp = lisp_val(parent(targ_word,"Token")->f("freq",est_val(0))); // This currently assumes one syllable words, need to process // the list more for multiple syllable words, or move the info // to the syllable. if(CONSP(l_tmp)) targ_pitch = get_c_float(car(l_tmp)); else targ_pitch = get_c_float(l_tmp); cand_pitch = parent(cand_word,"Token")->F("freq",0.0); if ( ! threshold_equal(targ_pitch,cand_pitch,threshold)) { cout << "PP: " << targ_pitch << " " << cand_pitch << endl; score += 0.5; } } if ( (next_targ_word && !next_cand_word) || (!next_targ_word && next_cand_word) ) { cout << "PITCH PENALTY NEXT WORD NON-WORD MISMATCH\n"; score += 0.5; } else if(next_targ_word && next_cand_word) { l_tmp = lisp_val(parent(next_targ_word,"Token")->f("freq",est_val(0))); if(CONSP(l_tmp)) targ_pitch = get_c_float(car(l_tmp)); else targ_pitch = get_c_float(l_tmp); cand_pitch = parent(next_cand_word,"Token")->F("freq",0.0); if ( ! threshold_equal(targ_pitch,cand_pitch,threshold)) { cout << "NP: " << targ_pitch << " " << cand_pitch << endl; score += 0.5; } } if (score == 0.0) cout << "NO PITCH PENALTY\n"; return score; } float EST_SingingTargetCost::duration_cost() const { const EST_Item *targ_word = tc_get_word(targ); const EST_Item *cand_word = tc_get_word(cand); const EST_Item *next_targ_word = tc_get_word(targ->next()); const EST_Item *next_cand_word = tc_get_word(cand->next()); float targ_dur, cand_dur; LISP l_tmp; float score = 0.0; if ( (targ_word && !cand_word) || (!targ_word && cand_word) ) score += 0.5; else if (targ_word && cand_word) { l_tmp = lisp_val(parent(targ_word,"Token")->f("dur",est_val(0))); if(CONSP(l_tmp)) targ_dur = get_c_float(car(l_tmp)); else targ_dur = get_c_float(l_tmp); cand_dur = parent(cand_word,"Token")->F("dur",0.0); if ( targ_dur != cand_dur ) score += 0.5; } if ( (next_targ_word && !next_cand_word) || (!next_targ_word && next_cand_word) ) score += 0.5; else if(next_targ_word && next_cand_word) { l_tmp = lisp_val(parent(next_targ_word,"Token")->f("dur",est_val(0))); if(CONSP(l_tmp)) targ_dur = get_c_float(car(l_tmp)); else targ_dur = get_c_float(l_tmp); cand_dur = parent(next_cand_word,"Token")->F("dur",0.0); if ( targ_dur != cand_dur ) score += 0.5; } return score; } float EST_SingingTargetCost::operator()(const EST_Item* targ, const EST_Item* cand) const { set_targ_and_cand(targ,cand); score = 0.0; weight_sum = 0.0; score += add_weight(50.0)*pitch_cost(); score += add_weight(50.0)*duration_cost(); score += add_weight(5.0)*stress_cost(); score += add_weight(5.0)*position_in_syllable_cost(); score += add_weight(5.0)*position_in_word_cost(); score += add_weight(5.0)*partofspeech_cost(); score += add_weight(5.0)*position_in_phrase_cost(); score += add_weight(5.0)*punctuation_cost(); score += add_weight(4.0)*left_context_cost(); score += add_weight(3.0)*right_context_cost(); score += add_weight(2.0)*bad_duration_cost(); // see also join cost. return score / weight_sum; } /* * DERIVED CLASS: EST_SchemeTargetCost * * This lets you implement your target cost in scheme, so you can * change it on the fly. Great for developement, but about 5 times as slow. * */ float EST_SchemeTargetCost::operator()( const EST_Item* targ, const EST_Item* cand ) const { LISP r,l; l = cons(tc, cons( siod(targ), cons( siod(cand), NIL) )); r = leval(l,NIL); if ((consp(r)) || (r == NIL) || !(numberp(r))) { cerr << "Lisp function: " << tc << " did not return float score" << endl; festival_error(); } else score = get_c_float(r); return score; } /* * Auxillary target cost functions */ static const EST_String simple_pos(const EST_String &s) { if( s == "nn" || s == "nnp" || s == "nns" || s == "nnps" || s == "fw" || s == "sym" || s == "ls") return "n"; if( s == "vbd" || s == "vb" || s == "vbn" || s == "vbz" || s == "vbp" || s == "vbg") return "v"; if( s == "jj" || s == "jjr" || s == "jjs" || s == "1" || s == "2" || s == "rb" || s == "rp" || s == "rbr" || s == "rbs") return "other"; return "func"; } static const EST_Utterance *tc_get_utt(const EST_Item *seg) { return seg->relation()->utt(); } static const EST_Item *tc_get_syl(const EST_Item *seg) { // if(!seg) // return 0; return parent(seg,"SylStructure"); } static const EST_Item *tc_get_word(const EST_Item *seg) { // if(!seg) // return 0; const EST_Item *syl = tc_get_syl(seg); if(syl) return parent(syl,"SylStructure"); else return 0; } /* adapted from base/ff.cc */ static EST_String ff_tobi_accent(const EST_Item *s) { // First tobi accent related to syllable EST_Item *nn = as(s,"Intonation"); EST_Item *p; for (p=daughter1(nn); p; p=p->next()) if (p->name().contains("*")) return p->name(); return "NONE"; } static EST_String ff_tobi_endtone(const EST_Item *s) { // First tobi endtone (phrase accent or boundary tone) EST_Item *nn = as(s,"Intonation"); EST_Item *p; for (p=daughter1(nn); p; p=p->next()) { EST_String l = p->name(); if ((l.contains("%")) || (l.contains("-"))) return p->name(); } return "NONE"; } static bool threshold_equal(float a, float b, float threshold) { if ( ( (a-b) < threshold ) && ( (a-b) > -threshold ) ) return true; else return false; } festival/src/modules/MultiSyn/TargetCostRescoring.cc0000644000175000017500000001273511465010356021275 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author: Korin Richmond */ /* Date: June 2006 */ /* --------------------------------------------------------------------- */ /* */ /* Code to allow the possibility of changing the costs (thus ranking) */ /* assigned to candidates units by the target cost function, typically */ /* based on what those original costs are... */ /* */ /*************************************************************************/ #include "TargetCostRescoring.h" #include "ling_class/EST_Item.h" #include "EST_TList.h" #include "EST_viterbi.h" #include "DiphoneVoiceModule.h" //for getJoinTime(const EST_Item *it) // Do some additional target cost here, based on the acoustic // properties of the units found matching this target unit. (Changes // the candidate list "candidates" in place.) This currently is // limited to looking at the durations of the top scoring candidates, // but also could be extended to include components for the pitch and // amplitude at various points for example... void rescoreCandidates( EST_VTCandidate *candidates, float beam_width, float mult ) { // first calculate the stats for the "model" float dur = 0.0; EST_Item *ph1 = 0; EST_Item *ph2 = 0; //EST_FVector *ph1_mid = 0; //EST_FVector *ph2_mid = 0; EST_TList scores; // get all scores to work out what durations are "suitable" for( EST_VTCandidate *it = candidates; it != 0; it=it->next ){ ph1 = it->s; ph2 = ph1->next(); // ph1_mid = fvector( ph1->f( "midcoef" ) ); // ph2_mid = fvector( ph2->f( "midcoef" ) ); dur = getJoinTime(ph2) - getJoinTime(ph1); // duration of diphone unit scores.append( ScorePair(it->score,dur, it) ); } sort( scores ); //cerr << scores << endl; // calculate simple mean duration of some or all of candidates float meandur = 0.0; int n = 0; if( beam_width < 0 ){ // just average all of them for( EST_Litem *li = scores.head(); li != 0; li = li->next() ){ meandur += scores(li)._dur; n++; } } else{ float score_cutoff = scores.first()._score + beam_width; for( EST_Litem *li = scores.head(); li != 0; li = li->next() ){ if( scores(li)._score > score_cutoff ) break; else{ meandur += scores(li)._dur; n++; } } } meandur /= n; // then tweak the scores based on that for( EST_Litem *li = scores.head(); li != 0; li = li->next() ){ float cand_dur = scores(li)._dur; // cerr << scores(li)._cand->score << " "; scores(li)._cand->score += (mult * abs( cand_dur - meandur ) ); // cerr << scores(li)._cand->score << endl; } } ostream& operator << ( ostream& out, const ScorePair &sp ) { out << sp._score << " " << sp._dur << "\n"; return out; } festival/src/modules/MultiSyn/EST_JoinCost.cc0000644000175000017500000001023710055371305017576 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: October 2002 */ /* --------------------------------------------------------------------- */ /* name speaks for itself */ /* */ /* */ /* */ /* */ /* */ /*************************************************************************/ #include "ling_class/EST_Relation.h" #include "ling_class/EST_Utterance.h" #include "EST_JoinCost.h" #include "EST_JoinCostCache.h" #include "safety.h" #include "ling_class/EST_Item.h" #include "EST_FMatrix.h" EST_JoinCost::~EST_JoinCost() { int cclen = costCaches.length(); for( int i=0; i &list, bool verbose ) { // for later speed use the costCache's index as it's id unsigned int jccid = costCaches.length(); costCaches.resize( jccid+1 ); //EST_TVector doesn't have push_back yet EST_JoinCostCache *jcc = new EST_JoinCostCache( jccid, list.length() ); CHECK_PTR(jcc); costCaches[jccid] = jcc; // this function adds the relevant indicies return jcc->computeAndCache( list, *this, verbose ); } festival/src/modules/MultiSyn/EST_JoinCostCache.cc0000644000175000017500000001272211465010356020525 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2004 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: January 2004 */ /* */ /*************************************************************************/ #include "EST_JoinCostCache.h" #include "EST_JoinCost.h" #include "EST_error.h" #include "safety.h" #include #include EST_JoinCostCache::EST_JoinCostCache( unsigned int id ) : numInstances(0), _id(id), cache(0), cachelen(0) { } EST_JoinCostCache::EST_JoinCostCache( unsigned int id, unsigned int n ) : numInstances(n), _id(id), cachelen((n*n)/2-n), deleteMemoryOnDeath(true) { cache = new unsigned char [cachelen]; CHECK_PTR( cache ); } EST_JoinCostCache::EST_JoinCostCache( unsigned int id, unsigned char *mem, unsigned int n, bool del ) : numInstances(n), _id(id), cache(mem), cachelen((n*n)/2-n), deleteMemoryOnDeath(del) { } EST_JoinCostCache::~EST_JoinCostCache() { if( cache != 0 ) if( deleteMemoryOnDeath ) delete [] cache; } unsigned char EST_JoinCostCache::val( unsigned int a, unsigned int b ) const { if( a>numInstances || b>numInstances ) EST_error( "Requested index greater than cache size" ); if( a == b ) return minVal; else if( b > a ) return cache[(b*(b-1)>>1)+a]; else return cache[(a*(a-1)>>1)+b]; return defVal; } bool EST_JoinCostCache::setval( unsigned int a, unsigned int b, unsigned char v ) { if( a>numInstances || b>numInstances ) EST_error( "Requested index greater than cache size" ); if( a == b ){ return true; } else if( b > a ){ cache[(b*(b-1)>>1)+a] = v; return true; } else{ cache[(a*(a-1)>>1)+b] = v; return true; } return false; } ostream& EST_JoinCostCache::write( ostream &os ) const { os << cachelen; // os.write( cache, cachelen ); return os; } bool EST_JoinCostCache::computeAndCache( const EST_TList &list, const EST_JoinCost &jc, bool verbose ) { unsigned char qcost; // quantized cost unsigned int qleveln = maxVal-minVal; float ulimit = 1.0-1/(float)(qleveln); float llimit = 0.0+1/(float)(qleveln); unsigned int i=0; EST_warning("EST_JoinCostCache::computeAndCache"); for( EST_Litem *it=list.head(); it; it=it->next(), ++i ){ unsigned int j=i+1; for( EST_Litem *jt=it->next(); jt; jt=jt->next(), ++j ){ float cost = jc( list(it), list(jt) ); if( cost >= ulimit ) qcost = maxVal; else if( cost <= llimit ) qcost = minVal; else qcost = static_cast(rint(cost*(float)qleveln)); setval( i, j, qcost ); } // yet to be convinced this is the best place for this... list(it)->set( "jccid", (int)this->id() ); list(it)->set( "jccindex", (int) i ); } return true; } festival/src/modules/MultiSyn/DiphoneBackoff.cc0000644000175000017500000001755711465010356020213 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 2004 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Rob Clark */ /* Date: Jan 2004 */ /* --------------------------------------------------------------------- */ /* Diphone backing off procedure */ /* */ /* Substitute a target phone for another (or multiple others). */ /* */ /* The class holds a list of rules. Each rule is represented by a list */ /* The rule is interpreted as "The head can be replaced by the tail" */ /* */ /* */ /*************************************************************************/ #include "DiphoneBackoff.h" const EST_String DiphoneBackoff::default_match = "_"; DiphoneBackoff::DiphoneBackoff(LISP l_backofflist) { EST_StrList list; LISP l; for ( l = l_backofflist ; l != NIL ; l = cdr(l)) { siod_list_to_strlist(car(l),list); if(list.length() < 2) EST_warning("BackoffList: ignoring invalid entry %s\n" , (const char*)list.first()); else backofflist.append(list); } } // The function EST_TList::append copies data, so this is not necessary. // // DiphoneBackoff::~DiphoneBackoff() // { // EST_Litem *p; // // for (p = backofflist.head(); p != 0; p = p->next()) // delete backofflist(p); // } /* * This version of backoff just takes two phone names as input and * returns a diphone name as output. It is entirely up to the caller * what to do with this information. * Note that it assumes substitutions are 1 phone for 1 phone. */ EST_String DiphoneBackoff::backoff(EST_String left, EST_String right) { EST_Litem *p; EST_String head,sub,result,rl,rr; rl = left; rr = right; p = backofflist.head(); while( p!= 0 ) { int i = 0; head = backofflist(p).nth(i++); sub = backofflist(p).nth(i++); if ( head == left || ( head == default_match && left != sub) ) { rl = sub; p = 0; } else if ( head == right || ( head == default_match && right != sub) ) { rr = sub; p = 0; } else p = p->next(); } if ( left != rl || right != rr ) result = EST_String::cat(rl,"_",rr); else result = EST_String::Empty; return result; } /* * This version of backoff takes a pointer to a target diphone in the * Segment as input. It changes the target phone sequence by * substituting a phone for one OR MORE phone as the rules * dictate. If the left phone of the diphone is altered a diphone * mismatch will occur in the between sucessive candidate lists. If * the right phone is substituted this will not happen, as the * candidate list for the next diphone will be constrcted from the * corrected list. */ int DiphoneBackoff::backoff(EST_Item *p1) { EST_Item *p2, *pp, *pps; EST_String n1,n2,head,sub,full_sub,bo; bool done = false; EST_Litem *p; if(! p1) EST_error("Backoff received null item."); if ( ! (p2 = p1->next()) ) EST_error("Backoff didn't get passed a diphone."); n1=p1->S("name"); n2=p2->S("name"); p = backofflist.head(); // for each rule. while( p!= 0 && !done ) { int i = 0 ; head = backofflist(p).nth(i++); pp = 0; // Match head of rule to left phone, or if head of the rule is the defualt substitution // do it if it hasn't already been done. if( (head == n1) || ( (head == default_match) && !is_defaultbackoff(p1) ) ) pp = p1; // if this fails, try the right phone. else if( (head == n2) || ( (head == default_match) && !is_defaultbackoff(p2) ) ) pp = p2; if(pp) { bo = pp->S("name"); sub = backofflist(p).nth(i++); full_sub = sub; pp->set("name",sub); set_backoff(pp); if(head.matches(default_match)) set_defaultbackoff(pp); while (i < backofflist(p).length()) { sub = backofflist(p).nth(i++); full_sub = EST_String::cat(full_sub," ",sub); pp->insert_after(); pps = pp->as_relation("SylStructure"); pp = pp->next(); // insert in SylStructure as well. pps->insert_after(pp); pp->set("name",sub); set_backoff(pp); if(head.matches(default_match)) set_defaultbackoff(pp); } EST_warning("Missing diphone: %s_%s. Changing %s to %s.\n", (const char *)n1, (const char *)n2, (const char *)bo, (const char *)full_sub); done = true; } p = p->next(); } if (done) return 0; else return 1; } ostream& DiphoneBackoff::print(ostream &st) const { EST_Litem *p; for (p = backofflist.head(); p != 0; p = p->next()) st << backofflist(p); return st; } ostream& operator << (ostream &st, const DiphoneBackoff dbo) { dbo.print(st); return st; } void DiphoneBackoff::set_defaultbackoff(EST_Item *it) const { it->set("defaultbackoff",1); } void DiphoneBackoff::set_backoff(EST_Item *it) const { if(it->f_present("backoff")) it->set("backoff",it->I("backoff")+1); else it->set("backoff",1); } int DiphoneBackoff::is_defaultbackoff(const EST_Item *it) const { if(it->f_present("defaultbackoff")) return 1; else return 0; } int DiphoneBackoff::is_backoff(const EST_Item *it) const { if(it->f_present("backoff")) return 1; else return 0; } festival/src/modules/MultiSyn/EST_DiphoneCoverage.cc0000644000175000017500000001425410240420117021103 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) */ /* */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Rob Clark */ /* Date: April 2005 */ /* --------------------------------------------------------------------- */ /* */ /* */ /* */ /* */ /* */ /* */ /*************************************************************************/ #include #include "EST_rw_status.h" #include "festival.h" #include "ling_class/EST_Item.h" #include "EST_TargetCost.h" #include "EST_DiphoneCoverage.h" /* * This file contains some functions which may be useful for diphone * selection. Not quite sure exactly how yet... * Target cost should probably be rewritten in terms of these functions. */ static EST_String pos_map[4] = { "INI", "MED", "FIN", "INT" } ; static EST_String stress_map[4] = { "UU" , "US" , "SU" , "SS"} ; static EST_String get_diphone_name(const EST_Item *seg1); static int get_stress_index(const EST_Item *seg1); static EST_String get_stress_name(const int index); static int get_syl_pos_index(const EST_Item *seg1); static EST_String get_syl_pos_name(const int index); void EST_DiphoneCoverage::add_stats(const EST_Utterance *utt) { EST_Relation *segs = utt->relation("Segment"); EST_Item *it=segs->head(); for( ; it->next(); it=it->next() ) if(it->next()) { EST_String key = EST_String::cat(get_diphone_name(it),"-", get_stress_name(get_stress_index(it)),"-", get_syl_pos_name(get_syl_pos_index(it))); int val = 0; if (strhash.present(key)) { val = strhash.val(key); strhash.remove_item(key); } ++val; strhash.add_item(key,val); } } void EST_DiphoneCoverage::print_stats(const EST_String filename) { ostream *outf; if (filename == "-") outf = &cout; else outf = new ofstream(filename); EST_THash::Entries them; for(them.begin(strhash); them; them++) *outf << them->k << " " << them->v << "\n"; if (outf != &cout) delete outf; } static EST_String get_diphone_name(const EST_Item *seg1) { return EST_String::cat(seg1->S("name"),"_",seg1->next()->S("name")); } static int get_stress_index(const EST_Item *seg1) { int i1 = 0, i2=0; if( ph_is_vowel(seg1->S("name")) && !ph_is_silence(seg1->S("name")) ) i1 = (parent(seg1,"SylStructure")->I("stress") > 0) ? 1 : 0; if( ph_is_vowel(seg1->next()->S("name")) && !ph_is_silence(seg1->next()->S("name")) ) i2 = (parent(seg1->next(),"SylStructure")->I("stress") > 0) ? 1 : 0; return i2+2*i1; } static EST_String get_stress_name(const int index) { return stress_map[index] ; } static int get_syl_pos_index(const EST_Item *seg1) { int pos = TCPOS_MEDIAL; const EST_Item *syl = parent(seg1,"SylStructure"); const EST_Item *next_syl = parent(seg1->next(),"SylStructure"); const EST_Item *next_next_syl = parent(seg1->next()->next(),"SylStructure"); const EST_Item *prev_syl = parent(seg1->prev(),"SylStructure"); if( syl != next_syl ) pos = TCPOS_INTER; else if( syl != prev_syl) pos = TCPOS_INITIAL; else if( next_syl != next_next_syl) pos = TCPOS_FINAL; return pos; } static EST_String get_syl_pos_name(const int index) { return pos_map[index] ; } festival/src/modules/MultiSyn/EST_FlatTargetCost.cc0000644000175000017500000003557511366534446020765 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Rob Clark */ /* Copyright (c) 2006 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Rob Clark */ /* Date: June 2006 */ /* --------------------------------------------------------------------- */ /* */ /* */ /* */ /* */ /* */ /* */ /*************************************************************************/ #include #include "festival.h" #include "ling_class/EST_Item.h" #include "EST_FlatTargetCost.h" #include "siod.h" static const int simple_phone(const EST_String&); static const int simple_id(const EST_String&); static const int simple_pos(const EST_String &s); static const int simple_punc(const EST_String &s); static const int get_bad_f0(const EST_Item *seg); static const EST_Item* tc_get_syl(const EST_Item *seg); static const EST_Item* tc_get_word(const EST_Item *seg); static EST_TStringHash phonehash(10); static int phone_count=0; //VAL_REGISTER_TYPE(tcdata,TCData) /* * BASE CLASS: EST_TargetCost */ /* Individual cost functions */ TCData *EST_FlatTargetCost::flatpack(EST_Item *seg) const { const EST_Item *syl, *nsyl, *nnsyl, *word; TCData *f =new TCData(TCHI_LAST); syl=tc_get_syl(seg); nsyl=tc_get_syl(seg->next()); if(seg->next()->next()) nnsyl=tc_get_syl(seg->next()->next()); else nnsyl = 0; // This segment features //cout << "SEG: " << seg->S("name") << " is vowel: " // << ph_is_vowel(seg->S("name")) << endl; if(ph_is_vowel(seg->S("name"))) (*f)[VOWEL]=1; else (*f)[VOWEL]=0; //cout << "SEG: " << seg->S("name") << " is sil: " // << ph_is_silence(seg->S("name")) << endl; if(ph_is_silence(seg->S("name"))) (*f)[SIL]=1; else (*f)[SIL]=0; if(seg->f_present("bad_dur")) (*f)[BAD_DUR]=1; else (*f)[BAD_DUR]=0; if(seg->next()->f_present("bad_dur")) (*f)[NBAD_DUR]=1; else (*f)[NBAD_DUR]=0; if(seg->f_present("bad_lex")) (*f)[BAD_OOL]=1; else (*f)[BAD_OOL]=0; if(seg->next()->f_present("bad_lex")) (*f)[NBAD_OOL]=1; else (*f)[NBAD_OOL]=0; (*f)[BAD_F0]=get_bad_f0(seg); // This segments syl features if(syl) { (*f)[SYL]=simple_id(syl->S("id")); (*f)[SYL_STRESS]=syl->I("stress"); //cout << "syl id: " << simple_id(syl->S("id")) //<< " stress: " << syl->I("stress") << endl; } else { (*f)[SYL]=0; (*f)[SYL_STRESS]=0; //cout << "no syl present " << endl; } // Next segment features //cout << "NSEG: " << seg->next()->S("name") << " is sil: " // << ph_is_silence(seg->next()->S("name")) << endl; if(ph_is_silence(seg->next()->S("name"))) (*f)[N_SIL]=1; else (*f)[N_SIL]=0; //cout << "NSEG: " << seg->next()->S("name") << " is vowel: " // << ph_is_vowel(seg->next()->S("name")) << endl; if(ph_is_vowel(seg->next()->S("name"))) (*f)[N_VOWEL]=1; else (*f)[N_VOWEL]=0; // Next seg syl features if(nsyl) { (*f)[NSYL]=simple_id(nsyl->S("id")); (*f)[NSYL_STRESS]=nsyl->I("stress"); //cout << "nsyl stress: " << nsyl->I("stress") << endl; } else { (*f)[NSYL]=0; (*f)[NSYL_STRESS]=0; //cout << "no nsyl: " << endl; } if(seg->next()->next()) { //cout << "RC: " << seg->next()->next()->S("name") //<< " " << simple_phone(seg->next()->next()->S("name")) // << endl; (*f)[RC]=simple_phone(seg->next()->next()->S("name")); (*f)[NNBAD_DUR]=seg->next()->next()->f_present("bad_dur"); } else { //cout << "NO RC\n"; (*f)[RC]=0; (*f)[NNBAD_DUR]=0; } // Next next seg syl features. if(nnsyl) { (*f)[NNSYL]=simple_id(nnsyl->S("id")); } else (*f)[NNSYL]=0; // Prev seg syl feature if(seg->prev()) { (*f)[LC]=simple_phone(seg->prev()->S("name")); (*f)[PBAD_DUR]=seg->prev()->f_present("bad_dur"); } else { (*f)[LC]=0; (*f)[PBAD_DUR]=0; } if(seg->prev() && (syl=tc_get_syl(seg->prev()))) (*f)[PSYL]=simple_id(syl->S("id")); else (*f)[PSYL]=0; // seg word feature if((word=tc_get_word(seg))) (*f)[WQRD]=simple_id(word->S("id")); else (*f)[WQRD]=0; // Next seg word features if((word=tc_get_word(seg->next()))) (*f)[NWQRD]=simple_id(word->S("id")); else (*f)[NWQRD]=0; // next next seg word feature if(seg->next()->next() && (word=tc_get_word(seg->next()->next()))) (*f)[NNWQRD]=simple_id(word->S("id")); else (*f)[NNWQRD]=0; // Prev seg word feature if(seg->prev() && (word=tc_get_word(seg->prev()))) (*f)[PWQRD]=simple_id(word->S("id")); else (*f)[PWQRD]=0; // segs sylpos (*f)[SYLPOS]=0; // medial if( f->a_no_check(SYL)!= f->a_no_check(NSYL) ) (*f)[SYLPOS]=1; // inter else if( f->a_no_check(SYL)!= f->a_no_check(PSYL) ) (*f)[SYLPOS]=2; // initial else if( f->a_no_check(NSYL) != f->a_no_check(NNSYL) ) (*f)[SYLPOS]=3; // final // segs wordpos (*f)[WQRDPOS]=0; // medial if( f->a_no_check(WQRD)!= f->a_no_check(NWQRD) ) (*f)[WQRDPOS]=1; // inter else if( f->a_no_check(WQRD)!= f->a_no_check(PWQRD) ) (*f)[WQRDPOS]=2; // initial else if( f->a_no_check(NWQRD) != f->a_no_check(NNWQRD) ) (*f)[WQRDPOS]=3; // final // pbreak if ((word=tc_get_word(seg))) { if ( word->S("pbreak") == "NB" ) (*f)[PBREAK]=0; else if ( word->S("pbreak") == "B" ) (*f)[PBREAK]=1; else (*f)[PBREAK]=2; } else (*f)[PBREAK]=-1; // seg punc and pos if((word=tc_get_word(seg))) { (*f)[POS]=simple_pos(word->S("pos")); (*f)[PUNC]=simple_punc(parent(word,"Token")->S("punc","NONE")); } else { (*f)[POS]=-1; (*f)[PUNC]=-1; } // next seg punc and pos if ((word=tc_get_word(seg->next()))) { (*f)[NPOS]=simple_pos(word->S("pos")); (*f)[NPUNC]=simple_punc(parent(word,"Token")->S("punc","NONE")); } else { (*f)[NPOS]=-1; (*f)[NPUNC]=-1; } return f; //seg->set_val("tcdata",est_val(f)); // copied? } float EST_FlatTargetCost::stress_cost() const { if( t->a_no_check(VOWEL) && ! t->a_no_check(SIL)) { // Can't assume candidate and target identities are the same // (because of backoff to a silence for example) if( c->a_no_check(SYL) == 0 || c->a_no_check(NSYL) ) return 1.0; if ( t->a_no_check(SYL_STRESS) != c->a_no_check(SYL_STRESS) ) return 1.0; if ( t->a_no_check(NSYL_STRESS) != c->a_no_check(NSYL_STRESS) ) return 1.0; } return 0.0; } float EST_FlatTargetCost::position_in_phrase_cost() const { if ( !t->a_no_check(WQRD) && !c->a_no_check(WQRD) ) return 0; if ( !t->a_no_check(WQRD) || !c->a_no_check(WQRD) ) return 1; return ( t->a_no_check(PBREAK) == c->a_no_check(PBREAK) ) ? 0 : 1; } float EST_FlatTargetCost::punctuation_cost() const { float score = 0.0; if ( (t->a_no_check(WQRD) && !c->a_no_check(WQRD)) || (!t->a_no_check(WQRD) && c->a_no_check(WQRD)) ) score += 0.5; else if (t->a_no_check(WQRD) && c->a_no_check(WQRD)) if ( t->a_no_check(PUNC) != c->a_no_check(PUNC) ) score += 0.5; if ( (t->a_no_check(NWQRD) && !c->a_no_check(NWQRD)) || (!t->a_no_check(NWQRD) && c->a_no_check(NWQRD)) ) score += 0.5; else if(t->a_no_check(NWQRD) && c->a_no_check(NWQRD)) if ( t->a_no_check(NPUNC) != c->a_no_check(NPUNC) ) score += 0.5; return score; } float EST_FlatTargetCost::partofspeech_cost() const { // Compare left phone half of diphone if(!t->a_no_check(WQRD) && !c->a_no_check(WQRD)) return 0; if(!t->a_no_check(WQRD) || !c->a_no_check(WQRD)) return 1; if( t->a_no_check(POS) != c->a_no_check(POS) ) return 1; // Compare right phone half of diphone if(!t->a_no_check(NWQRD) && !c->a_no_check(NWQRD)) return 0; if(!t->a_no_check(NWQRD) || !c->a_no_check(NWQRD)) return 1; if( t->a_no_check(NPOS) != c->a_no_check(NPOS) ) return 1; return 0; } float EST_FlatTargetCost::out_of_lex_cost() const { // bad_dur may at some stage be set on a target for resynthesis purposes. if( c->a_no_check(BAD_OOL) != t->a_no_check(BAD_OOL) ) return 1.0; if( c->a_no_check(NBAD_OOL) != t->a_no_check(NBAD_OOL) ) return 1.0; return 0.0; } float EST_FlatTargetCost::bad_duration_cost() const { // bad_dur may at some stage be set on a target for resynthesis purposes. if( c->a_no_check(BAD_DUR) != t->a_no_check(BAD_DUR) ) return 1.0; if( c->a_no_check(NBAD_DUR) != t->a_no_check(NBAD_DUR) ) return 1.0; // If the segments next to these segments are bad, then these ones are probably wrong too! if( c->a_no_check(PBAD_DUR) != t->a_no_check(PBAD_DUR) ) return 1.0; if( c->a_no_check(NNBAD_DUR) != t->a_no_check(NNBAD_DUR) ) return 1.0; return 0.0; } /* * DERIVED CLASS: EST_FlatTargetCost * * This is CSTR's proposed default target cost flat packed. Nothing * special, if you think you can do better derive your own class. */ float EST_FlatTargetCost::operator()(const TCData* targ, const TCData* cand) const { set_t_and_c(targ,cand); score = 0.0; weight_sum = 0.0; score += add_weight(10.0)*stress_cost(); score += add_weight(5.0)*position_in_syllable_cost(); score += add_weight(5.0)*position_in_word_cost(); score += add_weight(6.0)*partofspeech_cost(); score += add_weight(15.0)*position_in_phrase_cost(); score += add_weight(4.0)*left_context_cost(); score += add_weight(3.0)*right_context_cost(); score /= weight_sum; // These are considered really bad, and will result in a score > 1. score += 10.0*bad_duration_cost(); // see also join cost. score += 10.0*bad_f0_cost(); score += 10.0*punctuation_cost(); return score ; } /* * Auxillary target cost functions */ static const int simple_phone(const EST_String &phone) { if(phonehash.present(phone)) return phonehash.val(phone); phonehash.add_item(phone,++phone_count); return phone_count; } static const int simple_id(const EST_String &id) { return id.after("_").Int(); } static const int simple_punc(const EST_String &punc) { if ( punc == "NONE") return 0; else if ( punc == "," || punc == ":" || punc == ";" ) return 1; else if ( punc == "\"" || punc == "'" || punc == "-" ) return 1; else if ( punc == "(" || punc == ")" ) return 1; else if ( punc == ".") return 2; else if ( punc == "?") return 3; else return 0; } static const int simple_pos(const EST_String &s) { if( s == "nn" || s == "nnp" || s == "nns" || s == "nnps" || s == "fw" || s == "sym" || s == "ls") return 1; if( s == "vbd" || s == "vb" || s == "vbn" || s == "vbz" || s == "vbp" || s == "vbg") return 2; if( s == "jj" || s == "jjr" || s == "jjs" || s == "1" || s == "2" || s == "rb" || s == "rp" || s == "rbr" || s == "rbs") return 3; return 0; } static const int get_bad_f0(const EST_Item *seg) { // by default, the last element of join cost coef vector is // the f0 (i.e. fv->a_no_check( fv->n()-1 ) ) EST_String left(seg->S("name")); EST_String right(seg->next()->S("name")); EST_FVector *fv = 0; int penalty = 0; if( seg->f_present("midcoef") && ( ph_is_vowel( left ) || ph_is_approximant( left ) || ph_is_liquid( left ) || ph_is_nasal( left ) )){ fv = fvector( seg->f("midcoef") ); if( fv->a_no_check(fv->n()-1) == -1.0 ) // means unvoiced penalty += 1; } if( seg->next()->f_present("midcoef") && ( ph_is_vowel( right ) || ph_is_approximant( right ) || ph_is_liquid( right ) || ph_is_nasal( right ) ) ){ fv = fvector( seg->next()->f("midcoef") ); if( fv->a_no_check(fv->n()-1) == -1.0 ) // means unvoiced penalty += 1; } return penalty/2; } static const EST_Item *tc_get_syl(const EST_Item *seg) { // if(!seg) // return 0; return parent(seg,"SylStructure"); } static const EST_Item *tc_get_word(const EST_Item *seg) { // if(!seg) // return 0; const EST_Item *syl = tc_get_syl(seg); if(syl) return parent(syl,"SylStructure"); else return 0; } festival/src/modules/MultiSyn/UnitSelection.h0000644000175000017500000000641110055371305017761 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: Aug 2002 */ /* --------------------------------------------------------------------- */ /* */ /*************************************************************************/ #ifndef __UNITSELECTION_H__ #define __UNITSELECTION_H__ #include "festival.h" /**@name Generalis[ed/able] Unit Selection concatenative synthesis */ //@{ /** Object oriented approach for better or for worse... */ void register_UnitSelection_features(void); #endif // __UNITSELECTION_H__ festival/src/modules/MultiSyn/VoiceBase.h0000644000175000017500000001130311342750601017030 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: Aug 2002 */ /* --------------------------------------------------------------------- */ /* Abstract base class for "voices" - a top level interface to any code */ /* which knows how to take a preprocessed utterance object and fill in */ /* the information for subsequent synthesis by later festival modules */ /*************************************************************************/ #ifndef __VOICEBASE_H__ #define __VOICEBASE_H__ #include "EST_Val.h" #include "EST_Val_defs.h" // EST_TKVL.h is necessary because of header dependencies // which should probably be fixed at root, then this include // could be removed... #include "EST_TKVL.h" #include "siod_defs.h" class EST_Utterance; class VoiceBase { public: VoiceBase() : _verbosity(0), _name( EST_String::Empty ) {}; virtual ~VoiceBase() {}; virtual void initialise( bool ignore_bad_tag=false ) = 0; virtual EST_String name() { return _name; } virtual void set_name(EST_String n) { _name = n;} virtual void setVerbosity( unsigned int level ) { _verbosity=level; } virtual unsigned int verbosity() const { return _verbosity; } //virtual bool synthesiseWave( EST_Utterance *utt ) = 0; // this function should at best be moved to concatenative voice // subclass and the above one implemented instead in order to // generalise to non concatenative synthesis methods virtual void getUnitSequence( EST_Utterance *utt )=0; virtual unsigned int numDatabaseUnits() const = 0; virtual unsigned int numUnitTypes() const = 0; virtual bool unitAvailable( const EST_String &unit ) const = 0; virtual unsigned int numAvailableCandidates( const EST_String &unit ) const =0; private: //EST_Features params; unsigned int _verbosity; EST_String _name; }; SIOD_REGISTER_CLASS_DCLS(voice,VoiceBase) VAL_REGISTER_CLASS_DCLS(voice,VoiceBase) /**@name Generalis[ed/able] notion of a "voice" within festival */ //@{ /** Object oriented approach for better or for worse... */ #endif // __VOICEBASE_H__ festival/src/modules/MultiSyn/DiphoneUnitVoice.h0000644000175000017500000002112310613143414020403 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: Aug 2002 */ /* --------------------------------------------------------------------- */ /* first stab at a diphone unit selection "voice" - using a list of */ /* utterance objects */ /*************************************************************************/ #ifndef __DIPHONEUNITVOICE_H__ #define __DIPHONEUNITVOICE_H__ #include "VoiceBase.h" #include "DiphoneBackoff.h" #include "siod_defs.h" #include "EST_Val_defs.h" #include "EST_String.h" #include "EST_FlatTargetCost.h" #include "EST_types.h" // for EST_StrList class EST_Utterance; class EST_Relation; class EST_VTCandidate; class EST_VTPath; class EST_Features; class EST_Track; class EST_Wave; class EST_Item; class DiphoneVoiceModule; class EST_JoinCost; #include "EST_THash.h" template class EST_TList; typedef EST_TList ItemList; SIOD_REGISTER_TYPE_DCLS(itemlist, ItemList) VAL_REGISTER_TYPE_DCLS(itemlist, ItemList) SIOD_REGISTER_CLASS_DCLS(du_voice,DiphoneUnitVoice) VAL_REGISTER_CLASS_DCLS(du_voice,DiphoneUnitVoice) class DiphoneUnitVoice : public VoiceBase { public: DiphoneUnitVoice( const EST_StrList& basenames, const EST_String& uttDir, const EST_String& wavDir, const EST_String& pmDir, const EST_String& coefDir, unsigned int srate = 16000, const EST_String& uttExt = ".utt", const EST_String& wavExt = ".wav", const EST_String& pmExt = ".pm", const EST_String& coefExt = ".coef" ); virtual ~DiphoneUnitVoice(); virtual void initialise( bool ignore_bad_tag=false ); virtual unsigned int numDatabaseUnits() const; virtual unsigned int numUnitTypes() const; virtual bool synthesiseWave( EST_Utterance *utt ); virtual void getUnitSequence( EST_Utterance *utt ); void regetUnitSequence( EST_Utterance *utt ); void getCopyUnitUtterance( const EST_String &utt_fname, EST_Utterance **utt_out ) const; EST_VTCandidate* getCandidates( EST_Item *s, EST_Features &f ) const; void diphoneCoverage(const EST_String filename) const; virtual bool unitAvailable( const EST_String &diphone ) const; virtual unsigned int numAvailableCandidates( const EST_String &unit ) const; unsigned int numModules() const { return voiceModules.length(); } bool addVoiceModule( const EST_StrList& basenames, const EST_String& uttDir, const EST_String& wavDir, const EST_String& pmDir, const EST_String& coefDir, unsigned int srate = 16000, const EST_String& uttExt = ".utt", const EST_String& wavExt = ".wav", const EST_String& pmExt = ".pm", const EST_String& coefExt = ".coef" ); // assume responsibility to delete vm when done with it void registerVoiceModule( DiphoneVoiceModule *vm ); // del=true means it's ok to delete the join cost when we're done // with it void setJoinCost( EST_JoinCost *jcost, bool del=false ); const EST_JoinCost& getJoinCostCalculator( ) const { return *jc; } void setTargetCost( EST_TargetCost *tcost, bool del=false ); const EST_TargetCost& getTargetCostCalculator( ) const { return *tc; } void set_pruning_beam( float width ) { pruning_beam=width; } float get_pruning_beam( ) const { return pruning_beam; } void set_ob_pruning_beam( float width ){ ob_pruning_beam=width; } float get_ob_pruning_beam( ) const { return ob_pruning_beam; } void set_jc_f0_weight( float val ) { jc_f0_weight=val; } float get_jc_f0_weight() { return jc_f0_weight; } EST_JoinCost * get_jc() { return jc; } void set_jc_power_weight( float val ) { jc_power_weight=val; } float get_jc_power_weight() { return jc_power_weight; } void set_jc_spectral_weight( float val ) { jc_spectral_weight=val; } float get_jc_spectral_weight() { return jc_spectral_weight; } void set_tc_rescoring_beam( float width ){ tc_rescoring_beam = width; } float get_tc_rescoring_beam( ) const { return tc_rescoring_beam; } void set_tc_rescoring_weight( float weight ){ tc_rescoring_weight = weight; } float get_tc_rescoring_weight( ) const { return tc_rescoring_weight; } void set_target_cost_weight( float w ){ tc_weight=w; } float get_target_cost_weight() const { return tc_weight; } void set_join_cost_weight( float w ){ jc_weight=w; } float get_join_cost_weight() const { return jc_weight; } void set_prosodic_modification( int m ){ prosodic_modification=m; } int get_prosodic_modification() const { return prosodic_modification; } void set_wav_samplerate( unsigned int sr ) { wav_srate = sr; } unsigned int get_wav_samplerate( ) const { return wav_srate; } void precomputeJoinCosts( const EST_StrList &phones, bool verbose=true ); private: // don't allow copying of Voices (for now?) DiphoneUnitVoice( const DiphoneUnitVoice& ); DiphoneUnitVoice& operator=( const DiphoneUnitVoice& ); void addToCatalogue( const EST_Utterance *utt ); void getDiphone( const EST_VTCandidate *cand, EST_Track* coef, EST_Wave* sig, int *midframe, bool extendLeft=0, bool extendRight=0 ); int getPhoneList( const EST_String &phone, ItemList &list ); void fillUnitRelation( EST_Relation *units, const EST_VTPath *path ); private: EST_TList voiceModules; float pruning_beam; // beam pruning float ob_pruning_beam; // observation beam pruning float tc_rescoring_beam; float tc_rescoring_weight; float tc_weight; float jc_weight; float jc_f0_weight; // join cost f0 weight float jc_power_weight; // join cost f0 weight float jc_spectral_weight; // join cost spectral weight int prosodic_modification; unsigned int wav_srate; EST_JoinCost *jc; bool jc_delete; EST_TargetCost *tc; bool tc_delete; TCDataHash *tcdh; private: DiphoneBackoff *diphone_backoff_rules; // diphone backoff rules public: void set_diphone_backoff(DiphoneBackoff *dbo); }; #endif // __DIPHONEUNITVOICE_H__ festival/src/modules/MultiSyn/VoiceModuleBase.h0000644000175000017500000001065710466376642020230 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ //*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: Aug 2002 */ /* --------------------------------------------------------------------- */ /* Abstract base class for "voice modules" - a top level interface to */ /* any collection of speech material in the voice database */ /*************************************************************************/ #ifndef __VOICEMODULEBASE_H__ #define __VOICEMODULEBASE_H__ #include "EST_Val.h" #include "EST_Val_defs.h" // EST_TKVL.h is necessary because of header dependencies // which should probably be fixed at root, then this include // could be removed... #include "EST_TKVL.h" #include "siod_defs.h" #include "EST_FlatTargetCost.h" class EST_Utterance; class VoiceBase; class VoiceModuleBase { public: VoiceModuleBase( VoiceBase *parent = 0 ): _parentVoice(parent) {}; virtual ~VoiceModuleBase() {}; virtual void initialise(const EST_TargetCost *tc, bool ignore_bad_tag=false) = 0; virtual unsigned int numModuleUnits() const = 0; virtual unsigned int numUnitTypes() const = 0; virtual unsigned int numAvailableCandidates( const EST_String &unit ) const =0; private: //EST_Features params; const VoiceBase *_parentVoice; }; SIOD_REGISTER_CLASS_DCLS(voicemodule,VoiceModuleBase) VAL_REGISTER_CLASS_DCLS(voicemodule,VoiceModuleBase) /**@name Generalis[ed/able] notion of a unit selection "voice" module within festival. */ //@{ /** Object oriented approach for better or for worse... This is primarily intended to be a "module" from the point of view of a collection of *related* speech database material (such as a "weather" module) or a "news broadcast" module. At the very least, though, it is just a collection of speech database material. */ #endif // __VOICEMODULEBASE_H__ festival/src/modules/MultiSyn/DiphoneVoiceModule.h0000644000175000017500000001742310470644446020735 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: Aug 2002 */ /* --------------------------------------------------------------------- */ /* A diphone unit selection "voice module" */ /* (implemented using a list of utterance objects) */ /*************************************************************************/ #ifndef __DIPHONEVOICEMODULE_H__ #define __DIPHONEVOICEMODULE_H__ #include "VoiceModuleBase.h" #include "EST_DiphoneCoverage.h" #include "siod_defs.h" #include "EST_Val_defs.h" #include "EST_String.h" #include "EST_viterbi.h" #include "EST_types.h" // for EST_StrList #include "EST_FlatTargetCost.h" class EST_Utterance; class EST_Relation; class EST_VTCandidate; class EST_VTPath; class EST_Features; class EST_Track; class EST_Wave; class EST_Item; // return standard join point time for this segment // (one half of a diphone) float getJoinTime( const EST_Item *seg ); //template class EST_TStringHash; #include "EST_THash.h" template class EST_TList; typedef EST_TList ItemList; SIOD_REGISTER_CLASS_DCLS(du_voicemodule,DiphoneVoiceModule) VAL_REGISTER_CLASS_DCLS(du_voicemodule,DiphoneVoiceModule) // following is necessary to make some of a candidate's information // available in a faster way that EST_Item feature lookups (critically // the join cost coefficients EST_FVectors for example) // i.e. yet another temporary hack... (would be better if EST_viterbi // code allowed other things apart from just EST_Item* in order to // perform the search) class DiphoneCandidate { public: DiphoneCandidate( const EST_Item *phone1, const DiphoneVoiceModule *p, const EST_FVector *left, const EST_FVector *right ) : ph1(phone1), dvm( p ), l_coef(left), r_coef(right), ph1_jccid(-1), ph1_jccindex(-1), ph2_jccid(-1), ph2_jccindex(-1){}; const EST_Item *ph1; const DiphoneVoiceModule *dvm; const EST_FVector *l_coef; const EST_FVector *r_coef; int ph1_jccid, ph1_jccindex; int ph2_jccid, ph2_jccindex; }; VAL_REGISTER_CLASS_DCLS(diphonecandidate,DiphoneCandidate) class DiphoneVoiceModule : public VoiceModuleBase { public: DiphoneVoiceModule( const EST_StrList& basenames, const EST_String& uttDir, const EST_String& wavDir, const EST_String& pmDir, const EST_String& coefDir, unsigned int srate = 16000, const EST_String& uttExt = ".utt", const EST_String& wavExt = ".wav", const EST_String& pmExt = ".pm", const EST_String& coefExt = ".coef" ); virtual ~DiphoneVoiceModule(); virtual void initialise(const EST_TargetCost *tc, bool ignore_bad_tag=false ); virtual unsigned int numModuleUnits() const; virtual unsigned int numUnitTypes() const; virtual unsigned int numAvailableCandidates( const EST_String &unit ) const; ///// Some "debugging" functions - deliberately don't mind doing // slow things like returning copies of things. Such functions are // not intended to do important things, but just to make it easier // to work out whats "in" the voice database object. // return copy of utterance number bool getUtterance( EST_Utterance **utt, int n ) const; // return pointer to utterance which has feature "feat_name" // set to value "value" bool getUtterance( EST_Utterance **utt, const EST_String &feat_name, const EST_Val &value ) const; void getDiphoneCoverageStats(EST_DiphoneCoverage *dc) const; // int DiphoneVoiceModule::getCandidateList( const EST_Item& target, // const EST_TargetCost& tc, // EST_VTCandidate *head, // EST_VTCandidate *tail ) const; int getCandidateList( const EST_Item& target, const EST_TargetCost *tc, const TCDataHash *tcdh, const float tc_weight, EST_VTCandidate **head, EST_VTCandidate **tail ) const; // append all instances of a certain phone present in the utterances // in this voice. Returns the number added int getPhoneList( const EST_String &phone, ItemList &list ); private: // don't allow copying of Voices (for now?) DiphoneVoiceModule( const DiphoneVoiceModule& ); DiphoneVoiceModule& operator=( const DiphoneVoiceModule& ); // Flatpack void flatPack( EST_Relation *segs, const EST_TargetCost *tc) const; void addCoefficients( EST_Relation *segs, const EST_Track& coefs ); void addToCatalogue( const EST_Utterance *utt, int *num_ignored, bool ignore_bad=false ); void getDiphone( const EST_Item *phone1, EST_Track* coef, EST_Wave* sig, int* midframe, bool extendLeft=0, bool extendRight=0 ) const; friend class DiphoneUnitVoice; private: EST_StrList fileList; EST_String utt_dir; // utterance files EST_String utt_ext; EST_String pm_dir; // pitch marks EST_String pm_ext; EST_String coef_dir; // for coefficients that aren't pitch syncronous EST_String coef_ext; EST_String wave_dir; // waveform (or residual) EST_String wave_ext; unsigned int wav_srate; //sample rate of voice waveform data TCDataHash *tcdatahash; EST_TList *utt_dbase; EST_TStringHash *catalogue; }; #endif // __DIPHONEVOICEMODULE_H__ festival/src/modules/MultiSyn/EST_TargetCost.h0000644000175000017500000001423410576233724020002 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: October 2002 */ /* --------------------------------------------------------------------- */ /* Interface for family of target cost function objects which */ /* calculate a target score for given candidate */ /* */ /* */ /* */ /* */ /*************************************************************************/ #ifndef __EST_TARGETCOST_H__ #define __EST_TARGETCOST_H__ #include "siod.h" // for gc_protect( obj**) class EST_Item; /**@name Interface for Target Cost function object */ //@{ /** Object oriented approach for better or for worse... */ /* Positional enum */ enum tcpos_t { TCPOS_INITIAL, TCPOS_MEDIAL, TCPOS_FINAL, TCPOS_INTER, TCPOS_POSITIONS }; /* * BASE CLASS: EST_TargetCost */ class EST_TargetCost { public: EST_TargetCost() : defScore(0.0){}; virtual ~EST_TargetCost() {}; // Base class operator() doesn't do much, but it will work. virtual float operator()( const EST_Item* targp, const EST_Item* candp ) const { return defScore; } // Allow flatpacking virtual const bool is_flatpack() const {return false;} protected: float defScore; // Temp variables for use while calculating cost (here for speed) mutable float score; mutable float weight_sum; mutable const EST_Item *cand; mutable const EST_Item *targ; inline void set_cand(const EST_Item* seg) const { cand = seg; } inline void set_targ(const EST_Item* seg) const { targ = seg; } inline void set_targ_and_cand(const EST_Item* tseg, const EST_Item* cseg) const { set_targ(tseg); set_cand(cseg); } inline float add_weight(float w) const { weight_sum += w ; return w; } // General cost functions that derived classes may want to use. float apml_accent_cost() const; float stress_cost() const; float position_in_syllable_cost() const; float position_in_word_cost() const; float position_in_phrase_cost() const; float partofspeech_cost() const; float punctuation_cost() const; float left_context_cost() const; float right_context_cost() const; float bad_duration_cost() const; float out_of_lex_cost() const; float bad_f0_cost() const; }; /* * DERIVED CLASS: EST_DefaultTargetCost */ class EST_DefaultTargetCost : public EST_TargetCost { public: float operator()(const EST_Item* targ, const EST_Item* cand) const; }; /* * DERIVED CLASS: EST_APMLTargetCost */ class EST_APMLTargetCost : public EST_TargetCost { public: float operator()(const EST_Item* targ, const EST_Item* cand) const; }; /* * DERIVED CLASS: EST_SingingTargetCost */ class EST_SingingTargetCost : public EST_TargetCost { public: float operator()(const EST_Item* targ, const EST_Item* cand) const; protected: float pitch_cost() const; float duration_cost() const; }; /* * DERIVED CLASS: EST_SchemeTargetCost */ class EST_SchemeTargetCost : public EST_TargetCost { private: LISP tc; public: EST_SchemeTargetCost( LISP scheme_targetcost ) : EST_TargetCost(), tc(scheme_targetcost) { gc_protect( &tc ); } ~EST_SchemeTargetCost() { gc_unprotect( &tc ); } float operator()(const EST_Item* targ, const EST_Item* cand) const; }; #endif // __EST_TARGETCOST_H__ festival/src/modules/MultiSyn/TargetCostRescoring.h0000644000175000017500000001212110452235125021122 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: June 2006 */ /* --------------------------------------------------------------------- */ /* */ /* Code to allow the possibility of changing the costs (thus ranking) */ /* assigned to candidates units by the target cost function, typically */ /* based on what those original costs are... */ /* */ /* */ /*************************************************************************/ #ifndef __TARGETCOSTRESCORING_H__ #define __TARGETCOSTRESCORING_H__ #include #include "EST_viterbi.h" // do some additional target cost, based on the acoustic properties of // the units found matching this target unit. This function assumes the // viterbi candidates linked list is in the correct format, with the // correct information available. // toplevel function to call (Changes the candidate list in place.) // beam_width specifies which candidates are included in the // computation (i.e. only those candidates with target costs within // "beam_width" of the best candidate's score). // mult specifies an arbitrary weighting factor to be applied to the penalties // added to the candidates' target cost scores by this function (e.g. a mult of // 0.0 will stop the target cost scores being changed at all, whereas a mult of 10.0 // will make it have a large effect). void rescoreCandidates( EST_VTCandidate *candidates, float beam_width, float mult ); // our internal class for doing the work class ScorePair { public: ScorePair( ) : _score(0.0), _dur(0.0), _cand(0) {}; ScorePair( float score, float duration, EST_VTCandidate* cand ) : _score(score), _dur(duration), _cand(cand) {}; public: float _score; float _dur; EST_VTCandidate* _cand; }; inline bool operator > ( const ScorePair &A, const ScorePair &B ) { return ( A._score > B._score ) ? true : false; } inline bool operator < ( const ScorePair &A, const ScorePair &B ) { return ( A._score < B._score ) ? true : false; } inline bool operator == ( const ScorePair &A, const ScorePair &B ) { return ( A._score == B._score ) ? true : false; } ostream& operator<<( ostream& out, const ScorePair &sp ); #endif // __TARGETCOSTRESCORING_H__ festival/src/modules/MultiSyn/EST_JoinCost.h0000644000175000017500000002464610613143414017447 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: October 2002 */ /* --------------------------------------------------------------------- */ /* Interface for family of join cost function objects which */ /* calculate a join score for given candidates */ /* */ /* */ /* */ /* */ /*************************************************************************/ #ifndef __EST_JOINCOST_H__ #define __EST_JOINCOST_H__ class EST_Item; class EST_JoinCostCache; /////////////////////////////// //because of the inline #include "EST_JoinCostCache.h" #include "DiphoneVoiceModule.h" #include "safety.h" #include "ling_class/EST_Item.h" #include "EST_FMatrix.h" /////////////////////////////// /**@name Interface for Join Cost function object */ //@{ /** Object oriented approach for better or for worse... */ class EST_JoinCost { public: EST_JoinCost() : defCost(1), f0_weight (1.0), power_weight(1.0), spectral_weight(1.0) {}; ~EST_JoinCost(); bool computeAndCache( const EST_TList &list, bool verbose=true ); // join cost which avoids information lookup (or faster lookup), to // be used in Viterbi search inline float operator()( const DiphoneCandidate* left, const DiphoneCandidate* right ) const; // original join cost, retained because it's still used by cost calculations for // caching (and for posterity, and comparison) inline float operator()( const EST_Item *left, const EST_Item *right ) const; inline void set_f0_weight(float val) { f0_weight = val; } inline void set_power_weight(float val) { power_weight = val; } inline void set_spectral_weight(float val) { spectral_weight = val; } private: // member data all used by the older implementation (taking EST_Item* not // DiphoneCandidate*) float defCost; mutable const EST_Item *cachedItem; mutable const EST_FVector *cachedItemVector; mutable unsigned int cached_jccid; mutable unsigned int cached_jccindex; mutable bool costIsCached; mutable bool diphoneJoin; //would definitely be better somewhere else... float f0_weight; float power_weight; float spectral_weight; EST_TSimpleVector costCaches; // function which computes the actual cost between two vectors // of join cost coefficients inline float calcDistance( const EST_FVector *l, const EST_FVector *r ) const; }; ////////////////////////////////////////////////////////////////////////////////// // experiment to see if this is sensible or not // // for now, the left and right edges of a join are represented by EST_Items, // // which is particular to the implementation of the DiphoneUnitVoice. // // It would be desirable in the future though to abstract the interface to // // something like concatUnit::right_edge() and concatUnit::left_edge() since // // we would like a given join cost to be able to be generally applied to units of // // any length or form. inline float EST_JoinCost::operator()( const EST_Item* left, const EST_Item* right ) const { float d_overall; //default zero cost if units contiguous in database // (i.e. this is the cost between a phone and *itself* if( left == right->prev() ) return 0.0; // An "infinite" join cost for bad units. The idea here is that if // you use a unit marked as bad (either for duration or // pitchmarks, then the units either side of it must also be used. // if( left->f_present("bad") || right->f_present("bad") // || left->f_present("bad_dur") || right->f_present("bad_dur") ) // return 100.0; // else... // since the Viterbi class takes each path at time t and tries to // extend with all candidates at time t+1, it probably makes sense // to cache the feature vector for the "left" item to avoid looking // it up every time this function is called... if( cachedItem != left ){ cachedItem = left; if( left->f_present( "jccid" ) ){ costIsCached = true; cached_jccid = left->features().val( "jccid" ).Int(); cached_jccindex = left->features().val( "jccindex" ).Int(); } else{ costIsCached = false; if( left->f_present( "extendRight") ){ diphoneJoin = false; cachedItemVector = fvector( left->features().val( "endcoef" ) ); } else{ diphoneJoin = true; cachedItemVector = fvector( left->features().val( "midcoef" ) ); } } } if( costIsCached && right->f_present( "jccid" ) ){ unsigned int right_jccid = right->features().val( "jccid" ).Int(); unsigned int right_jccindex = right->features().val( "jccindex" ).Int(); if( cached_jccid == right_jccid ) d_overall = (float)(costCaches(cached_jccid)->val(cached_jccindex, right_jccindex))/255; else{ EST_warning( "JoinCost: inconsistent cache ids, setting max join cost" ); d_overall = 1.0; } } else{ const EST_FVector *l = cachedItemVector; const EST_FVector *r; if( diphoneJoin ) r = fvector( right->features().val( "midcoef" ) ); else r = fvector( right->features().val( "startcoef" ) ); d_overall = calcDistance( l, r ); } return d_overall; } inline float EST_JoinCost::operator()( const DiphoneCandidate* left, const DiphoneCandidate* right ) const { float dist; //default zero cost if units contiguous in database // (i.e. this is the cost between a phone and *itself* if( left->ph1->next() == right->ph1 ) return 0.0; //use cached costs in preference to calculating if( left->ph2_jccid >= 0 ) if( left->ph2_jccid == right->ph1_jccid ) dist = (float)(costCaches(left->ph2_jccid)->val(left->ph2_jccindex, right->ph1_jccindex))/255; else{ EST_warning( "JoinCost: inconsistent cache ids, setting max join cost" ); dist = 1.0; } else{ dist = calcDistance( left->r_coef, right->l_coef ); } return dist; } inline float EST_JoinCost::calcDistance( const EST_FVector *l, const EST_FVector *r ) const { float d_spectral, d_f0, d_power, d_overall; int l_length = l->length(); if (l_length != r->length()) EST_error("Can't compare vectors of differing length\n"); //////////////////////////////////////////////////////////////////////////// // f0 distance // // because unvoiced is represented as -1.0, we need to take // special measures when calculating f0 distances, to avoid // situation where something low in the speaker's range is closer // to 0.0 than other higher voiced speech. (this could especially // be problematic where bad pitchmarking or labelling occurs) float l_f0 = l->a_no_check( l_length-1 ); float r_f0 = r->a_no_check( l_length-1 ); if( l_f0 != -1.0 ){ if( r_f0 != -1.0 ){ d_f0 = pow((float)( l_f0 - r_f0 ), (float)2.0); d_f0 = sqrt( d_f0 ); } else d_f0 = 1.0; } else if( r_f0 != -1.0 ) d_f0 = 1.0; else d_f0 = 0.0; //////////////////////////////////////////////////////////////////////////// // power distance d_power = pow((float)((l->a_no_check(l_length-2) - r->a_no_check(l_length-2))), (float)2.0); d_power = sqrt( d_power ); //////////////////////////////////////////////////////////////////////////// // spectral distance float d = 0.0; l_length -= 2; // don't include f0 and power for(int i = 0; i < l_length; i++) d += pow((float)(l->a_no_check(i) - r->a_no_check(i)), (float)2.0); d_spectral = sqrt( d ); // equal weighting by default d_overall = (d_f0*f0_weight + d_power*power_weight + d_spectral*spectral_weight) / 3; return d_overall; } #endif // __EST_JOINCOST_H__ festival/src/modules/MultiSyn/EST_JoinCostCache.h0000644000175000017500000001114710127007361020363 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2004 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: January 2004 */ /* --------------------------------------------------------------------- */ /* Data type intended to handle a table of n^2 measures relating to */ /* n distinct entities (e.g. join costs for a phone type). We assume */ /* the measure is symmetric and hence we don't actually store n^2 values */ /* */ /*************************************************************************/ #ifndef __EST_JOINCOSTCACHE_H__ #define __EST_JOINCOSTCACHE_H__ /**@name Caching of join cost computations (need for speed) We assume the distance measure is symmetric and hence we don't actually store n^2 values for the n instances of any given phone. */ //@{ /** Object oriented approach for better and for worse... */ #include "EST_TList.h" #include "ling_class/EST_Item.h" #include using namespace std; class EST_JoinCost; class EST_JoinCostCache { public: EST_JoinCostCache( unsigned int id ); EST_JoinCostCache( unsigned int id, unsigned int n ); EST_JoinCostCache( unsigned int id, unsigned char *mem, unsigned int n, bool del=false ); ~EST_JoinCostCache(); unsigned int id() const {return _id;} ostream& write( ostream &os ) const; unsigned char val( unsigned int a, unsigned int b ) const; bool setval( unsigned int a, unsigned int b, unsigned char v ); bool computeAndCache( const EST_TList &list, const EST_JoinCost &jc, bool verbose=false ); private: EST_JoinCostCache( EST_JoinCostCache &); EST_JoinCostCache& operator=( EST_JoinCostCache &); private: unsigned int numInstances; unsigned int _id; unsigned char* cache; static const unsigned char minVal = 0x0; static const unsigned char maxVal = 0xff; static const unsigned char defVal = 0xff; unsigned int cachelen; bool deleteMemoryOnDeath; }; #endif // __EST_JOINCOSTCACHE_H__ festival/src/modules/MultiSyn/DiphoneBackoff.h0000644000175000017500000000710510055371305020037 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 2004 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Rob Clark */ /* Date: Jan 2004 */ /* --------------------------------------------------------------------- */ /* Diphone backing off procedure */ /*************************************************************************/ #ifndef __DIPHONEBACKOFF_H__ #define __DIPHONEBACKOFF_H__ #include "siod.h" #include "EST_types.h" #include "ling_class/EST_Utterance.h" class DiphoneBackoff { private: static const EST_String default_match; EST_StrListList backofflist; public: DiphoneBackoff(LISP l_backofflist); //~DiphoneBackoff(); EST_String backoff(EST_String left, EST_String right); int backoff(EST_Item *left_phone); ostream& print(ostream &st = cout) const; friend ostream& operator << (ostream &st, const DiphoneBackoff &dbo); private: int is_defaultbackoff(const EST_Item *it) const; void set_defaultbackoff(EST_Item *it) const; int is_backoff(const EST_Item *it) const; void set_backoff(EST_Item *it) const; }; #endif // __DIPHONEBACKOFF_H__ festival/src/modules/MultiSyn/safety.h0000644000175000017500000000646010055371305016473 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: Sept 2002 */ /* --------------------------------------------------------------------- */ /* */ /*************************************************************************/ #ifndef __SAFETY_H__ #define __SAFETY_H__ #include "EST_error.h" #define CHECK_NULL 1 #ifdef CHECK_NULL #define CHECK_PTR(p) if((p)==0){\ EST_error("memory allocation failed (file %s, line %d)",\ __FILE__,__LINE__);} #else #define CHECK_PTR(p) #endif //CHECK_NULL #endif //__SAFETY_H__ festival/src/modules/MultiSyn/EST_DiphoneCoverage.h0000644000175000017500000000731510227157430020757 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) */ /* */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Rob Clark */ /* Date: April 2005 */ /* --------------------------------------------------------------------- */ /* */ /* */ /* */ /* */ /* Diphone Coverage stats class */ /* */ /*************************************************************************/ #ifndef __EST_DIPHONECOVERAGE_H__ #define __EST_DIPHONECOVERAGE_H__ #include "EST_THash.h" class EST_DiphoneCoverage { public: EST_DiphoneCoverage() : strhash(100) {}; private: EST_TStringHash strhash; public: void add_stats(const EST_Utterance *utt); void print_stats(const EST_String filename="-"); }; #endif // __EST_DIPHONECOVERAGE_H__ festival/src/modules/MultiSyn/EST_FlatTargetCost.h0000644000175000017500000001231511366534446020612 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Rob Clark */ /* Copyright (c) 2006 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Rob Clark */ /* Date: June 2006 */ /* --------------------------------------------------------------------- */ /* */ /* */ /* */ /* */ /*************************************************************************/ #ifndef __EST_TARGETCOST_FLAT_H__ #define __EST_TARGETCOST_FLAT_H__ #include "EST_THash.h" #include "EST_TargetCost.h" /* we use WQRD in place of WORD below to keep VC++ happy */ enum tcdata_t { VOWEL, SIL, BAD_DUR, NBAD_DUR, BAD_OOL, NBAD_OOL, BAD_F0, SYL, SYL_STRESS, N_SIL, N_VOWEL, NSYL, NSYL_STRESS, RC, NNBAD_DUR, NNSYL, LC, PBAD_DUR, PSYL, WQRD, NWQRD, NNWQRD, PWQRD, SYLPOS, WQRDPOS, PBREAK, POS, PUNC, NPOS, NPUNC, TCHI_LAST } ; typedef EST_IVector TCData; VAL_REGISTER_TYPE_DCLS(ivector,TCData) typedef EST_THash TCDataHash; /* * DERIVED CLASS: EST_FlatTargetCost */ class EST_FlatTargetCost : public EST_TargetCost { public: EST_FlatTargetCost() : li(0){}; private: mutable const TCData *t; mutable const TCData *c; mutable const EST_Item *li; inline void set_t_and_c(const TCData* targ, const TCData* cand) const { t = targ; c = cand; } float stress_cost() const; inline float position_in_syllable_cost() const { return ( t->a_no_check(SYLPOS) == c->a_no_check(SYLPOS) ) ? 0 : 1; } inline float position_in_word_cost() const { return ( t->a_no_check(WQRDPOS) == c->a_no_check(WQRDPOS) ) ? 0 : 1; } float position_in_phrase_cost() const; float punctuation_cost() const; float partofspeech_cost() const; inline float left_context_cost() const { return ( t->a_no_check(LC) == c->a_no_check(LC) ) ? 0 : 1; } inline float right_context_cost() const { return ( t->a_no_check(RC) == c->a_no_check(RC) ) ? 0 : 1; } float bad_duration_cost() const; float out_of_lex_cost() const; inline float bad_f0_cost() const { return float(c->a_no_check(BAD_F0)) / 2.0; } public: float operator()(const EST_Item* targ, const EST_Item* cand) const { EST_error("EST_FlatTargetCost operator() called with EST_Items\n"); return 1; } float operator()(const TCData *targ, const TCData *cand) const; const bool is_flatpack() const { return true; } TCData *flatpack(EST_Item *seg) const; }; #endif // __EST_TARGETCOST_H__ festival/src/modules/MultiSyn/inst_tmpl/hash_s_itemlistp_t.cc0000644000175000017500000000672310055371305023233 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond (korin@cstr.ed.ac.uk) */ /* Date: September 2002 */ /* ----------------------------------------------------------------------*/ /* Instantiate TStringHash for ItemList* */ /*************************************************************************/ #include "EST_THash.h" #include "EST_String.h" #include "EST_Val.h" #include "DiphoneUnitVoice.h" //for ItemList //Declare_TStringHash_T(ItemList*,ItemListP) //above macro has been disabled, so do it manually. template <> EST_String EST_THash< EST_String, ItemList* >::Dummy_Key="DUMMY"; template <> ItemList* EST_THash< EST_String, ItemList* >::Dummy_Value=0; #if defined(INSTANTIATE_TEMPLATES) #include "../base_class/EST_THash.cc" Instantiate_TStringHash_T(ItemList *, ItemListP) #endif festival/src/modules/MultiSyn/inst_tmpl/hash_itemp_tcdatap_t.cc0000644000175000017500000000665310453677764023543 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Rob Clark */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Rob Clark (robert@cstr.ed.ac.uk) */ /* Date: June 2006 */ /* ----------------------------------------------------------------------*/ /* Instantiate THash for Item* TCData* */ /*************************************************************************/ #include "EST_THash.h" #include "EST_String.h" #include "EST_FlatTargetCost.h" // for TCData //Declare_TStringHash_T(ItemList*,ItemListP) //above macro has been disabled, so do it manually. template <> EST_Item* EST_THash< EST_Item*, TCData* >::Dummy_Key=0; template <> TCData* EST_THash< EST_Item*, TCData* >::Dummy_Value=0; #if defined(INSTANTIATE_TEMPLATES) #include "../base_class/EST_THash.cc" Instantiate_THash_T(EST_Item*, TCData*, TCDataP) #endif festival/src/modules/MultiSyn/inst_tmpl/list_uttp_t.cc0000644000175000017500000000656510055371305021727 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Date: September 2002 */ /* ------------------------------------------------------------------- */ /* Instantiate class list of pointers to EST_Utterance */ /* */ /* Author: Korin Richmond (korin@cstr.ed.ac.uk) */ /*************************************************************************/ #include "EST_TList.h" #include "EST_TSortable.h" #include "EST_Val.h" class EST_Utterance; Declare_TList_T(EST_Utterance *, EST_UtteranceP) #if defined(INSTANTIATE_TEMPLATES) #include "../base_class/EST_TList.cc" #include "../base_class/EST_TSortable.cc" Instantiate_TList_T_MIN(EST_Utterance *, EST_UtteranceP) #endif festival/src/modules/MultiSyn/inst_tmpl/list_itemp_t.cc0000644000175000017500000000653010055371305022041 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Date: September 2002 */ /* --------------------------------------------------------------------- */ /* Instantiate class list of pointers to EST_Utterance */ /* */ /* Author: Korin Richmond (korin@cstr.ed.ac.uk) */ /*************************************************************************/ #include "EST_TList.h" #include "EST_TSortable.h" #include "EST_Val.h" class EST_Item; Declare_TList_T(EST_Item *, EST_ItemP) #if defined(INSTANTIATE_TEMPLATES) #include "../base_class/EST_TList.cc" #include "../base_class/EST_TSortable.cc" Instantiate_TList_T(EST_Item *, EST_ItemP) #endif festival/src/modules/MultiSyn/inst_tmpl/list_voicemodulep_t.cc0000644000175000017500000000661510055371305023422 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2002 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Date: February 2002 */ /* ------------------------------------------------------------------- */ /* Instantiate class list of pointers to DiphoneVoiceModule */ /* */ /* Author: Korin Richmond (korin@cstr.ed.ac.uk) */ /*************************************************************************/ #include "EST_TList.h" #include "EST_TSortable.h" #include "EST_Val.h" class DiphoneVoiceModule; Declare_TList_T(DiphoneVoiceModule *, DiphoneVoiceModuleP) #if defined(INSTANTIATE_TEMPLATES) #include "../base_class/EST_TList.cc" #include "../base_class/EST_TSortable.cc" Instantiate_TList_T_MIN(DiphoneVoiceModule *, DiphoneVoiceModuleP) #endif festival/src/modules/MultiSyn/inst_tmpl/list_strlist_t.cc0000644000175000017500000000605210055371305022426 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 2004 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Rob Clark */ /* Date: Jan 2004 */ /* --------------------------------------------------------------------- */ /* Diphone backing off procedure */ /*************************************************************************/ #include "DiphoneBackoff.h" // Template rubish Declare_TList_T(EST_TList,STR_LIST) #if defined(INSTANTIATE_TEMPLATES) #include "../base_class/EST_TList.cc" Instantiate_TList_T(EST_StrList,STR_LIST); #endif festival/src/modules/MultiSyn/inst_tmpl/vector_jccp_t.cc0000644000175000017500000000676510055371305022203 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2004 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Date: February 2004 */ /* -------------------------------------------------------------------- */ /* Instantiate class list of pointers to EST_Utterance */ /* */ /* Author: Korin Richmond (korin@cstr.ed.ac.uk) */ /*************************************************************************/ #include "EST_TVector.h" #include "EST_TSimpleVector.h" class EST_JoinCostCache; Declare_TVector_Base_T(EST_JoinCostCache*,0,0,EST_JoinCostCacheP) Declare_TSimpleVector_T(EST_JoinCostCache*,EST_JoinCostCacheP) #if defined(INSTANTIATE_TEMPLATES) #include "../base_class/EST_TSimpleVector.cc" #include "../base_class/EST_TVector.cc" Instantiate_TVector_T_MIN(EST_JoinCostCache*,EST_JoinCostCacheP) Instantiate_TSimpleVector(EST_JoinCostCache*) #endif festival/src/modules/MultiSyn/inst_tmpl/list_scorepair_t.cc0000644000175000017500000000632110452235126022711 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* (University of Edinburgh, UK) and */ /* Korin Richmond */ /* Copyright (c) 2004 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* */ /* Author: Korin Richmond */ /* Date: June 2006 */ /* --------------------------------------------------------------------- */ /* Template instantiations for target cost rescoring internal work class */ /*************************************************************************/ #include "TargetCostRescoring.h" // manual template instantiation hassle Declare_TList(ScorePair) Declare_TSortable(ScorePair) #if defined(INSTANTIATE_TEMPLATES) #include "../base_class/EST_TList.cc" #include "../base_class/EST_TSortable.cc" Instantiate_TList(ScorePair); Instantiate_TSortable(ScorePair); #endif festival/src/modules/MultiSyn/inst_tmpl/Makefile0000644000175000017500000000644110452241416020477 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## (University of Edinburgh, UK) and ## ## Korin Richmond ## ## Copyright (c) 2002 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### # Makefile for base class instantiations # # # # Author: Korin Richmond (korin@cstr.ed.ac.uk) Aug 2002 # ########################################################################### TOP=../../../.. DIRNAME=src/modules/MultiSyn/inst_tmpl LOCAL_INCLUDES = -I$(TOP)/src/modules/MultiSyn TSRCS = hash_s_itemlistp_t.cc \ hash_itemp_tcdatap_t.cc \ list_uttp_t.cc \ list_itemp_t.cc \ list_voicemodulep_t.cc \ list_strlist_t.cc \ vector_jccp_t.cc \ list_scorepair_t.cc CPPSRCS = $(TSRCS) SRCS = $(CPPSRCS) OBJS = $(CPPSRCS:.cc=.o) FILES = $(SRCS) Makefile ALL = .buildlib $(BUILD_DIRS) include $(TOP)/config/common_make_rules festival/src/modules/hts_engine/Makefile0000644000175000017500000000670611576317213017040 0ustar awbawb# ----------------------------------------------------------------- # # The HMM-Based Speech Synthesis System (HTS) # # festopt_hts_engine developed by HTS Working Group # # http://hts-engine.sourceforge.net/ # # ----------------------------------------------------------------- # # # # Copyright (c) 2001-2011 Nagoya Institute of Technology # # Department of Computer Science # # # # 2001-2008 Tokyo Institute of Technology # # Interdisciplinary Graduate School of # # Science and Engineering # # # # All rights reserved. # # # # Redistribution and use in source and binary forms, with or # # without modification, are permitted provided that the following # # conditions are met: # # # # - Redistributions of source code must retain the above copyright # # notice, this list of conditions and the following disclaimer. # # - Redistributions in binary form must reproduce the above # # copyright notice, this list of conditions and the following # # disclaimer in the documentation and/or other materials provided # # with the distribution. # # - Neither the name of the HTS working group nor the names of its # # contributors may be used to endorse or promote products derived # # from this software without specific prior written permission. # # # # THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND # # CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, # # INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF # # MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE # # DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS # # BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, # # EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED # # TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, # # DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON # # ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, # # OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY # # OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE # # POSSIBILITY OF SUCH DAMAGE. # # ----------------------------------------------------------------- # TOP=../../.. DIRNAME=src/modules/hts_engine H = HTS_engine.h HTS_hidden.h CPPSRCS = fest2hts_engine.cc CSRCS = HTS_audio.c HTS_engine.c HTS_gstream.c HTS_label.c HTS_misc.c HTS_model.c HTS_pstream.c HTS_sstream.c HTS_vocoder.c SRCS = $(CPPSRCS) $(CSRCS) OBJS = $(CPPSRCS:.cc=.o) $(CSRCS:.c=.o) INFOS = README COPYING INSTALL AUTHORS FILES=Makefile $(SRCS) $(H) $(INFOS) LOCAL_INCLUDES = -I ../include -D FESTIVAL INLIB = $(TOP)/src/lib/libFestival.a ALL = .buildlib include $(TOP)/config/common_make_rules festival/src/modules/hts_engine/fest2hts_engine.cc0000644000175000017500000001756512110442174020771 0ustar awbawb/* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Module for Festival */ /* "festopt_hts_engine" developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2013 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ /* standard C libraries */ #include #include #include #include #include "festival.h" /* header of hts_engine API */ #include "HTS_engine.h" #include "HTS_hidden.h" /* Getfp: wrapper for fopen */ static FILE *Getfp(const char *name, const char *opt) { FILE *fp = fopen(name, opt); if (fp == NULL) { cerr << "Getfp: Cannot open " << name << endl; festival_error(); } return (fp); } /* HTS_Synthesize_Utt: generate speech from utt by using hts_engine API */ static LISP HTS_Synthesize_Utt(LISP utt) { EST_Utterance *u = get_c_utt(utt); EST_Item *item = 0; LISP hts_engine_params = NIL; LISP hts_output_params = NIL; char **fn_voices = NULL; char *fn_htsvoice = NULL; HTS_Engine engine; FILE *labfp = NULL; FILE *rawfp = NULL, *durfp = NULL; int sampling_rate; double speech_speed; /* get params */ hts_engine_params = siod_get_lval("hts_engine_params", "festopt_hts_engine: no parameters set for module"); hts_output_params = siod_get_lval("hts_output_params", "festopt_hts_engine: no output parameters set for module"); /* get model file names */ fn_voices = (char **) malloc(sizeof(char *)); fn_htsvoice = (char *) get_param_str("-m", hts_engine_params, "hts/htsvoice"); fn_voices[0] = fn_htsvoice; /* get speech speed */ speech_speed = (double) get_param_float("-r", hts_engine_params, 1.0); /* open input file pointers */ labfp = Getfp(get_param_str("-labelfile", hts_output_params, "utt.feats"), "r"); /* open output file pointers */ rawfp = Getfp(get_param_str("-or", hts_output_params, "tmp.raw"), "wb"); durfp = Getfp(get_param_str("-od", hts_output_params, "tmp.lab"), "wb"); /* initialize */ HTS_Engine_initialize(&engine); /* load models */ HTS_Engine_load(&engine, fn_voices, 1); free(fn_voices); /* turn off audio output */ HTS_Engine_set_audio_buff_size(&engine, 0); /* get sampling frequency */ sampling_rate = HTS_Engine_get_sampling_frequency(&engine); /* set speech speed */ HTS_Engine_set_speed(&engine, speech_speed); /* generate speech */ if (u->relation("Segment")->first()) { /* only if there segments */ HTS_Engine_synthesize_from_fn(&engine, get_param_str("-labelfile", hts_output_params, "utt.feats")); if (rawfp != NULL) HTS_Engine_save_generated_speech(&engine, rawfp); if (durfp != NULL) HTS_Engine_save_label(&engine, durfp); HTS_Engine_refresh(&engine); } /* free */ HTS_Engine_clear(&engine); /* close output file pointers */ if (rawfp != NULL) fclose(rawfp); if (durfp != NULL) fclose(durfp); /* close input file pointers */ if (labfp != NULL) fclose(labfp); /* Load back in the waveform */ EST_Wave *w = new EST_Wave; w->resample(sampling_rate); if (u->relation("Segment")->first()) /* only if there segments */ w->load_file(get_param_str("-or", hts_output_params, "tmp.raw"), "raw", sampling_rate, "short", str_to_bo("native"), 1); item = u->create_relation("Wave")->append(); item->set_val("wave", est_val(w)); /* Load back in the segment times */ EST_Relation *r = new EST_Relation; EST_Item *s, *o; r->load(get_param_str("-od", hts_output_params, "tmp.lab"), "htk"); for (o = r->first(), s = u->relation("Segment")->first(); (o != NULL) && (s != NULL); o = o->next(), s = s->next()) if (o->S("name").before("+").after("-").matches(s->S("name"))) s->set("end", o->F("end")); else cerr << "HTS_Synthesize_Utt: Output segment mismatch"; delete r; return utt; } #define HTS_VERSION "1.07" #define HTS_URL "http://hts-engine.sourceforge.net/" #define HTS_NCOPYRIGHT 1 /* HTS_get_copyright: write copyright to string */ void HTS_get_copyright(char *str) { int i, nCopyright = HTS_NCOPYRIGHT; char url[] = HTS_URL, version[] = HTS_VERSION; char *copyright[] = { HTS_COPYRIGHT }; sprintf(str, "\nThe HMM-Based Speech Synthesis Engine \"hts_engine API\"\n"); sprintf(str, "%shts_engine API version %s (%s)\n", str, version, url); for (i = 0; i < nCopyright; i++) { if (i == 0) sprintf(str, "%sCopyright (C) %s\n", str, copyright[i]); else sprintf(str, "%s %s\n", str, copyright[i]); } sprintf(str, "%sAll rights reserved.\n", str); return; } void festival_hts_engine_init(void) { char buf[4096]; HTS_get_copyright(buf); proclaim_module("hts_engine", buf); festival_def_utt_module("HTS_Synthesize", HTS_Synthesize_Utt, "(HTS_Synthesis UTT)\n Synthesize a waveform using the hts_engine and the current models"); } festival/src/modules/hts_engine/HTS_audio.c0000644000175000017500000004465612073131322017355 0ustar awbawb/* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ #ifndef HTS_AUDIO_C #define HTS_AUDIO_C #ifdef __cplusplus #define HTS_AUDIO_C_START extern "C" { #define HTS_AUDIO_C_END } #else #define HTS_AUDIO_C_START #define HTS_AUDIO_C_END #endif /* __CPLUSPLUS */ HTS_AUDIO_C_START; #if !defined(AUDIO_PLAY_WIN32) && !defined(AUDIO_PLAY_PORTAUDIO) && !defined(AUDIO_PLAY_NONE) #if defined(__WINCE__) || defined(_WINCE) || defined(_WINCE) || defined(__WINCE) || defined(__WIN32__) || defined(__WIN32) || defined(_WIN32) || defined(WIN32) || defined(__CYGWIN__) || defined(__MINGW32__) #define AUDIO_PLAY_WIN32 #else #define AUDIO_PLAY_NONE #endif /* __WINCE__ || _WINCE || _WINCE || __WINCE || __WIN32__ || __WIN32 || _WIN32 || WIN32 || __CYGWIN__ || __MINGW32__ */ #endif /* !AUDIO_PLAY_WIN32 && !AUDIO_PLAY_PORTAUDIO && !AUDIO_PLAY_NONE */ /* hts_engine libralies */ #include "HTS_hidden.h" #ifdef AUDIO_PLAY_WIN32 #include #include #define AUDIO_WAIT_BUFF_MS 10 /* wait time (0.01 sec) */ #define AUDIO_CHANNEL 1 /* monoral */ /* HTS_Audio: audio interface for Windows */ typedef struct _HTS_AudioInterface { HWAVEOUT hwaveout; /* audio device handle */ WAVEFORMATEX waveformatex; /* wave formatex */ unsigned char which_buff; /* double buffering flag */ HTS_Boolean now_buff_1; /* double buffering flag */ HTS_Boolean now_buff_2; /* double buffering flag */ WAVEHDR buff_1; /* buffer */ WAVEHDR buff_2; /* buffer */ } HTS_AudioInterface; /* HTS_AudioInterface_callback_function: callback function from audio device */ static void CALLBACK HTS_AudioInterface_callback_function(HWAVEOUT hwaveout, UINT msg, DWORD user_data, DWORD param1, DWORD param2) { WAVEHDR *wavehdr = (WAVEHDR *) param1; HTS_AudioInterface *audio_interface = (HTS_AudioInterface *) user_data; if (msg == MM_WOM_DONE && wavehdr && (wavehdr->dwFlags & WHDR_DONE)) { if (audio_interface->now_buff_1 == TRUE && wavehdr == &(audio_interface->buff_1)) { audio_interface->now_buff_1 = FALSE; } else if (audio_interface->now_buff_2 == TRUE && wavehdr == &(audio_interface->buff_2)) { audio_interface->now_buff_2 = FALSE; } } } /* HTS_AudioInterface_write: send buffer to audio device */ static void HTS_AudioInterface_write(HTS_AudioInterface * audio_interface, const short *buff, size_t buff_size) { MMRESULT result; if (audio_interface->which_buff == 1) { while (audio_interface->now_buff_1 == TRUE) Sleep(AUDIO_WAIT_BUFF_MS); audio_interface->now_buff_1 = TRUE; audio_interface->which_buff = 2; memcpy(audio_interface->buff_1.lpData, buff, buff_size * sizeof(short)); audio_interface->buff_1.dwBufferLength = buff_size * sizeof(short); result = waveOutWrite(audio_interface->hwaveout, &(audio_interface->buff_1), sizeof(WAVEHDR)); } else { while (audio_interface->now_buff_2 == TRUE) Sleep(AUDIO_WAIT_BUFF_MS); audio_interface->now_buff_2 = TRUE; audio_interface->which_buff = 1; memcpy(audio_interface->buff_2.lpData, buff, buff_size * sizeof(short)); audio_interface->buff_2.dwBufferLength = buff_size * sizeof(short); result = waveOutWrite(audio_interface->hwaveout, &(audio_interface->buff_2), sizeof(WAVEHDR)); } if (result != MMSYSERR_NOERROR) HTS_error(0, "hts_engine: Cannot send datablocks to your output audio device to play waveform.\n"); } /* HTS_AudioInterface_close: close audio device */ static void HTS_AudioInterface_close(HTS_AudioInterface * audio_interface) { MMRESULT result; /* stop audio */ result = waveOutReset(audio_interface->hwaveout); if (result != MMSYSERR_NOERROR) HTS_error(0, "hts_engine: Cannot stop and reset your output audio device.\n"); /* unprepare */ result = waveOutUnprepareHeader(audio_interface->hwaveout, &(audio_interface->buff_1), sizeof(WAVEHDR)); if (result != MMSYSERR_NOERROR) HTS_error(0, "hts_engine: Cannot cleanup the audio datablocks to play waveform.\n"); result = waveOutUnprepareHeader(audio_interface->hwaveout, &(audio_interface->buff_2), sizeof(WAVEHDR)); if (result != MMSYSERR_NOERROR) HTS_error(0, "hts_engine: Cannot cleanup the audio datablocks to play waveform.\n"); /* close */ result = waveOutClose(audio_interface->hwaveout); if (result != MMSYSERR_NOERROR) HTS_error(0, "hts_engine: Failed to close your output audio device.\n"); if (audio_interface->buff_1.lpData != NULL) HTS_free(audio_interface->buff_1.lpData); if (audio_interface->buff_2.lpData != NULL) HTS_free(audio_interface->buff_2.lpData); HTS_free(audio_interface); } static HTS_AudioInterface *HTS_AudioInterface_open(size_t sampling_frequency, size_t max_buff_size) { HTS_AudioInterface *audio_interface; MMRESULT result; /* make audio interface */ audio_interface = (HTS_AudioInterface *) HTS_calloc(1, sizeof(HTS_AudioInterface)); audio_interface->hwaveout = 0; audio_interface->which_buff = 1; audio_interface->now_buff_1 = FALSE; audio_interface->now_buff_2 = FALSE; /* format */ audio_interface->waveformatex.wFormatTag = WAVE_FORMAT_PCM; audio_interface->waveformatex.nChannels = AUDIO_CHANNEL; audio_interface->waveformatex.nSamplesPerSec = sampling_frequency; audio_interface->waveformatex.wBitsPerSample = sizeof(short) * 8; audio_interface->waveformatex.nBlockAlign = AUDIO_CHANNEL * audio_interface->waveformatex.wBitsPerSample / 8; audio_interface->waveformatex.nAvgBytesPerSec = sampling_frequency * audio_interface->waveformatex.nBlockAlign; /* open */ result = waveOutOpen(&audio_interface->hwaveout, WAVE_MAPPER, &audio_interface->waveformatex, (DWORD) HTS_AudioInterface_callback_function, (DWORD) audio_interface, CALLBACK_FUNCTION); if (result != MMSYSERR_NOERROR) { HTS_error(0, "hts_engine: Failed to open your output audio_interface device to play waveform.\n"); HTS_free(audio_interface); return NULL; } /* prepare */ audio_interface->buff_1.lpData = (LPSTR) HTS_calloc(max_buff_size, sizeof(short)); audio_interface->buff_1.dwBufferLength = max_buff_size * sizeof(short); audio_interface->buff_1.dwFlags = WHDR_BEGINLOOP | WHDR_ENDLOOP; audio_interface->buff_1.dwLoops = 1; audio_interface->buff_1.lpNext = 0; audio_interface->buff_1.reserved = 0; result = waveOutPrepareHeader(audio_interface->hwaveout, &(audio_interface->buff_1), sizeof(WAVEHDR)); if (result != MMSYSERR_NOERROR) { HTS_error(0, "hts_engine: Cannot initialize audio_interface datablocks to play waveform.\n"); HTS_free(audio_interface->buff_1.lpData); HTS_free(audio_interface); return NULL; } audio_interface->buff_2.lpData = (LPSTR) HTS_calloc(max_buff_size, sizeof(short)); audio_interface->buff_2.dwBufferLength = max_buff_size * sizeof(short); audio_interface->buff_2.dwFlags = WHDR_BEGINLOOP | WHDR_ENDLOOP; audio_interface->buff_2.dwLoops = 1; audio_interface->buff_2.lpNext = 0; audio_interface->buff_2.reserved = 0; result = waveOutPrepareHeader(audio_interface->hwaveout, &(audio_interface->buff_2), sizeof(WAVEHDR)); if (result != MMSYSERR_NOERROR) { HTS_error(0, "hts_engine: Cannot initialize audio_interface datablocks to play waveform.\n"); HTS_free(audio_interface->buff_1.lpData); HTS_free(audio_interface->buff_2.lpData); HTS_free(audio_interface); return NULL; } return audio_interface; } /* HTS_Audio_initialize: initialize audio */ void HTS_Audio_initialize(HTS_Audio * audio) { if (audio == NULL) return; audio->sampling_frequency = 0; audio->max_buff_size = 0; audio->buff = NULL; audio->buff_size = 0; audio->audio_interface = NULL; } /* HTS_Audio_set_parameter: set parameters for audio */ void HTS_Audio_set_parameter(HTS_Audio * audio, size_t sampling_frequency, size_t max_buff_size) { if (audio == NULL) return; if (audio->sampling_frequency == sampling_frequency && audio->max_buff_size == max_buff_size) return; HTS_Audio_clear(audio); if (sampling_frequency == 0 || max_buff_size == 0) return; audio->audio_interface = HTS_AudioInterface_open(sampling_frequency, max_buff_size); if (audio->audio_interface == NULL) return; audio->sampling_frequency = sampling_frequency; audio->max_buff_size = max_buff_size; audio->buff = (short *) HTS_calloc(max_buff_size, sizeof(short)); audio->buff_size = 0; } /* HTS_Audio_write: send data to audio */ void HTS_Audio_write(HTS_Audio * audio, short data) { if (audio == NULL) return; audio->buff[audio->buff_size++] = data; if (audio->buff_size >= audio->max_buff_size) { if (audio->audio_interface != NULL) HTS_AudioInterface_write((HTS_AudioInterface *) audio->audio_interface, audio->buff, audio->buff_size); audio->buff_size = 0; } } /* HTS_Audio_flush: flush remain data */ void HTS_Audio_flush(HTS_Audio * audio) { HTS_AudioInterface *audio_interface; if (audio == NULL || audio->audio_interface == NULL) return; audio_interface = (HTS_AudioInterface *) audio->audio_interface; if (audio->buff_size > 0) { HTS_AudioInterface_write(audio_interface, audio->buff, audio->buff_size); audio->buff_size = 0; } while (audio_interface->now_buff_1 == TRUE || audio_interface->now_buff_2 == TRUE) Sleep(AUDIO_WAIT_BUFF_MS); } /* HTS_Audio_clear: free audio */ void HTS_Audio_clear(HTS_Audio * audio) { HTS_AudioInterface *audio_interface; if (audio == NULL || audio->audio_interface == NULL) return; audio_interface = (HTS_AudioInterface *) audio->audio_interface; HTS_Audio_flush(audio); HTS_AudioInterface_close(audio_interface); if (audio->buff != NULL) free(audio->buff); HTS_Audio_initialize(audio); } #endif /* AUDIO_PLAY_WIN32 */ #ifdef AUDIO_PLAY_PORTAUDIO #include "portaudio.h" /* HTS_AudioInterface: audio output for PortAudio */ typedef struct _HTS_AudioInterface { PaStreamParameters parameters; /* parameters for output stream */ PaStream *stream; /* output stream */ } HTS_AudioInterface; /* HTS_AudioInterface_write: send data to audio device */ static void HTS_AudioInterface_write(HTS_AudioInterface * audio_interface, const short *buff, size_t buff_size) { PaError err; err = Pa_WriteStream(audio_interface->stream, buff, buff_size); if (err != paNoError && err != paOutputUnderflowed) HTS_error(0, "hts_engine: Cannot send datablocks to your output audio device to play waveform.\n"); } /* HTS_AudioInterface_close: close audio device */ static void HTS_AudioInterface_close(HTS_AudioInterface * audio_interface) { PaError err; err = Pa_StopStream(audio_interface->stream); if (err != paNoError) HTS_error(0, "hts_engine: Cannot stop your output audio device.\n"); err = Pa_CloseStream(audio_interface->stream); if (err != paNoError) HTS_error(0, "hts_engine: Failed to close your output audio device.\n"); Pa_Terminate(); HTS_free(audio_interface); } static HTS_AudioInterface *HTS_AudioInterface_open(size_t sampling_frequency, size_t max_buff_size) { HTS_AudioInterface *audio_interface; PaError err; audio_interface = HTS_calloc(1, sizeof(HTS_AudioInterface)); audio_interface->stream = NULL; err = Pa_Initialize(); if (err != paNoError) { HTS_error(0, "hts_engine: Failed to initialize your output audio device to play waveform.\n"); HTS_free(audio_interface); return NULL; } audio_interface->parameters.device = Pa_GetDefaultOutputDevice(); audio_interface->parameters.channelCount = 1; audio_interface->parameters.sampleFormat = paInt16; audio_interface->parameters.suggestedLatency = Pa_GetDeviceInfo(audio_interface->parameters.device)->defaultLowOutputLatency; audio_interface->parameters.hostApiSpecificStreamInfo = NULL; err = Pa_OpenStream(&audio_interface->stream, NULL, &audio_interface->parameters, sampling_frequency, max_buff_size, paClipOff, NULL, NULL); if (err != paNoError) { HTS_error(0, "hts_engine: Failed to open your output audio device to play waveform.\n"); Pa_Terminate(); HTS_free(audio_interface); return NULL; } err = Pa_StartStream(audio_interface->stream); if (err != paNoError) { HTS_error(0, "hts_engine: Failed to start your output audio device to play waveform.\n"); Pa_CloseStream(audio_interface->stream); Pa_Terminate(); HTS_free(audio_interface); return NULL; } return audio_interface; } /* HTS_Audio_initialize: initialize audio */ void HTS_Audio_initialize(HTS_Audio * audio) { if (audio == NULL) return; audio->sampling_frequency = 0; audio->max_buff_size = 0; audio->buff = NULL; audio->buff_size = 0; audio->audio_interface = NULL; } /* HTS_Audio_set_parameter: set parameters for audio */ void HTS_Audio_set_parameter(HTS_Audio * audio, size_t sampling_frequency, size_t max_buff_size) { if (audio == NULL) return; if (audio->sampling_frequency == sampling_frequency && audio->max_buff_size == max_buff_size) return; HTS_Audio_clear(audio); if (sampling_frequency == 0 || max_buff_size == 0) return; audio->audio_interface = HTS_AudioInterface_open(sampling_frequency, max_buff_size); if (audio->audio_interface == NULL) return; audio->sampling_frequency = sampling_frequency; audio->max_buff_size = max_buff_size; audio->buff = (short *) HTS_calloc(max_buff_size, sizeof(short)); audio->buff_size = 0; } /* HTS_Audio_write: send data to audio device */ void HTS_Audio_write(HTS_Audio * audio, short data) { if (audio == NULL) return; audio->buff[audio->buff_size++] = data; if (audio->buff_size >= audio->max_buff_size) { if (audio->audio_interface != NULL) HTS_AudioInterface_write((HTS_AudioInterface *) audio->audio_interface, audio->buff, audio->max_buff_size); audio->buff_size = 0; } } /* HTS_Audio_flush: flush remain data */ void HTS_Audio_flush(HTS_Audio * audio) { HTS_AudioInterface *audio_interface; if (audio == NULL || audio->audio_interface == NULL) return; audio_interface = (HTS_AudioInterface *) audio->audio_interface; if (audio->buff_size > 0) { HTS_AudioInterface_write((HTS_AudioInterface *) audio->audio_interface, audio->buff, audio->buff_size); audio->buff_size = 0; } } /* HTS_Audio_clear: free audio */ void HTS_Audio_clear(HTS_Audio * audio) { HTS_AudioInterface *audio_interface; if (audio == NULL || audio->audio_interface == NULL) return; audio_interface = (HTS_AudioInterface *) audio->audio_interface; HTS_Audio_flush(audio); HTS_AudioInterface_close(audio_interface); if (audio->buff != NULL) HTS_free(audio->buff); HTS_Audio_initialize(audio); } #endif /* AUDIO_PLAY_PORTAUDIO */ #ifdef AUDIO_PLAY_NONE /* HTS_Audio_initialize: initialize audio */ void HTS_Audio_initialize(HTS_Audio * audio) { } /* HTS_Audio_set_parameter: set parameters for audio */ void HTS_Audio_set_parameter(HTS_Audio * audio, size_t sampling_frequeny, size_t max_buff_size) { } /* HTS_Audio_write: send data to audio */ void HTS_Audio_write(HTS_Audio * audio, short data) { } /* HTS_Audio_flush: flush remain data */ void HTS_Audio_flush(HTS_Audio * audio) { } /* HTS_Audio_clear: free audio */ void HTS_Audio_clear(HTS_Audio * audio) { } #endif /* AUDIO_PLAY_NONE */ HTS_AUDIO_C_END; #endif /* !HTS_AUDIO_C */ festival/src/modules/hts_engine/HTS_engine.c0000644000175000017500000007223312110442173017512 0ustar awbawb/* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ #ifndef HTS_ENGINE_C #define HTS_ENGINE_C #ifdef __cplusplus #define HTS_ENGINE_C_START extern "C" { #define HTS_ENGINE_C_END } #else #define HTS_ENGINE_C_START #define HTS_ENGINE_C_END #endif /* __CPLUSPLUS */ HTS_ENGINE_C_START; #include /* for atof() */ #include /* for strcpy() */ #include /* for pow() */ /* hts_engine libraries */ #include "HTS_hidden.h" /* HTS_Engine_initialize: initialize engine */ void HTS_Engine_initialize(HTS_Engine * engine) { /* global */ engine->condition.sampling_frequency = 0; engine->condition.fperiod = 0; engine->condition.audio_buff_size = 0; engine->condition.stop = FALSE; engine->condition.volume = 1.0; engine->condition.msd_threshold = NULL; engine->condition.gv_weight = NULL; /* duration */ engine->condition.speed = 1.0; engine->condition.phoneme_alignment_flag = FALSE; /* spectrum */ engine->condition.stage = 0; engine->condition.use_log_gain = FALSE; engine->condition.alpha = 0.0; engine->condition.beta = 0.0; /* log F0 */ engine->condition.additional_half_tone = 0.0; /* interpolation weights */ engine->condition.duration_iw = NULL; engine->condition.parameter_iw = NULL; engine->condition.gv_iw = NULL; /* initialize audio */ HTS_Audio_initialize(&engine->audio); /* initialize model set */ HTS_ModelSet_initialize(&engine->ms); /* initialize label list */ HTS_Label_initialize(&engine->label); /* initialize state sequence set */ HTS_SStreamSet_initialize(&engine->sss); /* initialize pstream set */ HTS_PStreamSet_initialize(&engine->pss); /* initialize gstream set */ HTS_GStreamSet_initialize(&engine->gss); } /* HTS_Engine_load: load HTS voices */ HTS_Boolean HTS_Engine_load(HTS_Engine * engine, char **voices, size_t num_voices) { size_t i, j; size_t nstream; double average_weight; const char *option, *find; /* reset engine */ HTS_Engine_clear(engine); /* load voices */ if (HTS_ModelSet_load(&engine->ms, voices, num_voices) != TRUE) { HTS_Engine_clear(engine); return FALSE; } nstream = HTS_ModelSet_get_nstream(&engine->ms); average_weight = 1.0 / num_voices; /* global */ engine->condition.sampling_frequency = HTS_ModelSet_get_sampling_frequency(&engine->ms); engine->condition.fperiod = HTS_ModelSet_get_fperiod(&engine->ms); engine->condition.msd_threshold = (double *) HTS_calloc(nstream, sizeof(double)); for (i = 0; i < nstream; i++) engine->condition.msd_threshold[i] = 0.5; engine->condition.gv_weight = (double *) HTS_calloc(nstream, sizeof(double)); for (i = 0; i < nstream; i++) engine->condition.gv_weight[i] = 1.0; /* spectrum */ option = HTS_ModelSet_get_option(&engine->ms, 0); find = strstr(option, "GAMMA="); if (find != NULL) engine->condition.stage = (size_t) atoi(&find[strlen("GAMMA=")]); find = strstr(option, "LN_GAIN="); if (find != NULL) engine->condition.use_log_gain = atoi(&find[strlen("LN_GAIN=")]) == 1 ? TRUE : FALSE; find = strstr(option, "ALPHA="); if (find != NULL) engine->condition.alpha = atof(&find[strlen("ALPHA=")]); /* interpolation weights */ engine->condition.duration_iw = (double *) HTS_calloc(num_voices, sizeof(double)); for (i = 0; i < num_voices; i++) engine->condition.duration_iw[i] = average_weight; engine->condition.parameter_iw = (double **) HTS_calloc(nstream, sizeof(double *)); for (i = 0; i < nstream; i++) { engine->condition.parameter_iw[i] = (double *) HTS_calloc(num_voices, sizeof(double)); for (j = 0; j < num_voices; j++) engine->condition.parameter_iw[i][j] = average_weight; } engine->condition.gv_iw = (double **) HTS_calloc(nstream, sizeof(double *)); for (i = 0; i < nstream; i++) { engine->condition.gv_iw[i] = (double *) HTS_calloc(num_voices, sizeof(double)); for (j = 0; j < num_voices; j++) engine->condition.gv_iw[i][j] = average_weight; } return TRUE; } /* HTS_Engine_set_sampling_frequency: set sampling frequency */ void HTS_Engine_set_sampling_frequency(HTS_Engine * engine, size_t i) { if (i < 1) i = 1; engine->condition.sampling_frequency = i; HTS_Audio_set_parameter(&engine->audio, engine->condition.sampling_frequency, engine->condition.audio_buff_size); } /* HTS_Engine_get_sampling_frequency: get sampling frequency */ size_t HTS_Engine_get_sampling_frequency(HTS_Engine * engine) { return engine->condition.sampling_frequency; } /* HTS_Engine_set_fperiod: set frame period */ void HTS_Engine_set_fperiod(HTS_Engine * engine, size_t i) { if (i < 1) i = 1; engine->condition.fperiod = i; } /* HTS_Engine_get_fperiod: get frame period */ size_t HTS_Engine_get_fperiod(HTS_Engine * engine) { return engine->condition.fperiod; } /* HTS_Engine_set_audio_buff_size: set audio buffer size */ void HTS_Engine_set_audio_buff_size(HTS_Engine * engine, size_t i) { engine->condition.audio_buff_size = i; HTS_Audio_set_parameter(&engine->audio, engine->condition.sampling_frequency, engine->condition.audio_buff_size); } /* HTS_Engine_get_audio_buff_size: get audio buffer size */ size_t HTS_Engine_get_audio_buff_size(HTS_Engine * engine) { return engine->condition.audio_buff_size; } /* HTS_Engine_set_stop_flag: set stop flag */ void HTS_Engine_set_stop_flag(HTS_Engine * engine, HTS_Boolean b) { engine->condition.stop = b; } /* HTS_Engine_get_stop_flag: get stop flag */ HTS_Boolean HTS_Engine_get_stop_flag(HTS_Engine * engine) { return engine->condition.stop; } /* HTS_Engine_set_volume: set volume in db */ void HTS_Engine_set_volume(HTS_Engine * engine, double f) { engine->condition.volume = exp(f * DB); } /* HTS_Engine_get_volume: get volume in db */ double HTS_Engine_get_volume(HTS_Engine * engine) { return log(engine->condition.volume) / DB; } /* HTS_Egnine_set_msd_threshold: set MSD threshold */ void HTS_Engine_set_msd_threshold(HTS_Engine * engine, size_t stream_index, double f) { if (f < 0.0) f = 0.0; if (f > 1.0) f = 1.0; engine->condition.msd_threshold[stream_index] = f; } /* HTS_Engine_get_msd_threshold: get MSD threshold */ double HTS_Engine_get_msd_threshold(HTS_Engine * engine, size_t stream_index) { return engine->condition.msd_threshold[stream_index]; } /* HTS_Engine_set_gv_weight: set GV weight */ void HTS_Engine_set_gv_weight(HTS_Engine * engine, size_t stream_index, double f) { if (f < 0.0) f = 0.0; engine->condition.gv_weight[stream_index] = f; } /* HTS_Engine_get_gv_weight: get GV weight */ double HTS_Engine_get_gv_weight(HTS_Engine * engine, size_t stream_index) { return engine->condition.gv_weight[stream_index]; } /* HTS_Engine_set_speed: set speech speed */ void HTS_Engine_set_speed(HTS_Engine * engine, double f) { if (f < 1.0E-06) f = 1.0E-06; engine->condition.speed = f; } /* HTS_Engine_set_phoneme_alignment_flag: set flag for using phoneme alignment in label */ void HTS_Engine_set_phoneme_alignment_flag(HTS_Engine * engine, HTS_Boolean b) { engine->condition.phoneme_alignment_flag = b; } /* HTS_Engine_set_alpha: set alpha */ void HTS_Engine_set_alpha(HTS_Engine * engine, double f) { if (f < 0.0) f = 0.0; if (f > 1.0) f = 1.0; engine->condition.alpha = f; } /* HTS_Engine_get_alpha: get alpha */ double HTS_Engine_get_alpha(HTS_Engine * engine) { return engine->condition.alpha; } /* HTS_Engine_set_beta: set beta */ void HTS_Engine_set_beta(HTS_Engine * engine, double f) { if (f < 0.0) f = 0.0; if (f > 1.0) f = 1.0; engine->condition.beta = f; } /* HTS_Engine_get_beta: get beta */ double HTS_Engine_get_beta(HTS_Engine * engine) { return engine->condition.beta; } /* HTS_Engine_add_half_tone: add half tone */ void HTS_Engine_add_half_tone(HTS_Engine * engine, double f) { engine->condition.additional_half_tone = f; } /* HTS_Engine_set_duration_interpolation_weight: set interpolation weight for duration */ void HTS_Engine_set_duration_interpolation_weight(HTS_Engine * engine, size_t voice_index, double f) { engine->condition.duration_iw[voice_index] = f; } /* HTS_Engine_get_duration_interpolation_weight: get interpolation weight for duration */ double HTS_Engine_get_duration_interpolation_weight(HTS_Engine * engine, size_t voice_index) { return engine->condition.duration_iw[voice_index]; } /* HTS_Engine_set_parameter_interpolation_weight: set interpolation weight for parameter */ void HTS_Engine_set_parameter_interpolation_weight(HTS_Engine * engine, size_t voice_index, size_t stream_index, double f) { engine->condition.parameter_iw[voice_index][stream_index] = f; } /* HTS_Engine_get_parameter_interpolation_weight: get interpolation weight for parameter */ double HTS_Engine_get_parameter_interpolation_weight(HTS_Engine * engine, size_t voice_index, size_t stream_index) { return engine->condition.parameter_iw[voice_index][stream_index]; } /* HTS_Engine_set_gv_interpolation_weight: set interpolation weight for GV */ void HTS_Engine_set_gv_interpolation_weight(HTS_Engine * engine, size_t voice_index, size_t stream_index, double f) { engine->condition.gv_iw[voice_index][stream_index] = f; } /* HTS_Engine_get_gv_interpolation_weight: get interpolation weight for GV */ double HTS_Engine_get_gv_interpolation_weight(HTS_Engine * engine, size_t voice_index, size_t stream_index) { return engine->condition.gv_iw[voice_index][stream_index]; } /* HTS_Engine_get_total_state: get total number of state */ size_t HTS_Engine_get_total_state(HTS_Engine * engine) { return HTS_SStreamSet_get_total_state(&engine->sss); } /* HTS_Engine_set_state_mean: set mean value of state */ void HTS_Engine_set_state_mean(HTS_Engine * engine, size_t stream_index, size_t state_index, size_t vector_index, double f) { HTS_SStreamSet_set_mean(&engine->sss, stream_index, state_index, vector_index, f); } /* HTS_Engine_get_state_mean: get mean value of state */ double HTS_Engine_get_state_mean(HTS_Engine * engine, size_t stream_index, size_t state_index, size_t vector_index) { return HTS_SStreamSet_get_mean(&engine->sss, stream_index, state_index, vector_index); } /* HTS_Engine_get_state_duration: get state duration */ size_t HTS_Engine_get_state_duration(HTS_Engine * engine, size_t state_index) { return HTS_SStreamSet_get_duration(&engine->sss, state_index); } /* HTS_Engine_get_nvoices: get number of voices */ size_t HTS_Engine_get_nvoices(HTS_Engine * engine) { return HTS_ModelSet_get_nvoices(&engine->ms); } /* HTS_Engine_get_nstream: get number of stream */ size_t HTS_Engine_get_nstream(HTS_Engine * engine) { return HTS_ModelSet_get_nstream(&engine->ms); } /* HTS_Engine_get_nstate: get number of state */ size_t HTS_Engine_get_nstate(HTS_Engine * engine) { return HTS_ModelSet_get_nstate(&engine->ms); } /* HTS_Engine_get_total_frame: get total number of frame */ size_t HTS_Engine_get_total_frame(HTS_Engine * engine) { return HTS_GStreamSet_get_total_frame(&engine->gss); } /* HTS_Engine_get_nsamples: get number of samples */ size_t HTS_Engine_get_nsamples(HTS_Engine * engine) { return HTS_GStreamSet_get_total_nsamples(&engine->gss); } /* HTS_Engine_get_generated_parameter: output generated parameter */ double HTS_Engine_get_generated_parameter(HTS_Engine * engine, size_t stream_index, size_t frame_index, size_t vector_index) { return HTS_GStreamSet_get_parameter(&engine->gss, stream_index, frame_index, vector_index); } /* HTS_Engine_get_generated_speech: output generated speech */ double HTS_Engine_get_generated_speech(HTS_Engine * engine, size_t index) { return HTS_GStreamSet_get_speech(&engine->gss, index); } static HTS_Boolean HTS_Engine_synthesize(HTS_Engine * engine) { size_t i, state_index, model_index; double f; if (HTS_SStreamSet_create(&engine->sss, &engine->ms, &engine->label, engine->condition.phoneme_alignment_flag, engine->condition.speed, engine->condition.duration_iw, engine->condition.parameter_iw, engine->condition.gv_iw) != TRUE) { HTS_Engine_refresh(engine); return FALSE; } if (engine->condition.additional_half_tone != 0.0) { state_index = 0; model_index = 0; for (i = 0; i < HTS_Engine_get_total_state(engine); i++) { f = HTS_Engine_get_state_mean(engine, 1, i, 0); f += engine->condition.additional_half_tone * HALF_TONE; if (f < MIN_LF0) f = MIN_LF0; else if (f > MAX_LF0) f = MAX_LF0; HTS_Engine_set_state_mean(engine, 1, i, 0, f); state_index++; if (state_index >= HTS_Engine_get_nstate(engine)) { state_index = 0; model_index++; } } } if (HTS_PStreamSet_create(&engine->pss, &engine->sss, engine->condition.msd_threshold, engine->condition.gv_weight) != TRUE) { HTS_Engine_refresh(engine); return FALSE; } if (HTS_GStreamSet_create(&engine->gss, &engine->pss, engine->condition.stage, engine->condition.use_log_gain, engine->condition.sampling_frequency, engine->condition.fperiod, engine->condition.alpha, engine->condition.beta, &engine->condition.stop, engine->condition.volume, engine->condition.audio_buff_size > 0 ? &engine->audio : NULL) != TRUE) { HTS_Engine_refresh(engine); return FALSE; } return TRUE; } /* HTS_Engine_synthesize_from_fn: synthesize speech from file name */ HTS_Boolean HTS_Engine_synthesize_from_fn(HTS_Engine * engine, const char *fn) { HTS_Engine_refresh(engine); HTS_Label_load_from_fn(&engine->label, engine->condition.sampling_frequency, engine->condition.fperiod, fn); return HTS_Engine_synthesize(engine); } /* HTS_Engine_synthesize_from_strings: synthesize speech from strings */ HTS_Boolean HTS_Engine_synthesize_from_strings(HTS_Engine * engine, char **lines, size_t num_lines) { HTS_Engine_refresh(engine); HTS_Label_load_from_strings(&engine->label, engine->condition.sampling_frequency, engine->condition.fperiod, lines, num_lines); return HTS_Engine_synthesize(engine); } /* HTS_Engine_save_information: save trace information */ void HTS_Engine_save_information(HTS_Engine * engine, FILE * fp) { size_t i, j, k, l, m, n; double temp; HTS_Condition *condition = &engine->condition; HTS_ModelSet *ms = &engine->ms; HTS_Label *label = &engine->label; HTS_SStreamSet *sss = &engine->sss; HTS_PStreamSet *pss = &engine->pss; /* global parameter */ fprintf(fp, "[Global parameter]\n"); fprintf(fp, "Sampring frequency -> %8lu(Hz)\n", (unsigned long) condition->sampling_frequency); fprintf(fp, "Frame period -> %8lu(point)\n", (unsigned long) condition->fperiod); fprintf(fp, " %8.5f(msec)\n", 1e+3 * condition->fperiod / condition->sampling_frequency); fprintf(fp, "All-pass constant -> %8.5f\n", (float) condition->alpha); fprintf(fp, "Gamma -> %8.5f\n", (float) (condition->stage == 0 ? 0.0 : -1.0 / condition->stage)); if (condition->stage != 0) { if (condition->use_log_gain == TRUE) fprintf(fp, "Log gain flag -> TRUE\n"); else fprintf(fp, "Log gain flag -> FALSE\n"); } fprintf(fp, "Postfiltering coefficient -> %8.5f\n", (float) condition->beta); fprintf(fp, "Audio buffer size -> %8lu(sample)\n", (unsigned long) condition->audio_buff_size); fprintf(fp, "\n"); /* duration parameter */ fprintf(fp, "[Duration parameter]\n"); fprintf(fp, "Number of states -> %8lu\n", (unsigned long) HTS_ModelSet_get_nstate(ms)); fprintf(fp, " Interpolation size -> %8lu\n", (unsigned long) HTS_ModelSet_get_nvoices(ms)); /* check interpolation */ for (i = 0, temp = 0.0; i < HTS_ModelSet_get_nvoices(ms); i++) temp += condition->duration_iw[i]; for (i = 0; i < HTS_ModelSet_get_nvoices(ms); i++) if (condition->duration_iw[i] != 0.0) condition->duration_iw[i] /= temp; for (i = 0; i < HTS_ModelSet_get_nvoices(ms); i++) fprintf(fp, " Interpolation weight[%2lu] -> %8.0f(%%)\n", (unsigned long) i, (float) (100 * condition->duration_iw[i])); fprintf(fp, "\n"); fprintf(fp, "[Stream parameter]\n"); for (i = 0; i < HTS_ModelSet_get_nstream(ms); i++) { /* stream parameter */ fprintf(fp, "Stream[%2lu] vector length -> %8lu\n", (unsigned long) i, (unsigned long) HTS_ModelSet_get_vector_length(ms, i)); fprintf(fp, " Dynamic window size -> %8lu\n", (unsigned long) HTS_ModelSet_get_window_size(ms, i)); /* interpolation */ fprintf(fp, " Interpolation size -> %8lu\n", (unsigned long) HTS_ModelSet_get_nvoices(ms)); for (j = 0, temp = 0.0; j < HTS_ModelSet_get_nvoices(ms); j++) temp += condition->parameter_iw[i][j]; for (j = 0; j < HTS_ModelSet_get_nvoices(ms); j++) if (condition->parameter_iw[i][j] != 0.0) condition->parameter_iw[i][j] /= temp; for (j = 0; j < HTS_ModelSet_get_nvoices(ms); j++) fprintf(fp, " Interpolation weight[%2lu] -> %8.0f(%%)\n", (unsigned long) j, (float) (100 * condition->parameter_iw[i][j])); /* MSD */ if (HTS_ModelSet_is_msd(ms, i)) { /* for MSD */ fprintf(fp, " MSD flag -> TRUE\n"); fprintf(fp, " MSD threshold -> %8.5f\n", condition->msd_threshold[i]); } else { /* for non MSD */ fprintf(fp, " MSD flag -> FALSE\n"); } /* GV */ if (HTS_ModelSet_use_gv(ms, i)) { fprintf(fp, " GV flag -> TRUE\n"); fprintf(fp, " GV weight -> %8.0f(%%)\n", (float) (100 * condition->gv_weight[i])); fprintf(fp, " GV interpolation size -> %8lu\n", (unsigned long) HTS_ModelSet_get_nvoices(ms)); /* interpolation */ for (j = 0, temp = 0.0; j < HTS_ModelSet_get_nvoices(ms); j++) temp += condition->gv_iw[i][j]; for (j = 0; j < HTS_ModelSet_get_nvoices(ms); j++) if (condition->gv_iw[i][j] != 0.0) condition->gv_iw[i][j] /= temp; for (j = 0; j < HTS_ModelSet_get_nvoices(ms); j++) fprintf(fp, " GV interpolation weight[%2lu] -> %8.0f(%%)\n", (unsigned long) j, (float) (100 * condition->gv_iw[i][j])); } else { fprintf(fp, " GV flag -> FALSE\n"); } } fprintf(fp, "\n"); /* generated sequence */ fprintf(fp, "[Generated sequence]\n"); fprintf(fp, "Number of HMMs -> %8lu\n", (unsigned long) HTS_Label_get_size(label)); fprintf(fp, "Number of stats -> %8lu\n", (unsigned long) HTS_Label_get_size(label) * HTS_ModelSet_get_nstate(ms)); fprintf(fp, "Length of this speech -> %8.3f(sec)\n", (float) ((double) HTS_PStreamSet_get_total_frame(pss) * condition->fperiod / condition->sampling_frequency)); fprintf(fp, " -> %8lu(frames)\n", (unsigned long) HTS_PStreamSet_get_total_frame(pss) * condition->fperiod); for (i = 0; i < HTS_Label_get_size(label); i++) { fprintf(fp, "HMM[%2lu]\n", (unsigned long) i); fprintf(fp, " Name -> %s\n", HTS_Label_get_string(label, i)); fprintf(fp, " Duration\n"); for (j = 0; j < HTS_ModelSet_get_nvoices(ms); j++) { fprintf(fp, " Interpolation[%2lu]\n", (unsigned long) j); HTS_ModelSet_get_duration_index(ms, j, HTS_Label_get_string(label, i), &k, &l); fprintf(fp, " Tree index -> %8lu\n", (unsigned long) k); fprintf(fp, " PDF index -> %8lu\n", (unsigned long) l); } for (j = 0; j < HTS_ModelSet_get_nstate(ms); j++) { fprintf(fp, " State[%2lu]\n", (unsigned long) j + 2); fprintf(fp, " Length -> %8lu(frames)\n", (unsigned long) HTS_SStreamSet_get_duration(sss, i * HTS_ModelSet_get_nstate(ms) + j)); for (k = 0; k < HTS_ModelSet_get_nstream(ms); k++) { fprintf(fp, " Stream[%2lu]\n", (unsigned long) k); if (HTS_ModelSet_is_msd(ms, k)) { if (HTS_SStreamSet_get_msd(sss, k, i * HTS_ModelSet_get_nstate(ms) + j) > condition->msd_threshold[k]) fprintf(fp, " MSD flag -> TRUE\n"); else fprintf(fp, " MSD flag -> FALSE\n"); } for (l = 0; l < HTS_ModelSet_get_nvoices(ms); l++) { fprintf(fp, " Interpolation[%2lu]\n", (unsigned long) l); HTS_ModelSet_get_parameter_index(ms, l, k, j + 2, HTS_Label_get_string(label, i), &m, &n); fprintf(fp, " Tree index -> %8lu\n", (unsigned long) m); fprintf(fp, " PDF index -> %8lu\n", (unsigned long) n); } } } } } /* HTS_Engine_save_label: save label with time */ void HTS_Engine_save_label(HTS_Engine * engine, FILE * fp) { size_t i, j; size_t frame, state, duration; HTS_Label *label = &engine->label; HTS_SStreamSet *sss = &engine->sss; size_t nstate = HTS_ModelSet_get_nstate(&engine->ms); double rate = engine->condition.fperiod * 1.0e+07 / engine->condition.sampling_frequency; for (i = 0, state = 0, frame = 0; i < HTS_Label_get_size(label); i++) { for (j = 0, duration = 0; j < nstate; j++) duration += HTS_SStreamSet_get_duration(sss, state++); fprintf(fp, "%lu %lu %s\n", (unsigned long) (frame * rate), (unsigned long) ((frame + duration) * rate), HTS_Label_get_string(label, i)); frame += duration; } } /* HTS_Engine_save_generated_parameter: save generated parameter */ void HTS_Engine_save_generated_parameter(HTS_Engine * engine, size_t stream_index, FILE * fp) { size_t i, j; float temp; HTS_GStreamSet *gss = &engine->gss; for (i = 0; i < HTS_GStreamSet_get_total_frame(gss); i++) for (j = 0; j < HTS_GStreamSet_get_vector_length(gss, stream_index); j++) { temp = (float) HTS_GStreamSet_get_parameter(gss, stream_index, i, j); fwrite(&temp, sizeof(float), 1, fp); } } /* HTS_Engine_save_generated_speech: save generated speech */ void HTS_Engine_save_generated_speech(HTS_Engine * engine, FILE * fp) { size_t i; double x; short temp; HTS_GStreamSet *gss = &engine->gss; for (i = 0; i < HTS_GStreamSet_get_total_nsamples(gss); i++) { x = HTS_GStreamSet_get_speech(gss, i); if (x > 32767.0) temp = 32767; else if (x < -32768.0) temp = -32768; else temp = (short) x; fwrite(&temp, sizeof(short), 1, fp); } } /* HTS_Engine_save_riff: save RIFF format file */ void HTS_Engine_save_riff(HTS_Engine * engine, FILE * fp) { size_t i; double x; short temp; HTS_GStreamSet *gss = &engine->gss; char data_01_04[] = { 'R', 'I', 'F', 'F' }; int data_05_08 = HTS_GStreamSet_get_total_nsamples(gss) * sizeof(short) + 36; char data_09_12[] = { 'W', 'A', 'V', 'E' }; char data_13_16[] = { 'f', 'm', 't', ' ' }; int data_17_20 = 16; short data_21_22 = 1; /* PCM */ short data_23_24 = 1; /* monoral */ int data_25_28 = engine->condition.sampling_frequency; int data_29_32 = engine->condition.sampling_frequency * sizeof(short); short data_33_34 = sizeof(short); short data_35_36 = (short) (sizeof(short) * 8); char data_37_40[] = { 'd', 'a', 't', 'a' }; int data_41_44 = HTS_GStreamSet_get_total_nsamples(gss) * sizeof(short); /* write header */ HTS_fwrite_little_endian(data_01_04, sizeof(char), 4, fp); HTS_fwrite_little_endian(&data_05_08, sizeof(int), 1, fp); HTS_fwrite_little_endian(data_09_12, sizeof(char), 4, fp); HTS_fwrite_little_endian(data_13_16, sizeof(char), 4, fp); HTS_fwrite_little_endian(&data_17_20, sizeof(int), 1, fp); HTS_fwrite_little_endian(&data_21_22, sizeof(short), 1, fp); HTS_fwrite_little_endian(&data_23_24, sizeof(short), 1, fp); HTS_fwrite_little_endian(&data_25_28, sizeof(int), 1, fp); HTS_fwrite_little_endian(&data_29_32, sizeof(int), 1, fp); HTS_fwrite_little_endian(&data_33_34, sizeof(short), 1, fp); HTS_fwrite_little_endian(&data_35_36, sizeof(short), 1, fp); HTS_fwrite_little_endian(data_37_40, sizeof(char), 4, fp); HTS_fwrite_little_endian(&data_41_44, sizeof(int), 1, fp); /* write data */ for (i = 0; i < HTS_GStreamSet_get_total_nsamples(gss); i++) { x = HTS_GStreamSet_get_speech(gss, i); if (x > 32767.0) temp = 32767; else if (x < -32768.0) temp = -32768; else temp = (short) x; HTS_fwrite_little_endian(&temp, sizeof(short), 1, fp); } } /* HTS_Engine_refresh: free model per one time synthesis */ void HTS_Engine_refresh(HTS_Engine * engine) { /* free generated parameter stream set */ HTS_GStreamSet_clear(&engine->gss); /* free parameter stream set */ HTS_PStreamSet_clear(&engine->pss); /* free state stream set */ HTS_SStreamSet_clear(&engine->sss); /* free label list */ HTS_Label_clear(&engine->label); /* stop flag */ engine->condition.stop = FALSE; } /* HTS_Engine_clear: free engine */ void HTS_Engine_clear(HTS_Engine * engine) { size_t i; if (engine->condition.msd_threshold != NULL) HTS_free(engine->condition.msd_threshold); if (engine->condition.duration_iw != NULL) HTS_free(engine->condition.duration_iw); if (engine->condition.gv_weight != NULL) HTS_free(engine->condition.gv_weight); if (engine->condition.parameter_iw != NULL) { for (i = 0; i < HTS_ModelSet_get_nstream(&engine->ms); i++) HTS_free(engine->condition.parameter_iw[i]); HTS_free(engine->condition.parameter_iw); } if (engine->condition.gv_iw != NULL) { for (i = 0; i < HTS_ModelSet_get_nstream(&engine->ms); i++) HTS_free(engine->condition.gv_iw[i]); HTS_free(engine->condition.gv_iw); } HTS_ModelSet_clear(&engine->ms); HTS_Audio_clear(&engine->audio); HTS_Engine_initialize(engine); } HTS_ENGINE_C_END; #endif /* !HTS_ENGINE_C */ festival/src/modules/hts_engine/HTS_gstream.c0000644000175000017500000002062312073131324017704 0ustar awbawb/* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ #ifndef HTS_GSTREAM_C #define HTS_GSTREAM_C #ifdef __cplusplus #define HTS_GSTREAM_C_START extern "C" { #define HTS_GSTREAM_C_END } #else #define HTS_GSTREAM_C_START #define HTS_GSTREAM_C_END #endif /* __CPLUSPLUS */ HTS_GSTREAM_C_START; /* hts_engine libraries */ #include "HTS_hidden.h" /* HTS_GStreamSet_initialize: initialize generated parameter stream set */ void HTS_GStreamSet_initialize(HTS_GStreamSet * gss) { gss->nstream = 0; gss->total_frame = 0; gss->total_nsample = 0; gss->gstream = NULL; gss->gspeech = NULL; } /* HTS_GStreamSet_create: generate speech */ HTS_Boolean HTS_GStreamSet_create(HTS_GStreamSet * gss, HTS_PStreamSet * pss, size_t stage, HTS_Boolean use_log_gain, size_t sampling_rate, size_t fperiod, double alpha, double beta, HTS_Boolean * stop, double volume, HTS_Audio * audio) { size_t i, j, k; size_t msd_frame; HTS_Vocoder v; size_t nlpf = 0; double *lpf = NULL; /* check */ if (gss->gstream || gss->gspeech) { HTS_error(1, "HTS_GStreamSet_create: HTS_GStreamSet is not initialized.\n"); return FALSE; } /* initialize */ gss->nstream = HTS_PStreamSet_get_nstream(pss); gss->total_frame = HTS_PStreamSet_get_total_frame(pss); gss->total_nsample = fperiod * gss->total_frame; gss->gstream = (HTS_GStream *) HTS_calloc(gss->nstream, sizeof(HTS_GStream)); for (i = 0; i < gss->nstream; i++) { gss->gstream[i].vector_length = HTS_PStreamSet_get_vector_length(pss, i); gss->gstream[i].par = (double **) HTS_calloc(gss->total_frame, sizeof(double *)); for (j = 0; j < gss->total_frame; j++) gss->gstream[i].par[j] = (double *) HTS_calloc(gss->gstream[i].vector_length, sizeof(double)); } gss->gspeech = (double *) HTS_calloc(gss->total_nsample, sizeof(double)); /* copy generated parameter */ for (i = 0; i < gss->nstream; i++) { if (HTS_PStreamSet_is_msd(pss, i)) { /* for MSD */ for (j = 0, msd_frame = 0; j < gss->total_frame; j++) if (HTS_PStreamSet_get_msd_flag(pss, i, j)) { for (k = 0; k < gss->gstream[i].vector_length; k++) gss->gstream[i].par[j][k] = HTS_PStreamSet_get_parameter(pss, i, msd_frame, k); msd_frame++; } else for (k = 0; k < gss->gstream[i].vector_length; k++) gss->gstream[i].par[j][k] = HTS_NODATA; } else { /* for non MSD */ for (j = 0; j < gss->total_frame; j++) for (k = 0; k < gss->gstream[i].vector_length; k++) gss->gstream[i].par[j][k] = HTS_PStreamSet_get_parameter(pss, i, j, k); } } /* check */ if (gss->nstream != 2 && gss->nstream != 3) { HTS_error(1, "HTS_GStreamSet_create: The number of streams should be 2 or 3.\n"); HTS_GStreamSet_clear(gss); return FALSE; } if (HTS_PStreamSet_get_vector_length(pss, 1) != 1) { HTS_error(1, "HTS_GStreamSet_create: The size of lf0 static vector should be 1.\n"); HTS_GStreamSet_clear(gss); return FALSE; } if (gss->nstream >= 3 && gss->gstream[2].vector_length % 2 == 0) { HTS_error(1, "HTS_GStreamSet_create: The number of low-pass filter coefficient should be odd numbers."); HTS_GStreamSet_clear(gss); return FALSE; } /* synthesize speech waveform */ HTS_Vocoder_initialize(&v, gss->gstream[0].vector_length - 1, stage, use_log_gain, sampling_rate, fperiod); if (gss->nstream >= 3) nlpf = (gss->gstream[2].vector_length - 1) / 2; for (i = 0; i < gss->total_frame && (*stop) == FALSE; i++) { j = i * fperiod; if (gss->nstream >= 3) lpf = &gss->gstream[2].par[i][0]; HTS_Vocoder_synthesize(&v, gss->gstream[0].vector_length - 1, gss->gstream[1].par[i][0], &gss->gstream[0].par[i][0], nlpf, lpf, alpha, beta, volume, &gss->gspeech[j], audio); } HTS_Vocoder_clear(&v); if (audio) HTS_Audio_flush(audio); return TRUE; } /* HTS_GStreamSet_get_total_nsamples: get total number of sample */ size_t HTS_GStreamSet_get_total_nsamples(HTS_GStreamSet * gss) { return gss->total_nsample; } /* HTS_GStreamSet_get_total_frame: get total number of frame */ size_t HTS_GStreamSet_get_total_frame(HTS_GStreamSet * gss) { return gss->total_frame; } /* HTS_GStreamSet_get_vector_length: get features length */ size_t HTS_GStreamSet_get_vector_length(HTS_GStreamSet * gss, size_t stream_index) { return gss->gstream[stream_index].vector_length; } /* HTS_GStreamSet_get_speech: get synthesized speech parameter */ double HTS_GStreamSet_get_speech(HTS_GStreamSet * gss, size_t sample_index) { return gss->gspeech[sample_index]; } /* HTS_GStreamSet_get_parameter: get generated parameter */ double HTS_GStreamSet_get_parameter(HTS_GStreamSet * gss, size_t stream_index, size_t frame_index, size_t vector_index) { return gss->gstream[stream_index].par[frame_index][vector_index]; } /* HTS_GStreamSet_clear: free generated parameter stream set */ void HTS_GStreamSet_clear(HTS_GStreamSet * gss) { size_t i, j; if (gss->gstream) { for (i = 0; i < gss->nstream; i++) { for (j = 0; j < gss->total_frame; j++) HTS_free(gss->gstream[i].par[j]); HTS_free(gss->gstream[i].par); } HTS_free(gss->gstream); } if (gss->gspeech) HTS_free(gss->gspeech); HTS_GStreamSet_initialize(gss); } HTS_GSTREAM_C_END; #endif /* !HTS_GSTREAM_C */ festival/src/modules/hts_engine/HTS_label.c0000644000175000017500000002165512073131325017330 0ustar awbawb/* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ #ifndef HTS_LABEL_C #define HTS_LABEL_C #ifdef __cplusplus #define HTS_LABEL_C_START extern "C" { #define HTS_LABEL_C_END } #else #define HTS_LABEL_C_START #define HTS_LABEL_C_END #endif /* __CPLUSPLUS */ HTS_LABEL_C_START; #include /* for atof() */ #include /* for isgraph(),isdigit() */ /* hts_engine libraries */ #include "HTS_hidden.h" static HTS_Boolean isdigit_string(char *str) { int i; if (sscanf(str, "%d", &i) == 1) return TRUE; else return FALSE; } /* HTS_Label_initialize: initialize label */ void HTS_Label_initialize(HTS_Label * label) { label->head = NULL; label->size = 0; } /* HTS_Label_check_time: check label */ static void HTS_Label_check_time(HTS_Label * label) { HTS_LabelString *lstring = label->head; HTS_LabelString *next = NULL; if (lstring) lstring->start = 0.0; while (lstring) { next = lstring->next; if (!next) break; if (lstring->end < 0.0 && next->start >= 0.0) lstring->end = next->start; else if (lstring->end >= 0.0 && next->start < 0.0) next->start = lstring->end; if (lstring->start < 0.0) lstring->start = -1.0; if (lstring->end < 0.0) lstring->end = -1.0; lstring = next; } } /* HTS_Label_load: load label */ static void HTS_Label_load(HTS_Label * label, size_t sampling_rate, size_t fperiod, HTS_File * fp) { char buff[HTS_MAXBUFLEN]; HTS_LabelString *lstring = NULL; double start, end; const double rate = (double) sampling_rate / ((double) fperiod * 1e+7); if (label->head || label->size != 0) { HTS_error(1, "HTS_Label_load_from_fp: label is not initialized.\n"); return; } /* parse label file */ while (HTS_get_token_from_fp(fp, buff)) { if (!isgraph((int) buff[0])) break; label->size++; if (lstring) { lstring->next = (HTS_LabelString *) HTS_calloc(1, sizeof(HTS_LabelString)); lstring = lstring->next; } else { /* first time */ lstring = (HTS_LabelString *) HTS_calloc(1, sizeof(HTS_LabelString)); label->head = lstring; } if (isdigit_string(buff)) { /* has frame infomation */ start = atof(buff); HTS_get_token_from_fp(fp, buff); end = atof(buff); HTS_get_token_from_fp(fp, buff); lstring->start = rate * start; lstring->end = rate * end; } else { lstring->start = -1.0; lstring->end = -1.0; } lstring->next = NULL; lstring->name = HTS_strdup(buff); } HTS_Label_check_time(label); } /* HTS_Label_load_from_fn: load label from file name */ void HTS_Label_load_from_fn(HTS_Label * label, size_t sampling_rate, size_t fperiod, const char *fn) { HTS_File *fp = HTS_fopen_from_fn(fn, "r"); HTS_Label_load(label, sampling_rate, fperiod, fp); HTS_fclose(fp); } /* HTS_Label_load_from_strings: load label from strings */ void HTS_Label_load_from_strings(HTS_Label * label, size_t sampling_rate, size_t fperiod, char **lines, size_t num_lines) { char buff[HTS_MAXBUFLEN]; HTS_LabelString *lstring = NULL; size_t i; size_t data_index; double start, end; const double rate = (double) sampling_rate / ((double) fperiod * 1e+7); if (label->head || label->size != 0) { HTS_error(1, "HTS_Label_load_from_fp: label list is not initialized.\n"); return; } /* copy label */ for (i = 0; i < num_lines; i++) { if (!isgraph((int) lines[i][0])) break; label->size++; if (lstring) { lstring->next = (HTS_LabelString *) HTS_calloc(1, sizeof(HTS_LabelString)); lstring = lstring->next; } else { /* first time */ lstring = (HTS_LabelString *) HTS_calloc(1, sizeof(HTS_LabelString)); label->head = lstring; } data_index = 0; if (isdigit_string(lines[i])) { /* has frame infomation */ HTS_get_token_from_string(lines[i], &data_index, buff); start = atof(buff); HTS_get_token_from_string(lines[i], &data_index, buff); end = atof(buff); HTS_get_token_from_string(lines[i], &data_index, buff); lstring->name = HTS_strdup(buff); lstring->start = rate * start; lstring->end = rate * end; } else { lstring->start = -1.0; lstring->end = -1.0; lstring->name = HTS_strdup(lines[i]); } lstring->next = NULL; } HTS_Label_check_time(label); } /* HTS_Label_get_size: get number of label string */ size_t HTS_Label_get_size(HTS_Label * label) { return label->size; } /* HTS_Label_get_string: get label string */ const char *HTS_Label_get_string(HTS_Label * label, size_t index) { size_t i; HTS_LabelString *lstring = label->head; for (i = 0; i < index && lstring; i++) lstring = lstring->next; if (!lstring) return NULL; return lstring->name; } /* HTS_Label_get_start_frame: get start frame */ double HTS_Label_get_start_frame(HTS_Label * label, size_t index) { size_t i; HTS_LabelString *lstring = label->head; for (i = 0; i < index && lstring; i++) lstring = lstring->next; if (!lstring) return -1.0; return lstring->start; } /* HTS_Label_get_end_frame: get end frame */ double HTS_Label_get_end_frame(HTS_Label * label, size_t index) { size_t i; HTS_LabelString *lstring = label->head; for (i = 0; i < index && lstring; i++) lstring = lstring->next; if (!lstring) return -1.0; return lstring->end; } /* HTS_Label_clear: free label */ void HTS_Label_clear(HTS_Label * label) { HTS_LabelString *lstring, *next_lstring; for (lstring = label->head; lstring; lstring = next_lstring) { next_lstring = lstring->next; HTS_free(lstring->name); HTS_free(lstring); } HTS_Label_initialize(label); } HTS_LABEL_C_END; #endif /* !HTS_LABEL_C */ festival/src/modules/hts_engine/HTS_misc.c0000644000175000017500000003741712165025207017212 0ustar awbawb/* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ #ifndef HTS_MISC_C #define HTS_MISC_C #ifdef __cplusplus #define HTS_MISC_C_START extern "C" { #define HTS_MISC_C_END } #else #define HTS_MISC_C_START #define HTS_MISC_C_END #endif /* __CPLUSPLUS */ HTS_MISC_C_START; #include /* for exit(),calloc(),free() */ #include /* for va_list */ #include /* for strcpy(),strlen() */ /* hts_engine libraries */ #include "HTS_hidden.h" #ifdef FESTIVAL #include "EST_walloc.h" #endif /* FESTIVAL */ #define HTS_FILE 0 #define HTS_DATA 1 typedef struct _HTS_Data { unsigned char *data; size_t size; size_t index; } HTS_Data; /* HTS_fopen_from_fn: wrapper for fopen */ HTS_File *HTS_fopen_from_fn(const char *name, const char *opt) { HTS_File *fp = (HTS_File *) HTS_calloc(1, sizeof(HTS_File)); fp->type = HTS_FILE; fp->pointer = (void *) fopen(name, opt); if (fp->pointer == NULL) { HTS_error(0, "HTS_fopen: Cannot open %s.\n", name); HTS_free(fp); return NULL; } return fp; } /* HTS_fopen_from_fp: wrapper for fopen */ HTS_File *HTS_fopen_from_fp(HTS_File * fp, size_t size) { if (fp == NULL || size == 0) return NULL; else if (fp->type == HTS_FILE) { HTS_Data *d; HTS_File *f; d = (HTS_Data *) HTS_calloc(1, sizeof(HTS_Data)); d->data = (unsigned char *) HTS_calloc(size, sizeof(unsigned char)); d->size = size; d->index = 0; if (fread(d->data, sizeof(unsigned char), size, (FILE *) fp->pointer) != size) { free(d->data); free(d); return NULL; } f = (HTS_File *) HTS_calloc(1, sizeof(HTS_File)); f->type = HTS_DATA; f->pointer = (void *) d; return f; } else if (fp->type == HTS_DATA) { HTS_File *f; HTS_Data *tmp1, *tmp2; tmp1 = (HTS_Data *) fp->pointer; if (tmp1->index + size > tmp1->size) return NULL; tmp2 = (HTS_Data *) HTS_calloc(1, sizeof(HTS_Data)); tmp2->data = (unsigned char *) HTS_calloc(size, sizeof(unsigned char)); tmp2->size = size; tmp2->index = 0; memcpy(tmp2->data, &tmp1->data[tmp1->index], size); tmp1->index += size; f = (HTS_File *) HTS_calloc(1, sizeof(HTS_File)); f->type = HTS_DATA; f->pointer = (void *) tmp2; return f; } HTS_error(0, "HTS_fopen_from_fp: Unknown file type.\n"); return NULL; } /* HTS_fopen_from_data: wrapper for fopen */ HTS_File *HTS_fopen_from_data(void *data, size_t size) { HTS_Data *d; HTS_File *f; if (data == NULL || size == 0) return NULL; d = (HTS_Data *) HTS_calloc(1, sizeof(HTS_Data)); d->data = (unsigned char *) HTS_calloc(size, sizeof(unsigned char)); d->size = size; d->index = 0; memcpy(d->data, data, size); f = (HTS_File *) HTS_calloc(1, sizeof(HTS_File)); f->type = HTS_DATA; f->pointer = (void *) d; return f; } /* HTS_fclose: wrapper for fclose */ void HTS_fclose(HTS_File * fp) { if (fp == NULL) { return; } else if (fp->type == HTS_FILE) { if (fp->pointer != NULL) fclose((FILE *) fp->pointer); HTS_free(fp); return; } else if (fp->type == HTS_DATA) { if (fp->pointer != NULL) { HTS_Data *d = (HTS_Data *) fp->pointer; if (d->data != NULL) HTS_free(d->data); HTS_free(d); } HTS_free(fp); return; } HTS_error(0, "HTS_fclose: Unknown file type.\n"); } /* HTS_fgetc: wrapper for fgetc */ int HTS_fgetc(HTS_File * fp) { if (fp == NULL) { return EOF; } else if (fp->type == HTS_FILE) { return fgetc((FILE *) fp->pointer); } else if (fp->type == HTS_DATA) { HTS_Data *d = (HTS_Data *) fp->pointer; if (d->size <= d->index) return EOF; return (int) d->data[d->index++]; } HTS_error(0, "HTS_fgetc: Unknown file type.\n"); return EOF; } /* HTS_feof: wrapper for feof */ int HTS_feof(HTS_File * fp) { if (fp == NULL) { return 1; } else if (fp->type == HTS_FILE) { return feof((FILE *) fp->pointer); } else if (fp->type == HTS_DATA) { HTS_Data *d = (HTS_Data *) fp->pointer; return d->size <= d->index ? 1 : 0; } HTS_error(0, "HTS_feof: Unknown file type.\n"); return 1; } /* HTS_fseek: wrapper for fseek */ int HTS_fseek(HTS_File * fp, long offset, int origin) { if (fp == NULL) { return 1; } else if (fp->type == HTS_FILE) { return fseek((FILE *) fp->pointer, offset, origin); } else if (fp->type == HTS_DATA) { HTS_Data *d = (HTS_Data *) fp->pointer; if (origin == SEEK_SET) { d->index = (size_t) offset; } else if (origin == SEEK_CUR) { d->index += offset; } else if (origin == SEEK_END) { d->index = d->size + offset; } else { return 1; } return 0; } HTS_error(0, "HTS_fseek: Unknown file type.\n"); return 1; } /* HTS_ftell: rapper for ftell */ size_t HTS_ftell(HTS_File * fp) { if (fp == NULL) { return 0; } else if (fp->type == HTS_FILE) { fpos_t pos; fgetpos((FILE *) fp->pointer, &pos); #if defined(_WIN32) || defined(__CYGWIN__) || defined(__APPLE__) return (size_t) pos; #else return (size_t) pos.__pos; #endif /* _WIN32 || __APPLE__ */ } HTS_error(0, "HTS_ftell: Unknown file type.\n"); return 0; } /* HTS_fread: wrapper for fread */ static size_t HTS_fread(void *buf, size_t size, size_t n, HTS_File * fp) { if (fp == NULL || size == 0 || n == 0) { return 0; } else if (fp->type == HTS_FILE) { return fread(buf, size, n, (FILE *) fp->pointer); } else if (fp->type == HTS_DATA) { HTS_Data *d = (HTS_Data *) fp->pointer; size_t i, length = size * n; unsigned char *c = (unsigned char *) buf; for (i = 0; i < length; i++) { if (d->index < d->size) c[i] = d->data[d->index++]; else break; } if (i == 0) return 0; else return i / size; } HTS_error(0, "HTS_fread: Unknown file type.\n"); return 0; } /* HTS_byte_swap: byte swap */ static void HTS_byte_swap(void *p, size_t size, size_t block) { char *q, tmp; size_t i, j; q = (char *) p; for (i = 0; i < block; i++) { for (j = 0; j < (size / 2); j++) { tmp = *(q + j); *(q + j) = *(q + (size - 1 - j)); *(q + (size - 1 - j)) = tmp; } q += size; } } /* HTS_fread_big_endian: fread with byteswap */ size_t HTS_fread_big_endian(void *buf, size_t size, size_t n, HTS_File * fp) { size_t block = HTS_fread(buf, size, n, fp); #ifdef WORDS_LITTLEENDIAN HTS_byte_swap(buf, size, block); #endif /* WORDS_LITTLEENDIAN */ return block; } /* HTS_fread_little_endian: fread with byteswap */ size_t HTS_fread_little_endian(void *buf, size_t size, size_t n, HTS_File * fp) { size_t block = HTS_fread(buf, size, n, fp); #ifdef WORDS_BIGENDIAN HTS_byte_swap(buf, size, block); #endif /* WORDS_BIGENDIAN */ return block; } /* HTS_fwrite_little_endian: fwrite with byteswap */ size_t HTS_fwrite_little_endian(const void *buf, size_t size, size_t n, FILE * fp) { #ifdef WORDS_BIGENDIAN HTS_byte_swap(buf, size, n * size); #endif /* WORDS_BIGENDIAN */ return fwrite(buf, size, n, fp); } /* HTS_get_pattern_token: get pattern token (single/double quote can be used) */ HTS_Boolean HTS_get_pattern_token(HTS_File * fp, char *buff) { char c; size_t i; HTS_Boolean squote = FALSE, dquote = FALSE; if (fp == NULL || HTS_feof(fp)) return FALSE; c = HTS_fgetc(fp); while (c == ' ' || c == '\n') { if (HTS_feof(fp)) return FALSE; c = HTS_fgetc(fp); } if (c == '\'') { /* single quote case */ if (HTS_feof(fp)) return FALSE; c = HTS_fgetc(fp); squote = TRUE; } if (c == '\"') { /*double quote case */ if (HTS_feof(fp)) return FALSE; c = HTS_fgetc(fp); dquote = TRUE; } if (c == ',') { /*special character ',' */ strcpy(buff, ","); return TRUE; } i = 0; while (1) { buff[i++] = c; c = HTS_fgetc(fp); if (squote && c == '\'') break; if (dquote && c == '\"') break; if (!squote && !dquote) { if (c == ' ') break; if (c == '\n') break; if (HTS_feof(fp)) break; } } buff[i] = '\0'; return TRUE; } /* HTS_get_token: get token from file pointer (separators are space, tab, and line break) */ HTS_Boolean HTS_get_token_from_fp(HTS_File * fp, char *buff) { char c; size_t i; if (fp == NULL || HTS_feof(fp)) return FALSE; c = HTS_fgetc(fp); while (c == ' ' || c == '\n' || c == '\t') { if (HTS_feof(fp)) return FALSE; c = HTS_fgetc(fp); if (c == EOF) return FALSE; } for (i = 0; c != ' ' && c != '\n' && c != '\t';) { buff[i++] = c; if (HTS_feof(fp)) break; c = HTS_fgetc(fp); if (c == EOF) break; } buff[i] = '\0'; return TRUE; } /* HTS_get_token_with_separator: get token from file pointer with specified separator */ HTS_Boolean HTS_get_token_from_fp_with_separator(HTS_File * fp, char *buff, char separator) { char c; size_t i; if (fp == NULL || HTS_feof(fp)) return FALSE; c = HTS_fgetc(fp); while (c == separator) { if (HTS_feof(fp)) return FALSE; c = HTS_fgetc(fp); if (c == EOF) return FALSE; } for (i = 0; c != separator;) { buff[i++] = c; if (HTS_feof(fp)) break; c = HTS_fgetc(fp); if (c == EOF) break; } buff[i] = '\0'; return TRUE; } /* HTS_get_token_from_string: get token from string (separators are space, tab, and line break) */ HTS_Boolean HTS_get_token_from_string(const char *string, size_t * index, char *buff) { char c; size_t i; c = string[(*index)]; if (c == '\0') return FALSE; c = string[(*index)++]; if (c == '\0') return FALSE; while (c == ' ' || c == '\n' || c == '\t') { if (c == '\0') return FALSE; c = string[(*index)++]; } for (i = 0; c != ' ' && c != '\n' && c != '\t' && c != '\0'; i++) { buff[i] = c; c = string[(*index)++]; } buff[i] = '\0'; return TRUE; } /* HTS_get_token_from_string_with_separator: get token from string with specified separator */ HTS_Boolean HTS_get_token_from_string_with_separator(const char *str, size_t * index, char *buff, char separator) { char c; size_t start; size_t len = 0; if (str == NULL) return FALSE; c = str[(*index)]; if (c == '\0') return FALSE; while (c == separator) { if (c == '\0') return FALSE; (*index)++; c = str[(*index)]; } start = (*index); while (c != separator && c != '\0') { buff[len++] = c; (*index)++; c = str[(*index)]; } if (c != '\0') (*index)++; buff[len] = '\0'; if (len > 0) return TRUE; else return FALSE; } /* HTS_calloc: wrapper for calloc */ void *HTS_calloc(const size_t num, const size_t size) { size_t n = num * size; #ifdef FESTIVAL void *mem = (void *) safe_wcalloc(n); #else void *mem = (void *) malloc(n); #endif /* FESTIVAL */ memset(mem, 0, n); if (mem == NULL) HTS_error(1, "HTS_calloc: Cannot allocate memory.\n"); return mem; } /* HTS_Free: wrapper for free */ void HTS_free(void *ptr) { #ifdef FESTIVAL wfree(ptr); #else free(ptr); #endif /* FESTIVAL */ } /* HTS_strdup: wrapper for strdup */ char *HTS_strdup(const char *string) { #ifdef FESTIVAL return (wstrdup(string)); #else char *buff = (char *) HTS_calloc(strlen(string) + 1, sizeof(char)); strcpy(buff, string); return buff; #endif /* FESTIVAL */ } /* HTS_alloc_matrix: allocate double matrix */ double **HTS_alloc_matrix(size_t x, size_t y) { size_t i; double **p = (double **) HTS_calloc(x, sizeof(double *)); for (i = 0; i < x; i++) p[i] = (double *) HTS_calloc(y, sizeof(double)); return p; } /* HTS_free_matrix: free double matrix */ void HTS_free_matrix(double **p, size_t x) { size_t i; for (i = 0; i < x; i++) HTS_free(p[i]); HTS_free(p); } /* HTS_error: output error message */ void HTS_error(int error, const char *message, ...) { va_list arg; fflush(stdout); fflush(stderr); if (error > 0) fprintf(stderr, "\nError: "); else fprintf(stderr, "\nWarning: "); va_start(arg, message); vfprintf(stderr, message, arg); va_end(arg); fflush(stderr); if (error > 0) exit(error); } HTS_MISC_C_END; #endif /* !HTS_MISC_C */ festival/src/modules/hts_engine/HTS_model.c0000644000175000017500000015751112073262755017366 0ustar awbawb/* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ #ifndef HTS_MODEL_C #define HTS_MODEL_C #ifdef __cplusplus #define HTS_MODEL_C_START extern "C" { #define HTS_MODEL_C_END } #else #define HTS_MODEL_C_START #define HTS_MODEL_C_END #endif /* __CPLUSPLUS */ HTS_MODEL_C_START; #include /* for atoi(),abs() */ #include /* for strlen(),strstr(),strrchr(),strcmp() */ #include /* for isdigit() */ /* hts_engine libraries */ #include "HTS_hidden.h" /* HTS_dp_match: recursive matching */ static HTS_Boolean HTS_dp_match(const char *string, const char *pattern, size_t pos, size_t max) { if (pos > max) return FALSE; if (string[0] == '\0' && pattern[0] == '\0') return TRUE; if (pattern[0] == '*') { if (HTS_dp_match(string + 1, pattern, pos + 1, max) == 1) return TRUE; else return HTS_dp_match(string, pattern + 1, pos, max); } if (string[0] == pattern[0] || pattern[0] == '?') { if (HTS_dp_match(string + 1, pattern + 1, pos + 1, max + 1) == 1) return TRUE; } return FALSE; } /* HTS_pattern_match: pattern matching function */ static HTS_Boolean HTS_pattern_match(const char *string, const char *pattern) { size_t i, j; size_t buff_length, max = 0, nstar = 0, nquestion = 0; char buff[HTS_MAXBUFLEN]; size_t pattern_length = strlen(pattern); for (i = 0; i < pattern_length; i++) { switch (pattern[i]) { case '*': nstar++; break; case '?': nquestion++; max++; break; default: max++; } } if (nstar == 2 && nquestion == 0 && pattern[0] == '*' && pattern[i - 1] == '*') { /* only string matching is required */ buff_length = i - 2; for (i = 0, j = 1; i < buff_length; i++, j++) buff[i] = pattern[j]; buff[buff_length] = '\0'; if (strstr(string, buff) != NULL) return TRUE; else return FALSE; } else return HTS_dp_match(string, pattern, 0, strlen(string) - max); } /* HTS_is_num: check given buffer is number or not */ static HTS_Boolean HTS_is_num(const char *buff) { size_t i; size_t length = strlen(buff); for (i = 0; i < length; i++) if (!(isdigit((int) buff[i]) || (buff[i] == '-'))) return FALSE; return TRUE; } /* HTS_name2num: convert name of node to number */ static size_t HTS_name2num(const char *buff) { size_t i; for (i = strlen(buff) - 1; '0' <= buff[i] && buff[i] <= '9' && i >= 0; i--); i++; return (size_t) atoi(&buff[i]); } /* HTS_get_state_num: return the number of state */ static size_t HTS_get_state_num(const char *string) { char *left, *right; left = strchr(string, '['); if (left == NULL) return 0; left++; right = strchr(left, ']'); if (right == NULL) return 0; return (size_t) atoi(left); } /* HTS_Question_initialize: initialize question */ static void HTS_Question_initialize(HTS_Question * question) { question->string = NULL; question->head = NULL; question->next = NULL; } /* HTS_Question_clear: clear loaded question */ static void HTS_Question_clear(HTS_Question * question) { HTS_Pattern *pattern, *next_pattern; if (question->string != NULL) HTS_free(question->string); for (pattern = question->head; pattern; pattern = next_pattern) { next_pattern = pattern->next; HTS_free(pattern->string); HTS_free(pattern); } HTS_Question_initialize(question); } /* HTS_Question_load: Load questions from file */ static HTS_Boolean HTS_Question_load(HTS_Question * question, HTS_File * fp) { char buff[HTS_MAXBUFLEN]; HTS_Pattern *pattern, *last_pattern; if (question == NULL || fp == NULL) return FALSE; HTS_Question_clear(question); /* get question name */ if (HTS_get_pattern_token(fp, buff) == FALSE) return FALSE; question->string = HTS_strdup(buff); /* get pattern list */ if (HTS_get_pattern_token(fp, buff) == FALSE) { HTS_Question_clear(question); return FALSE; } last_pattern = NULL; if (strcmp(buff, "{") == 0) { while (1) { if (HTS_get_pattern_token(fp, buff) == FALSE) { HTS_Question_clear(question); return FALSE; } pattern = (HTS_Pattern *) HTS_calloc(1, sizeof(HTS_Pattern)); if (question->head != NULL) last_pattern->next = pattern; else /* first time */ question->head = pattern; pattern->string = HTS_strdup(buff); pattern->next = NULL; if (HTS_get_pattern_token(fp, buff) == FALSE) { HTS_Question_clear(question); return FALSE; } if (!strcmp(buff, "}")) break; last_pattern = pattern; } } return TRUE; } /* HTS_Question_match: check given string match given question */ static HTS_Boolean HTS_Question_match(HTS_Question * question, const char *string) { HTS_Pattern *pattern; for (pattern = question->head; pattern; pattern = pattern->next) if (HTS_pattern_match(string, pattern->string)) return TRUE; return FALSE; } /* HTS_Question_find: find question from question list */ static HTS_Question *HTS_Question_find(HTS_Question * question, const char *string) { for (; question; question = question->next) if (strcmp(string, question->string) == 0) return question; return NULL; } /* HTS_Node_initialzie: initialize node */ static void HTS_Node_initialize(HTS_Node * node) { node->index = 0; node->pdf = 0; node->yes = NULL; node->no = NULL; node->next = NULL; node->quest = NULL; } /* HTS_Node_clear: recursive function to free node */ static void HTS_Node_clear(HTS_Node * node) { if (node->yes != NULL) { HTS_Node_clear(node->yes); HTS_free(node->yes); } if (node->no != NULL) { HTS_Node_clear(node->no); HTS_free(node->no); } HTS_Node_initialize(node); } /* HTS_Node_find: find node for given number */ static HTS_Node *HTS_Node_find(HTS_Node * node, int num) { for (; node; node = node->next) if (node->index == num) return node; return NULL; } /* HTS_Tree_initialize: initialize tree */ static void HTS_Tree_initialize(HTS_Tree * tree) { tree->head = NULL; tree->next = NULL; tree->root = NULL; tree->state = 0; } /* HTS_Tree_clear: clear given tree */ static void HTS_Tree_clear(HTS_Tree * tree) { HTS_Pattern *pattern, *next_pattern; for (pattern = tree->head; pattern; pattern = next_pattern) { next_pattern = pattern->next; HTS_free(pattern->string); HTS_free(pattern); } if (tree->root != NULL) { HTS_Node_clear(tree->root); HTS_free(tree->root); } HTS_Tree_initialize(tree); } /* HTS_Tree_parse_pattern: parse pattern specified for each tree */ static void HTS_Tree_parse_pattern(HTS_Tree * tree, char *string) { char *left, *right; HTS_Pattern *pattern, *last_pattern; tree->head = NULL; last_pattern = NULL; /* parse tree pattern */ if ((left = strchr(string, '{')) != NULL) { /* pattern is specified */ string = left + 1; if (*string == '(') ++string; right = strrchr(string, '}'); if (string < right && *(right - 1) == ')') --right; *right = ','; /* parse pattern */ while ((left = strchr(string, ',')) != NULL) { pattern = (HTS_Pattern *) HTS_calloc(1, sizeof(HTS_Pattern)); if (tree->head) { last_pattern->next = pattern; } else { tree->head = pattern; } *left = '\0'; pattern->string = HTS_strdup(string); string = left + 1; pattern->next = NULL; last_pattern = pattern; } } } /* HTS_Tree_load: load trees */ static HTS_Boolean HTS_Tree_load(HTS_Tree * tree, HTS_File * fp, HTS_Question * question) { char buff[HTS_MAXBUFLEN]; HTS_Node *node, *last_node; if (tree == NULL || fp == NULL) return FALSE; if (HTS_get_pattern_token(fp, buff) == FALSE) { HTS_Tree_clear(tree); return FALSE; } node = (HTS_Node *) HTS_calloc(1, sizeof(HTS_Node)); HTS_Node_initialize(node); tree->root = last_node = node; if (strcmp(buff, "{") == 0) { while (HTS_get_pattern_token(fp, buff) == TRUE && strcmp(buff, "}") != 0) { node = HTS_Node_find(last_node, atoi(buff)); if (node == NULL) { HTS_error(0, "HTS_Tree_load: Cannot find node %d.\n", atoi(buff)); HTS_Tree_clear(tree); return FALSE; } if (HTS_get_pattern_token(fp, buff) == FALSE) { HTS_Tree_clear(tree); return FALSE; } node->quest = HTS_Question_find(question, buff); if (node->quest == NULL) { HTS_error(0, "HTS_Tree_load: Cannot find question %s.\n", buff); HTS_Tree_clear(tree); return FALSE; } node->yes = (HTS_Node *) HTS_calloc(1, sizeof(HTS_Node)); node->no = (HTS_Node *) HTS_calloc(1, sizeof(HTS_Node)); HTS_Node_initialize(node->yes); HTS_Node_initialize(node->no); if (HTS_get_pattern_token(fp, buff) == FALSE) { node->quest = NULL; free(node->yes); free(node->no); HTS_Tree_clear(tree); return FALSE; } if (HTS_is_num(buff)) node->no->index = atoi(buff); else node->no->pdf = HTS_name2num(buff); node->no->next = last_node; last_node = node->no; if (HTS_get_pattern_token(fp, buff) == FALSE) { node->quest = NULL; free(node->yes); free(node->no); HTS_Tree_clear(tree); return FALSE; } if (HTS_is_num(buff)) node->yes->index = atoi(buff); else node->yes->pdf = HTS_name2num(buff); node->yes->next = last_node; last_node = node->yes; } } else { node->pdf = HTS_name2num(buff); } return TRUE; } /* HTS_Node_search: tree search */ static size_t HTS_Tree_search_node(HTS_Tree * tree, const char *string) { HTS_Node *node = tree->root; while (node != NULL) { if (node->quest == NULL) return node->pdf; if (HTS_Question_match(node->quest, string)) { if (node->yes->pdf > 0) return node->yes->pdf; node = node->yes; } else { if (node->no->pdf > 0) return node->no->pdf; node = node->no; } } HTS_error(0, "HTS_Tree_search_node: Cannot find node.\n"); return 1; } /* HTS_Window_initialize: initialize dynamic window */ static void HTS_Window_initialize(HTS_Window * win) { win->size = 0; win->l_width = NULL; win->r_width = NULL; win->coefficient = NULL; win->max_width = 0; } /* HTS_Window_clear: free dynamic window */ static void HTS_Window_clear(HTS_Window * win) { size_t i; if (win->coefficient != NULL) { for (i = 0; i < win->size; i++) { win->coefficient[i] += win->l_width[i]; HTS_free(win->coefficient[i]); } HTS_free(win->coefficient); } if (win->l_width) HTS_free(win->l_width); if (win->r_width) HTS_free(win->r_width); HTS_Window_initialize(win); } /* HTS_Window_load: load dynamic windows */ static HTS_Boolean HTS_Window_load(HTS_Window * win, HTS_File ** fp, size_t size) { size_t i, j; size_t fsize, length; char buff[HTS_MAXBUFLEN]; HTS_Boolean result = TRUE; /* check */ if (win == NULL || fp == NULL || size == 0) return FALSE; win->size = size; win->l_width = (int *) HTS_calloc(win->size, sizeof(int)); win->r_width = (int *) HTS_calloc(win->size, sizeof(int)); win->coefficient = (double **) HTS_calloc(win->size, sizeof(double *)); /* set delta coefficents */ for (i = 0; i < win->size; i++) { if (HTS_get_token_from_fp(fp[i], buff) == FALSE) { result = FALSE; fsize = 1; } else { fsize = atoi(buff); if (fsize == 0) { result = FALSE; fsize = 1; } } /* read coefficients */ win->coefficient[i] = (double *) HTS_calloc(fsize, sizeof(double)); for (j = 0; j < fsize; j++) { if (HTS_get_token_from_fp(fp[i], buff) == FALSE) { result = FALSE; win->coefficient[i][j] = 0.0; } else { win->coefficient[i][j] = (double) atof(buff); } } /* set pointer */ length = fsize / 2; win->coefficient[i] += length; win->l_width[i] = -1 * (int) length; win->r_width[i] = (int) length; if (fsize % 2 == 0) win->r_width[i]--; } /* calcurate max_width to determine size of band matrix */ win->max_width = 0; for (i = 0; i < win->size; i++) { if (win->max_width < (size_t) abs(win->l_width[i])) win->max_width = abs(win->l_width[i]); if (win->max_width < (size_t) abs(win->r_width[i])) win->max_width = abs(win->r_width[i]); } if (result == FALSE) { HTS_Window_clear(win); return FALSE; } return TRUE; } /* HTS_Model_initialize: initialize model */ static void HTS_Model_initialize(HTS_Model * model) { model->vector_length = 0; model->num_windows = 0; model->is_msd = FALSE; model->ntree = 0; model->npdf = NULL; model->pdf = NULL; model->tree = NULL; model->question = NULL; } /* HTS_Model_clear: free pdfs and trees */ static void HTS_Model_clear(HTS_Model * model) { size_t i, j; HTS_Question *question, *next_question; HTS_Tree *tree, *next_tree; for (question = model->question; question; question = next_question) { next_question = question->next; HTS_Question_clear(question); HTS_free(question); } for (tree = model->tree; tree; tree = next_tree) { next_tree = tree->next; HTS_Tree_clear(tree); HTS_free(tree); } if (model->pdf) { for (i = 2; i <= model->ntree + 1; i++) { for (j = 1; j <= model->npdf[i]; j++) { HTS_free(model->pdf[i][j]); } model->pdf[i]++; HTS_free(model->pdf[i]); } model->pdf += 2; HTS_free(model->pdf); } if (model->npdf) { model->npdf += 2; HTS_free(model->npdf); } HTS_Model_initialize(model); } /* HTS_Model_load_tree: load trees */ static HTS_Boolean HTS_Model_load_tree(HTS_Model * model, HTS_File * fp) { char buff[HTS_MAXBUFLEN]; HTS_Question *question, *last_question; HTS_Tree *tree, *last_tree; size_t state; /* check */ if (model == NULL) { HTS_error(0, "HTS_Model_load_tree: File for trees is not specified.\n"); return FALSE; } if (fp == NULL) { model->ntree = 1; return TRUE; } model->ntree = 0; last_question = NULL; last_tree = NULL; while (!HTS_feof(fp)) { HTS_get_pattern_token(fp, buff); /* parse questions */ if (strcmp(buff, "QS") == 0) { question = (HTS_Question *) HTS_calloc(1, sizeof(HTS_Question)); HTS_Question_initialize(question); if (HTS_Question_load(question, fp) == FALSE) { free(question); HTS_Model_clear(model); return FALSE; } if (model->question) last_question->next = question; else model->question = question; question->next = NULL; last_question = question; } /* parse trees */ state = HTS_get_state_num(buff); if (state != 0) { tree = (HTS_Tree *) HTS_calloc(1, sizeof(HTS_Tree)); HTS_Tree_initialize(tree); tree->state = state; HTS_Tree_parse_pattern(tree, buff); if (HTS_Tree_load(tree, fp, model->question) == FALSE) { free(tree); HTS_Model_clear(model); return FALSE; } if (model->tree) last_tree->next = tree; else model->tree = tree; tree->next = NULL; last_tree = tree; model->ntree++; } } /* No Tree information in tree file */ if (model->tree == NULL) model->ntree = 1; return TRUE; } /* HTS_Model_load_pdf: load pdfs */ static HTS_Boolean HTS_Model_load_pdf(HTS_Model * model, HTS_File * fp, size_t vector_length, size_t num_windows, HTS_Boolean is_msd) { unsigned int i; size_t j, k; HTS_Boolean result = TRUE; size_t len; /* check */ if (model == NULL || fp == NULL || model->ntree <= 0) { HTS_error(1, "HTS_Model_load_pdf: File for pdfs is not specified.\n"); return FALSE; } /* read MSD flag */ model->vector_length = vector_length; model->num_windows = num_windows; model->is_msd = is_msd; model->npdf = (size_t *) HTS_calloc(model->ntree, sizeof(size_t)); model->npdf -= 2; /* read the number of pdfs */ for (j = 2; j <= model->ntree + 1; j++) { if (HTS_fread_little_endian(&i, sizeof(unsigned int), 1, fp) != 1) { result = FALSE; break; } model->npdf[j] = (size_t) i; } for (j = 2; j <= model->ntree + 1; j++) { if (model->npdf[j] <= 0) { HTS_error(1, "HTS_Model_load_pdf: # of pdfs at %d-th state should be positive.\n", j); result = FALSE; break; } } if (result == FALSE) { model->npdf += 2; free(model->npdf); HTS_Model_initialize(model); return FALSE; } model->pdf = (float ***) HTS_calloc(model->ntree, sizeof(float **)); model->pdf -= 2; /* read means and variances */ if (is_msd) /* for MSD */ len = model->vector_length * model->num_windows * 2 + 1; else len = model->vector_length * model->num_windows * 2; for (j = 2; j <= model->ntree + 1; j++) { model->pdf[j] = (float **) HTS_calloc(model->npdf[j], sizeof(float *)); model->pdf[j]--; for (k = 1; k <= model->npdf[j]; k++) { model->pdf[j][k] = (float *) HTS_calloc(len, sizeof(float)); if (HTS_fread_little_endian(model->pdf[j][k], sizeof(float), len, fp) != len) result = FALSE; } } if (result == FALSE) { HTS_Model_clear(model); return FALSE; } return TRUE; } /* HTS_Model_load: load pdf and tree */ static HTS_Boolean HTS_Model_load(HTS_Model * model, HTS_File * pdf, HTS_File * tree, size_t vector_length, size_t num_windows, HTS_Boolean is_msd) { /* check */ if (model == NULL || pdf == NULL || vector_length == 0 || num_windows == 0) return FALSE; /* reset */ HTS_Model_clear(model); /* load tree */ if (HTS_Model_load_tree(model, tree) != TRUE) { HTS_Model_clear(model); return FALSE; } /* load pdf */ if (HTS_Model_load_pdf(model, pdf, vector_length, num_windows, is_msd) != TRUE) { HTS_Model_clear(model); return FALSE; } return TRUE; } /* HTS_Model_get_index: get index of tree and PDF */ static void HTS_Model_get_index(HTS_Model * model, size_t state_index, const char *string, size_t * tree_index, size_t * pdf_index) { HTS_Tree *tree; HTS_Pattern *pattern; HTS_Boolean find; (*tree_index) = 2; (*pdf_index) = 1; if (model->tree == NULL) return; find = FALSE; for (tree = model->tree; tree; tree = tree->next) { if (tree->state == state_index) { pattern = tree->head; if (!pattern) find = TRUE; for (; pattern; pattern = pattern->next) if (HTS_pattern_match(string, pattern->string)) { find = TRUE; break; } if (find) break; } (*tree_index)++; } if (tree != NULL) { (*pdf_index) = HTS_Tree_search_node(tree, string); } else { (*pdf_index) = HTS_Tree_search_node(model->tree, string); } } /* HTS_ModelSet_initialize: initialize model set */ void HTS_ModelSet_initialize(HTS_ModelSet * ms) { ms->hts_voice_version = NULL; ms->sampling_frequency = 0; ms->frame_period = 0; ms->num_voices = 0; ms->num_states = 0; ms->num_streams = 0; ms->stream_type = NULL; ms->fullcontext_format = NULL; ms->fullcontext_version = NULL; ms->gv_off_context = NULL; ms->option = NULL; ms->duration = NULL; ms->window = NULL; ms->stream = NULL; ms->gv = NULL; } /* HTS_ModelSet_clear: free model set */ void HTS_ModelSet_clear(HTS_ModelSet * ms) { size_t i, j; if (ms->hts_voice_version != NULL) free(ms->hts_voice_version); if (ms->stream_type != NULL) free(ms->stream_type); if (ms->fullcontext_format != NULL) free(ms->fullcontext_format); if (ms->fullcontext_version != NULL) free(ms->fullcontext_version); if (ms->gv_off_context != NULL) { HTS_Question_clear(ms->gv_off_context); free(ms->gv_off_context); } if (ms->option != NULL) { for (i = 0; i < ms->num_streams; i++) if (ms->option[i] != NULL) free(ms->option[i]); free(ms->option); } if (ms->duration != NULL) { for (i = 0; i < ms->num_voices; i++) HTS_Model_clear(&ms->duration[i]); free(ms->duration); } if (ms->window != NULL) { for (i = 0; i < ms->num_streams; i++) HTS_Window_clear(&ms->window[i]); free(ms->window); } if (ms->stream != NULL) { for (i = 0; i < ms->num_voices; i++) { for (j = 0; j < ms->num_streams; j++) HTS_Model_clear(&ms->stream[i][j]); free(ms->stream[i]); } HTS_free(ms->stream); } if (ms->gv != NULL) { for (i = 0; i < ms->num_voices; i++) { for (j = 0; j < ms->num_streams; j++) HTS_Model_clear(&ms->gv[i][j]); free(ms->gv[i]); } free(ms->gv); } HTS_ModelSet_initialize(ms); } /* HTS_match_head_string: return true if head of str is equal to pattern */ static HTS_Boolean HTS_match_head_string(const char *str, const char *pattern, size_t * matched_size) { (*matched_size) = 0; while (1) { if (pattern[(*matched_size)] == '\0') return TRUE; if (str[(*matched_size)] == '\0') return FALSE; if (str[(*matched_size)] != pattern[(*matched_size)]) return FALSE; (*matched_size)++; } } /* HTS_strequal: strcmp wrapper */ static HTS_Boolean HTS_strequal(const char *s1, const char *s2) { if (s1 == NULL && s2 == NULL) return TRUE; else if (s1 == NULL || s2 == NULL) return FALSE; else return strcmp(s1, s2) == 0 ? TRUE : FALSE; } /* HTS_ModelSet_load: load model set */ HTS_Boolean HTS_ModelSet_load(HTS_ModelSet * ms, char **voices, size_t num_voices) { size_t i, j, k, s, e; HTS_Boolean error = FALSE; HTS_File *fp; char buff1[HTS_MAXBUFLEN]; char buff2[HTS_MAXBUFLEN]; size_t matched_size; char **stream_type_list = NULL; size_t *vector_length = NULL; HTS_Boolean *is_msd = NULL; size_t *num_windows = NULL; HTS_Boolean *use_gv = NULL; char *gv_off_context = NULL; /* temporary values */ char *temp_hts_voice_version; size_t temp_sampling_frequency; size_t temp_frame_period; size_t temp_num_states; size_t temp_num_streams; char *temp_stream_type; char *temp_fullcontext_format; char *temp_fullcontext_version; char *temp_gv_off_context; size_t *temp_vector_length; HTS_Boolean *temp_is_msd; size_t *temp_num_windows; HTS_Boolean *temp_use_gv; char **temp_option; char *temp_duration_pdf; char *temp_duration_tree; char ***temp_stream_win; char **temp_stream_pdf; char **temp_stream_tree; char **temp_gv_pdf; char **temp_gv_tree; long start_of_data; HTS_File *pdf_fp = NULL; HTS_File *tree_fp = NULL; HTS_File **win_fp = NULL; HTS_File *gv_off_context_fp = NULL; HTS_ModelSet_clear(ms); if (ms == NULL || voices == NULL || num_voices < 1) return FALSE; ms->num_voices = num_voices; for (i = 0; i < num_voices && error == FALSE; i++) { /* open file */ fp = HTS_fopen_from_fn(voices[i], "rb"); if (fp == NULL) { error = TRUE; break; } /* reset GLOBAL options */ temp_hts_voice_version = NULL; temp_sampling_frequency = 0; temp_frame_period = 0; temp_num_states = 0; temp_num_streams = 0; temp_stream_type = NULL; temp_fullcontext_format = NULL; temp_fullcontext_version = NULL; temp_gv_off_context = NULL; if (HTS_get_token_from_fp_with_separator(fp, buff1, '\n') != TRUE) { error = TRUE; break; } /* load GLOBAL options */ if (HTS_strequal(buff1, "[GLOBAL]") != TRUE) { error = TRUE; break; } while (1) { if (HTS_get_token_from_fp_with_separator(fp, buff1, '\n') != TRUE) { error = TRUE; break; } if (HTS_strequal(buff1, "[STREAM]") == TRUE) { break; } else if (HTS_match_head_string(buff1, "HTS_VOICE_VERSION:", &matched_size) == TRUE) { if (temp_hts_voice_version != NULL) free(temp_hts_voice_version); temp_hts_voice_version = HTS_strdup(&buff1[matched_size]); } else if (HTS_match_head_string(buff1, "SAMPLING_FREQUENCY:", &matched_size) == TRUE) { temp_sampling_frequency = (size_t) atoi(&buff1[matched_size]); } else if (HTS_match_head_string(buff1, "FRAME_PERIOD:", &matched_size) == TRUE) { temp_frame_period = (size_t) atoi(&buff1[matched_size]); } else if (HTS_match_head_string(buff1, "NUM_STATES:", &matched_size) == TRUE) { temp_num_states = (size_t) atoi(&buff1[matched_size]); } else if (HTS_match_head_string(buff1, "NUM_STREAMS:", &matched_size) == TRUE) { temp_num_streams = (size_t) atoi(&buff1[matched_size]); } else if (HTS_match_head_string(buff1, "STREAM_TYPE:", &matched_size) == TRUE) { if (temp_stream_type != NULL) free(temp_stream_type); temp_stream_type = HTS_strdup(&buff1[matched_size]); } else if (HTS_match_head_string(buff1, "FULLCONTEXT_FORMAT:", &matched_size) == TRUE) { if (temp_fullcontext_format != NULL) free(temp_fullcontext_format); temp_fullcontext_format = HTS_strdup(&buff1[matched_size]); } else if (HTS_match_head_string(buff1, "FULLCONTEXT_VERSION:", &matched_size) == TRUE) { if (temp_fullcontext_version != NULL) free(temp_fullcontext_version); temp_fullcontext_version = HTS_strdup(&buff1[matched_size]); } else if (HTS_match_head_string(buff1, "GV_OFF_CONTEXT:", &matched_size) == TRUE) { if (temp_gv_off_context != NULL) free(temp_gv_off_context); temp_gv_off_context = HTS_strdup(&buff1[matched_size]); } else if (HTS_match_head_string(buff1, "COMMENT:", &matched_size) == TRUE) { } else { HTS_error(0, "HTS_ModelSet_load: Unknown option %s.\n", buff1); } } /* check GLOBAL options */ if (i == 0) { ms->hts_voice_version = temp_hts_voice_version; ms->sampling_frequency = temp_sampling_frequency; ms->frame_period = temp_frame_period; ms->num_states = temp_num_states; ms->num_streams = temp_num_streams; ms->stream_type = temp_stream_type; ms->fullcontext_format = temp_fullcontext_format; ms->fullcontext_version = temp_fullcontext_version; gv_off_context = temp_gv_off_context; } else { if (HTS_strequal(ms->hts_voice_version, temp_hts_voice_version) != TRUE) error = TRUE; if (ms->sampling_frequency != temp_sampling_frequency) error = TRUE; if (ms->frame_period != temp_frame_period) error = TRUE; if (ms->num_states != temp_num_states) error = TRUE; if (ms->num_streams != temp_num_streams) error = TRUE; if (HTS_strequal(ms->stream_type, temp_stream_type) != TRUE) error = TRUE; if (HTS_strequal(ms->fullcontext_format, temp_fullcontext_format) != TRUE) error = TRUE; if (HTS_strequal(ms->fullcontext_version, temp_fullcontext_version) != TRUE) error = TRUE; if (HTS_strequal(gv_off_context, temp_gv_off_context) != TRUE) error = TRUE; if (temp_hts_voice_version != NULL) free(temp_hts_voice_version); if (temp_stream_type != NULL) free(temp_stream_type); if (temp_fullcontext_format != NULL) free(temp_fullcontext_format); if (temp_fullcontext_version != NULL) free(temp_fullcontext_version); if (temp_gv_off_context != NULL) free(temp_gv_off_context); } /* find stream names */ if (i == 0) { stream_type_list = (char **) HTS_calloc(ms->num_streams, sizeof(char *)); for (j = 0, matched_size = 0; j < ms->num_streams; j++) { if (HTS_get_token_from_string_with_separator(ms->stream_type, &matched_size, buff2, ',') == TRUE) { stream_type_list[j] = HTS_strdup(buff2); } else { stream_type_list[j] = NULL; error = TRUE; } } } if (error != FALSE) { HTS_fclose(fp); break; } /* reset STREAM options */ temp_vector_length = (size_t *) HTS_calloc(ms->num_streams, sizeof(size_t)); for (j = 0; j < ms->num_streams; j++) temp_vector_length[j] = 0; temp_is_msd = (HTS_Boolean *) HTS_calloc(ms->num_streams, sizeof(HTS_Boolean)); for (j = 0; j < ms->num_streams; j++) temp_is_msd[j] = FALSE; temp_num_windows = (size_t *) HTS_calloc(ms->num_streams, sizeof(size_t)); for (j = 0; j < ms->num_streams; j++) temp_num_windows[j] = 0; temp_use_gv = (HTS_Boolean *) HTS_calloc(ms->num_streams, sizeof(HTS_Boolean)); for (j = 0; j < ms->num_streams; j++) temp_use_gv[j] = FALSE; temp_option = (char **) HTS_calloc(ms->num_streams, sizeof(char *)); for (j = 0; j < ms->num_streams; j++) temp_option[j] = NULL; /* load STREAM options */ while (1) { if (HTS_get_token_from_fp_with_separator(fp, buff1, '\n') != TRUE) { error = TRUE; break; } if (strcmp(buff1, "[POSITION]") == 0) { break; } else if (HTS_match_head_string(buff1, "VECTOR_LENGTH[", &matched_size) == TRUE) { if (HTS_get_token_from_string_with_separator(buff1, &matched_size, buff2, ']') == TRUE) { if (buff1[matched_size++] == ':') { for (j = 0; j < ms->num_streams; j++) if (strcmp(stream_type_list[j], buff2) == 0) { temp_vector_length[j] = (size_t) atoi(&buff1[matched_size]); break; } } } } else if (HTS_match_head_string(buff1, "IS_MSD[", &matched_size) == TRUE) { if (HTS_get_token_from_string_with_separator(buff1, &matched_size, buff2, ']') == TRUE) { if (buff1[matched_size++] == ':') { for (j = 0; j < ms->num_streams; j++) if (strcmp(stream_type_list[j], buff2) == 0) { temp_is_msd[j] = (buff1[matched_size] == '1') ? TRUE : FALSE; break; } } } } else if (HTS_match_head_string(buff1, "NUM_WINDOWS[", &matched_size) == TRUE) { if (HTS_get_token_from_string_with_separator(buff1, &matched_size, buff2, ']') == TRUE) { if (buff1[matched_size++] == ':') { for (j = 0; j < ms->num_streams; j++) if (strcmp(stream_type_list[j], buff2) == 0) { temp_num_windows[j] = (size_t) atoi(&buff1[matched_size]); break; } } } } else if (HTS_match_head_string(buff1, "USE_GV[", &matched_size) == TRUE) { if (HTS_get_token_from_string_with_separator(buff1, &matched_size, buff2, ']') == TRUE) { if (buff1[matched_size++] == ':') { for (j = 0; j < ms->num_streams; j++) if (strcmp(stream_type_list[j], buff2) == 0) { temp_use_gv[j] = (buff1[matched_size] == '1') ? TRUE : FALSE; break; } } } } else if (HTS_match_head_string(buff1, "OPTION[", &matched_size) == TRUE) { if (HTS_get_token_from_string_with_separator(buff1, &matched_size, buff2, ']') == TRUE) { if (buff1[matched_size++] == ':') { for (j = 0; j < ms->num_streams; j++) if (strcmp(stream_type_list[j], buff2) == 0) { if (temp_option[j] != NULL) free(temp_option[j]); temp_option[j] = HTS_strdup(&buff1[matched_size]); break; } } } } else { HTS_error(0, "HTS_ModelSet_load: Unknown option %s.\n", buff1); } } /* check STREAM options */ if (i == 0) { vector_length = temp_vector_length; is_msd = temp_is_msd; num_windows = temp_num_windows; use_gv = temp_use_gv; ms->option = temp_option; } else { for (j = 0; j < ms->num_streams; j++) if (vector_length[j] != temp_vector_length[j]) error = TRUE; for (j = 0; j < ms->num_streams; j++) if (is_msd[j] != is_msd[j]) error = TRUE; for (j = 0; j < ms->num_streams; j++) if (num_windows[j] != temp_num_windows[j]) error = TRUE; for (j = 0; j < ms->num_streams; j++) if (use_gv[j] != temp_use_gv[j]) error = TRUE; for (j = 0; j < ms->num_streams; j++) if (HTS_strequal(ms->option[j], temp_option[j]) != TRUE) error = TRUE; free(temp_vector_length); free(temp_is_msd); free(temp_num_windows); free(temp_use_gv); for (j = 0; j < ms->num_streams; j++) if (temp_option[j] != NULL) free(temp_option[j]); free(temp_option); } if (error != FALSE) { HTS_fclose(fp); break; } /* reset POSITION */ temp_duration_pdf = NULL; temp_duration_tree = NULL; temp_stream_win = (char ***) HTS_calloc(ms->num_streams, sizeof(char **)); for (j = 0; j < ms->num_streams; j++) { temp_stream_win[j] = (char **) HTS_calloc(num_windows[j], sizeof(char *)); for (k = 0; k < num_windows[j]; k++) temp_stream_win[j][k] = NULL; } temp_stream_pdf = (char **) HTS_calloc(ms->num_streams, sizeof(char *)); for (j = 0; j < ms->num_streams; j++) temp_stream_pdf[j] = NULL; temp_stream_tree = (char **) HTS_calloc(ms->num_streams, sizeof(char *)); for (j = 0; j < ms->num_streams; j++) temp_stream_tree[j] = NULL; temp_gv_pdf = (char **) HTS_calloc(ms->num_streams, sizeof(char *)); for (j = 0; j < ms->num_streams; j++) temp_gv_pdf[j] = NULL; temp_gv_tree = (char **) HTS_calloc(ms->num_streams, sizeof(char *)); for (j = 0; j < ms->num_streams; j++) temp_gv_tree[j] = NULL; /* load POSITION */ while (1) { if (HTS_get_token_from_fp_with_separator(fp, buff1, '\n') != TRUE) { error = TRUE; break; } if (strcmp(buff1, "[DATA]") == 0) { break; } else if (HTS_match_head_string(buff1, "DURATION_PDF:", &matched_size) == TRUE) { if (temp_duration_pdf != NULL) free(temp_duration_pdf); temp_duration_pdf = HTS_strdup(&buff1[matched_size]); } else if (HTS_match_head_string(buff1, "DURATION_TREE:", &matched_size) == TRUE) { if (temp_duration_tree != NULL) free(temp_duration_tree); temp_duration_tree = HTS_strdup(&buff1[matched_size]); } else if (HTS_match_head_string(buff1, "STREAM_WIN[", &matched_size) == TRUE) { if (HTS_get_token_from_string_with_separator(buff1, &matched_size, buff2, ']') == TRUE) { if (buff1[matched_size++] == ':') { for (j = 0; j < ms->num_streams; j++) { if (strcmp(stream_type_list[j], buff2) == 0) { for (k = 0; k < num_windows[j]; k++) { if (HTS_get_token_from_string_with_separator(buff1, &matched_size, buff2, ',') == TRUE) temp_stream_win[j][k] = HTS_strdup(buff2); else error = TRUE; } break; } } } } } else if (HTS_match_head_string(buff1, "STREAM_PDF[", &matched_size) == TRUE) { if (HTS_get_token_from_string_with_separator(buff1, &matched_size, buff2, ']') == TRUE) { if (buff1[matched_size++] == ':') { for (j = 0; j < ms->num_streams; j++) { if (strcmp(stream_type_list[j], buff2) == 0) { if (temp_stream_pdf[j] != NULL) free(temp_stream_pdf[j]); temp_stream_pdf[j] = HTS_strdup(&buff1[matched_size]); break; } } } } } else if (HTS_match_head_string(buff1, "STREAM_TREE[", &matched_size) == TRUE) { if (HTS_get_token_from_string_with_separator(buff1, &matched_size, buff2, ']') == TRUE) { if (buff1[matched_size++] == ':') { for (j = 0; j < ms->num_streams; j++) { if (strcmp(stream_type_list[j], buff2) == 0) { if (temp_stream_tree[j] != NULL) free(temp_stream_tree[j]); temp_stream_tree[j] = HTS_strdup(&buff1[matched_size]); break; } } } } } else if (HTS_match_head_string(buff1, "GV_PDF[", &matched_size) == TRUE) { if (HTS_get_token_from_string_with_separator(buff1, &matched_size, buff2, ']') == TRUE) { if (buff1[matched_size++] == ':') { for (j = 0; j < ms->num_streams; j++) { if (strcmp(stream_type_list[j], buff2) == 0) { if (temp_gv_pdf[j] != NULL) free(temp_gv_pdf[j]); temp_gv_pdf[j] = HTS_strdup(&buff1[matched_size]); break; } } } } } else if (HTS_match_head_string(buff1, "GV_TREE[", &matched_size) == TRUE) { if (HTS_get_token_from_string_with_separator(buff1, &matched_size, buff2, ']') == TRUE) { if (buff1[matched_size++] == ':') { for (j = 0; j < ms->num_streams; j++) { if (strcmp(stream_type_list[j], buff2) == 0) { if (temp_gv_tree[j] != NULL) free(temp_gv_tree[j]); temp_gv_tree[j] = HTS_strdup(&buff1[matched_size]); break; } } } } } else { HTS_error(0, "HTS_ModelSet_load: Unknown option %s.\n", buff1); } } /* check POSITION */ if (temp_duration_pdf == NULL) error = TRUE; for (j = 0; j < ms->num_streams; j++) for (k = 0; k < num_windows[j]; k++) if (temp_stream_win[j][k] == NULL) error = TRUE; for (j = 0; j < ms->num_streams; j++) if (temp_stream_pdf[j] == NULL) error = TRUE; /* prepare memory */ if (i == 0) { ms->duration = (HTS_Model *) HTS_calloc(num_voices, sizeof(HTS_Model)); for (j = 0; j < num_voices; j++) HTS_Model_initialize(&ms->duration[j]); ms->window = (HTS_Window *) HTS_calloc(ms->num_streams, sizeof(HTS_Window)); for (j = 0; j < ms->num_streams; j++) HTS_Window_initialize(&ms->window[j]); ms->stream = (HTS_Model **) HTS_calloc(num_voices, sizeof(HTS_Model *)); for (j = 0; j < num_voices; j++) { ms->stream[j] = (HTS_Model *) HTS_calloc(ms->num_streams, sizeof(HTS_Model)); for (k = 0; k < ms->num_streams; k++) HTS_Model_initialize(&ms->stream[j][k]); } ms->gv = (HTS_Model **) HTS_calloc(num_voices, sizeof(HTS_Model *)); for (j = 0; j < num_voices; j++) { ms->gv[j] = (HTS_Model *) HTS_calloc(ms->num_streams, sizeof(HTS_Model)); for (k = 0; k < ms->num_streams; k++) HTS_Model_initialize(&ms->gv[j][k]); } } start_of_data = HTS_ftell(fp); /* load duration */ pdf_fp = NULL; tree_fp = NULL; matched_size = 0; if (HTS_get_token_from_string_with_separator(temp_duration_pdf, &matched_size, buff2, '-') == TRUE) { s = (size_t) atoi(buff2); e = (size_t) atoi(&temp_duration_pdf[matched_size]); HTS_fseek(fp, (long) s, SEEK_CUR); pdf_fp = HTS_fopen_from_fp(fp, e - s + 1); HTS_fseek(fp, start_of_data, SEEK_SET); } matched_size = 0; if (HTS_get_token_from_string_with_separator(temp_duration_tree, &matched_size, buff2, '-') == TRUE) { s = (size_t) atoi(buff2); e = (size_t) atoi(&temp_duration_tree[matched_size]); HTS_fseek(fp, (long) s, SEEK_CUR); tree_fp = HTS_fopen_from_fp(fp, e - s + 1); HTS_fseek(fp, start_of_data, SEEK_SET); } if (HTS_Model_load(&ms->duration[i], pdf_fp, tree_fp, ms->num_states, 1, FALSE) != TRUE) error = TRUE; HTS_fclose(pdf_fp); HTS_fclose(tree_fp); /* load windows */ for (j = 0; j < ms->num_streams; j++) { win_fp = (HTS_File **) HTS_calloc(num_windows[j], sizeof(HTS_File *)); for (k = 0; k < num_windows[j]; k++) win_fp[k] = NULL; for (k = 0; k < num_windows[j]; k++) { matched_size = 0; if (HTS_get_token_from_string_with_separator(temp_stream_win[j][k], &matched_size, buff2, '-') == TRUE) { s = (size_t) atoi(buff2); e = (size_t) atoi(&temp_stream_win[j][k][matched_size]); HTS_fseek(fp, (long) s, SEEK_CUR); win_fp[k] = HTS_fopen_from_fp(fp, e - s + 1); HTS_fseek(fp, start_of_data, SEEK_SET); } } if (HTS_Window_load(&ms->window[j], win_fp, num_windows[j]) != TRUE) error = TRUE; for (k = 0; k < num_windows[j]; k++) HTS_fclose(win_fp[k]); free(win_fp); } /* load streams */ for (j = 0; j < ms->num_streams; j++) { pdf_fp = NULL; tree_fp = NULL; matched_size = 0; if (HTS_get_token_from_string_with_separator(temp_stream_pdf[j], &matched_size, buff2, '-') == TRUE) { s = (size_t) atoi(buff2); e = (size_t) atoi(&temp_stream_pdf[j][matched_size]); HTS_fseek(fp, (long) s, SEEK_CUR); pdf_fp = HTS_fopen_from_fp(fp, e - s + 1); HTS_fseek(fp, start_of_data, SEEK_SET); } matched_size = 0; if (HTS_get_token_from_string_with_separator(temp_stream_tree[j], &matched_size, buff2, '-') == TRUE) { s = (size_t) atoi(buff2); e = (size_t) atoi(&temp_stream_tree[j][matched_size]); HTS_fseek(fp, (long) s, SEEK_CUR); tree_fp = HTS_fopen_from_fp(fp, e - s + 1); HTS_fseek(fp, start_of_data, SEEK_SET); } if (HTS_Model_load(&ms->stream[i][j], pdf_fp, tree_fp, vector_length[j], num_windows[j], is_msd[j]) != TRUE) error = TRUE; HTS_fclose(pdf_fp); HTS_fclose(tree_fp); } /* load GVs */ for (j = 0; j < ms->num_streams; j++) { pdf_fp = NULL; tree_fp = NULL; matched_size = 0; if (HTS_get_token_from_string_with_separator(temp_gv_pdf[j], &matched_size, buff2, '-') == TRUE) { s = (size_t) atoi(buff2); e = (size_t) atoi(&temp_gv_pdf[j][matched_size]); HTS_fseek(fp, (long) s, SEEK_CUR); pdf_fp = HTS_fopen_from_fp(fp, e - s + 1); HTS_fseek(fp, start_of_data, SEEK_SET); } matched_size = 0; if (HTS_get_token_from_string_with_separator(temp_gv_tree[j], &matched_size, buff2, '-') == TRUE) { s = (size_t) atoi(buff2); e = (size_t) atoi(&temp_gv_tree[j][matched_size]); HTS_fseek(fp, (long) s, SEEK_CUR); tree_fp = HTS_fopen_from_fp(fp, e - s + 1); HTS_fseek(fp, start_of_data, SEEK_SET); } if (use_gv[j] == TRUE) { if (HTS_Model_load(&ms->gv[i][j], pdf_fp, tree_fp, vector_length[j], 1, FALSE) != TRUE) error = TRUE; } HTS_fclose(pdf_fp); HTS_fclose(tree_fp); } /* free */ if (temp_duration_pdf != NULL) free(temp_duration_pdf); if (temp_duration_tree != NULL) free(temp_duration_tree); for (j = 0; j < ms->num_streams; j++) { for (k = 0; k < num_windows[j]; k++) if (temp_stream_win[j][k] != NULL) free(temp_stream_win[j][k]); free(temp_stream_win[j]); } free(temp_stream_win); for (j = 0; j < ms->num_streams; j++) if (temp_stream_pdf[j] != NULL) free(temp_stream_pdf[j]); free(temp_stream_pdf); for (j = 0; j < ms->num_streams; j++) if (temp_stream_tree[j] != NULL) free(temp_stream_tree[j]); free(temp_stream_tree); for (j = 0; j < ms->num_streams; j++) if (temp_gv_pdf[j] != NULL) free(temp_gv_pdf[j]); free(temp_gv_pdf); for (j = 0; j < ms->num_streams; j++) if (temp_gv_tree[j] != NULL) free(temp_gv_tree[j]); free(temp_gv_tree); /* fclose */ HTS_fclose(fp); if (error != FALSE) break; } if (gv_off_context != NULL) { sprintf(buff1, "GV-Off { %s }", gv_off_context); gv_off_context_fp = HTS_fopen_from_data((void *) buff1, strlen(buff1) + 1); ms->gv_off_context = (HTS_Question *) HTS_calloc(1, sizeof(HTS_Question)); HTS_Question_initialize(ms->gv_off_context); HTS_Question_load(ms->gv_off_context, gv_off_context_fp); HTS_fclose(gv_off_context_fp); free(gv_off_context); } if (stream_type_list != NULL) { for (i = 0; i < ms->num_streams; i++) if (stream_type_list[i] != NULL) free(stream_type_list[i]); free(stream_type_list); } if (vector_length != NULL) free(vector_length); if (is_msd != NULL) free(is_msd); if (num_windows != NULL) free(num_windows); if (use_gv != NULL) free(use_gv); return !error; } /* HTS_ModelSet_get_sampling_frequency: get sampling frequency of HTS voices */ size_t HTS_ModelSet_get_sampling_frequency(HTS_ModelSet * ms) { return ms->sampling_frequency; } /* HTS_ModelSet_get_fperiod: get frame period of HTS voices */ size_t HTS_ModelSet_get_fperiod(HTS_ModelSet * ms) { return ms->frame_period; } /* HTS_ModelSet_get_fperiod: get stream option */ const char *HTS_ModelSet_get_option(HTS_ModelSet * ms, size_t stream_index) { return ms->option[stream_index]; } /* HTS_ModelSet_get_gv_flag: get GV flag */ HTS_Boolean HTS_ModelSet_get_gv_flag(HTS_ModelSet * ms, const char *string) { if (ms->gv_off_context == NULL) return TRUE; else if (HTS_Question_match(ms->gv_off_context, string) == TRUE) return FALSE; else return TRUE; } /* HTS_ModelSet_get_nstate: get number of state */ size_t HTS_ModelSet_get_nstate(HTS_ModelSet * ms) { return ms->num_states; } /* HTS_ModelSet_get_nstream: get number of stream */ size_t HTS_ModelSet_get_nstream(HTS_ModelSet * ms) { return ms->num_streams; } /* HTS_ModelSet_get_nvoices: get number of stream */ size_t HTS_ModelSet_get_nvoices(HTS_ModelSet * ms) { return ms->num_voices; } /* HTS_ModelSet_get_vector_length: get vector length */ size_t HTS_ModelSet_get_vector_length(HTS_ModelSet * ms, size_t stream_index) { return ms->stream[0][stream_index].vector_length; } /* HTS_ModelSet_is_msd: get MSD flag */ HTS_Boolean HTS_ModelSet_is_msd(HTS_ModelSet * ms, size_t stream_index) { return ms->stream[0][stream_index].is_msd; } /* HTS_ModelSet_get_window_size: get dynamic window size */ size_t HTS_ModelSet_get_window_size(HTS_ModelSet * ms, size_t stream_index) { return ms->window[stream_index].size; } /* HTS_ModelSet_get_window_left_width: get left width of dynamic window */ int HTS_ModelSet_get_window_left_width(HTS_ModelSet * ms, size_t stream_index, size_t window_index) { return ms->window[stream_index].l_width[window_index]; } /* HTS_ModelSet_get_window_right_width: get right width of dynamic window */ int HTS_ModelSet_get_window_right_width(HTS_ModelSet * ms, size_t stream_index, size_t window_index) { return ms->window[stream_index].r_width[window_index]; } /* HTS_ModelSet_get_window_coefficient: get coefficient of dynamic window */ double HTS_ModelSet_get_window_coefficient(HTS_ModelSet * ms, size_t stream_index, size_t window_index, size_t coefficient_index) { return ms->window[stream_index].coefficient[window_index][coefficient_index]; } /* HTS_ModelSet_get_window_max_width: get max width of dynamic window */ size_t HTS_ModelSet_get_window_max_width(HTS_ModelSet * ms, size_t stream_index) { return ms->window[stream_index].max_width; } /* HTS_ModelSet_use_gv: get GV flag */ HTS_Boolean HTS_ModelSet_use_gv(HTS_ModelSet * ms, size_t stream_index) { if (ms->gv[0][stream_index].vector_length != 0) return TRUE; else return FALSE; } /* HTS_Model_add_parameter: get parameter using interpolation weight */ static void HTS_Model_add_parameter(HTS_Model * model, size_t state_index, const char *string, double *mean, double *vari, double *msd, double weight) { size_t i; size_t tree_index, pdf_index; size_t len = model->vector_length * model->num_windows; HTS_Model_get_index(model, state_index, string, &tree_index, &pdf_index); for (i = 0; i < len; i++) { mean[i] += weight * model->pdf[tree_index][pdf_index][i]; vari[i] += weight * model->pdf[tree_index][pdf_index][i + len]; } if (msd != NULL && model->is_msd == TRUE) *msd += weight * model->pdf[tree_index][pdf_index][len + len]; } /* HTS_ModelSet_get_duration_index: get duration PDF & tree index */ void HTS_ModelSet_get_duration_index(HTS_ModelSet * ms, size_t voice_index, const char *string, size_t * tree_index, size_t * pdf_index) { HTS_Model_get_index(&ms->duration[voice_index], 2, string, tree_index, pdf_index); } /* HTS_ModelSet_get_duration: get duration using interpolation weight */ void HTS_ModelSet_get_duration(HTS_ModelSet * ms, const char *string, const double *iw, double *mean, double *vari) { size_t i; size_t len = ms->num_states; for (i = 0; i < len; i++) { mean[i] = 0.0; vari[i] = 0.0; } for (i = 0; i < ms->num_voices; i++) HTS_Model_add_parameter(&ms->duration[i], 2, string, mean, vari, NULL, iw[i]); } /* HTS_ModelSet_get_parameter_index: get paramter PDF & tree index */ void HTS_ModelSet_get_parameter_index(HTS_ModelSet * ms, size_t voice_index, size_t stream_index, size_t state_index, const char *string, size_t * tree_index, size_t * pdf_index) { HTS_Model_get_index(&ms->stream[voice_index][stream_index], state_index, string, tree_index, pdf_index); } /* HTS_ModelSet_get_parameter: get parameter using interpolation weight */ void HTS_ModelSet_get_parameter(HTS_ModelSet * ms, size_t stream_index, size_t state_index, const char *string, const double *iw, double *mean, double *vari, double *msd) { size_t i; size_t len = ms->stream[0][stream_index].vector_length * ms->stream[0][stream_index].num_windows; for (i = 0; i < len; i++) { mean[i] = 0.0; vari[i] = 0.0; } if (msd != NULL) *msd = 0.0; for (i = 0; i < ms->num_voices; i++) HTS_Model_add_parameter(&ms->stream[i][stream_index], state_index, string, mean, vari, msd, iw[i]); } /* HTS_ModelSet_get_gv_index: get gv PDF & tree index */ void HTS_ModelSet_get_gv_index(HTS_ModelSet * ms, size_t voice_index, size_t stream_index, const char *string, size_t * tree_index, size_t * pdf_index) { HTS_Model_get_index(&ms->gv[voice_index][stream_index], 2, string, tree_index, pdf_index); } /* HTS_ModelSet_get_gv: get GV using interpolation weight */ void HTS_ModelSet_get_gv(HTS_ModelSet * ms, size_t stream_index, const char *string, const double *iw, double *mean, double *vari) { size_t i; size_t len = ms->stream[0][stream_index].vector_length; for (i = 0; i < len; i++) { mean[i] = 0.0; vari[i] = 0.0; } for (i = 0; i < ms->num_voices; i++) HTS_Model_add_parameter(&ms->gv[i][stream_index], 2, string, mean, vari, NULL, iw[i]); } HTS_MODEL_C_END; #endif /* !HTS_MODEL_C */ festival/src/modules/hts_engine/HTS_pstream.c0000644000175000017500000005012712110442174017717 0ustar awbawb/* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ #ifndef HTS_PSTREAM_C #define HTS_PSTREAM_C #ifdef __cplusplus #define HTS_PSTREAM_C_START extern "C" { #define HTS_PSTREAM_C_END } #else #define HTS_PSTREAM_C_START #define HTS_PSTREAM_C_END #endif /* __CPLUSPLUS */ HTS_PSTREAM_C_START; #include /* for sqrt() */ /* hts_engine libraries */ #include "HTS_hidden.h" /* HTS_finv: calculate 1.0/variance function */ static double HTS_finv(const double x) { if (x >= INFTY2) return 0.0; if (x <= -INFTY2) return 0.0; if (x <= INVINF2 && x >= 0) return INFTY; if (x >= -INVINF2 && x < 0) return -INFTY; return (1.0 / x); } /* HTS_PStream_calc_wuw_and_wum: calcurate W'U^{-1}W and W'U^{-1}M */ static void HTS_PStream_calc_wuw_and_wum(HTS_PStream * pst, size_t m) { size_t t, i, j; int shift; double wu; for (t = 0; t < pst->length; t++) { /* initialize */ pst->sm.wum[t] = 0.0; for (i = 0; i < pst->width; i++) pst->sm.wuw[t][i] = 0.0; /* calc WUW & WUM */ for (i = 0; i < pst->win_size; i++) for (shift = pst->win_l_width[i]; shift <= pst->win_r_width[i]; shift++) if ((t + shift >= 0) && (t + shift < pst->length) && (pst->win_coefficient[i][-shift] != 0.0)) { wu = pst->win_coefficient[i][-shift] * pst->sm.ivar[t + shift][i * pst->vector_length + m]; pst->sm.wum[t] += wu * pst->sm.mean[t + shift][i * pst->vector_length + m]; for (j = 0; (j < pst->width) && (t + j < pst->length); j++) if (((int) j <= pst->win_r_width[i] + shift) && (pst->win_coefficient[i][j - shift] != 0.0)) pst->sm.wuw[t][j] += wu * pst->win_coefficient[i][j - shift]; } } } /* HTS_PStream_ldl_factorization: Factorize W'*U^{-1}*W to L*D*L' (L: lower triangular, D: diagonal) */ static void HTS_PStream_ldl_factorization(HTS_PStream * pst) { size_t t, i, j; for (t = 0; t < pst->length; t++) { for (i = 1; (i < pst->width) && (t >= i); i++) pst->sm.wuw[t][0] -= pst->sm.wuw[t - i][i] * pst->sm.wuw[t - i][i] * pst->sm.wuw[t - i][0]; for (i = 1; i < pst->width; i++) { for (j = 1; (i + j < pst->width) && (t >= j); j++) pst->sm.wuw[t][i] -= pst->sm.wuw[t - j][j] * pst->sm.wuw[t - j][i + j] * pst->sm.wuw[t - j][0]; pst->sm.wuw[t][i] /= pst->sm.wuw[t][0]; } } } /* HTS_PStream_forward_substitution: forward subtitution for mlpg */ static void HTS_PStream_forward_substitution(HTS_PStream * pst) { size_t t, i; for (t = 0; t < pst->length; t++) { pst->sm.g[t] = pst->sm.wum[t]; for (i = 1; (i < pst->width) && (t >= i); i++) pst->sm.g[t] -= pst->sm.wuw[t - i][i] * pst->sm.g[t - i]; } } /* HTS_PStream_backward_substitution: backward subtitution for mlpg */ static void HTS_PStream_backward_substitution(HTS_PStream * pst, size_t m) { size_t rev, t, i; for (rev = 0; rev < pst->length; rev++) { t = pst->length - 1 - rev; pst->par[t][m] = pst->sm.g[t] / pst->sm.wuw[t][0]; for (i = 1; (i < pst->width) && (t + i < pst->length); i++) pst->par[t][m] -= pst->sm.wuw[t][i] * pst->par[t + i][m]; } } /* HTS_PStream_calc_gv: subfunction for mlpg using GV */ static void HTS_PStream_calc_gv(HTS_PStream * pst, size_t m, double *mean, double *vari) { size_t t; *mean = 0.0; for (t = 0; t < pst->length; t++) if (pst->gv_switch[t]) *mean += pst->par[t][m]; *mean /= pst->gv_length; *vari = 0.0; for (t = 0; t < pst->length; t++) if (pst->gv_switch[t]) *vari += (pst->par[t][m] - *mean) * (pst->par[t][m] - *mean); *vari /= pst->gv_length; } /* HTS_PStream_conv_gv: subfunction for mlpg using GV */ static void HTS_PStream_conv_gv(HTS_PStream * pst, size_t m) { size_t t; double ratio; double mean; double vari; HTS_PStream_calc_gv(pst, m, &mean, &vari); ratio = sqrt(pst->gv_mean[m] / vari); for (t = 0; t < pst->length; t++) if (pst->gv_switch[t]) pst->par[t][m] = ratio * (pst->par[t][m] - mean) + mean; } /* HTS_PStream_calc_derivative: subfunction for mlpg using GV */ static double HTS_PStream_calc_derivative(HTS_PStream * pst, size_t m) { size_t t, i; double mean; double vari; double dv; double h; double gvobj; double hmmobj; double w = 1.0 / (pst->win_size * pst->length); HTS_PStream_calc_gv(pst, m, &mean, &vari); gvobj = -0.5 * W2 * vari * pst->gv_vari[m] * (vari - 2.0 * pst->gv_mean[m]); dv = -2.0 * pst->gv_vari[m] * (vari - pst->gv_mean[m]) / pst->length; for (t = 0; t < pst->length; t++) { pst->sm.g[t] = pst->sm.wuw[t][0] * pst->par[t][m]; for (i = 1; i < pst->width; i++) { if (t + i < pst->length) pst->sm.g[t] += pst->sm.wuw[t][i] * pst->par[t + i][m]; if (t + 1 > i) pst->sm.g[t] += pst->sm.wuw[t - i][i] * pst->par[t - i][m]; } } for (t = 0, hmmobj = 0.0; t < pst->length; t++) { hmmobj += W1 * w * pst->par[t][m] * (pst->sm.wum[t] - 0.5 * pst->sm.g[t]); h = -W1 * w * pst->sm.wuw[t][1 - 1] - W2 * 2.0 / (pst->length * pst->length) * ((pst->length - 1) * pst->gv_vari[m] * (vari - pst->gv_mean[m]) + 2.0 * pst->gv_vari[m] * (pst->par[t][m] - mean) * (pst->par[t][m] - mean)); if (pst->gv_switch[t]) pst->sm.g[t] = 1.0 / h * (W1 * w * (-pst->sm.g[t] + pst->sm.wum[t]) + W2 * dv * (pst->par[t][m] - mean)); else pst->sm.g[t] = 1.0 / h * (W1 * w * (-pst->sm.g[t] + pst->sm.wum[t])); } return (-(hmmobj + gvobj)); } /* HTS_PStream_gv_parmgen: function for mlpg using GV */ static void HTS_PStream_gv_parmgen(HTS_PStream * pst, size_t m) { size_t t, i; double step = STEPINIT; double prev = 0.0; double obj; if (pst->gv_length == 0) return; HTS_PStream_conv_gv(pst, m); if (GV_MAX_ITERATION > 0) { HTS_PStream_calc_wuw_and_wum(pst, m); for (i = 1; i <= GV_MAX_ITERATION; i++) { obj = HTS_PStream_calc_derivative(pst, m); if (i > 1) { if (obj > prev) step *= STEPDEC; if (obj < prev) step *= STEPINC; } for (t = 0; t < pst->length; t++) pst->par[t][m] += step * pst->sm.g[t]; prev = obj; } } } /* HTS_PStream_mlpg: generate sequence of speech parameter vector maximizing its output probability for given pdf sequence */ static void HTS_PStream_mlpg(HTS_PStream * pst) { size_t m; if (pst->length == 0) return; for (m = 0; m < pst->vector_length; m++) { HTS_PStream_calc_wuw_and_wum(pst, m); HTS_PStream_ldl_factorization(pst); /* LDL factorization */ HTS_PStream_forward_substitution(pst); /* forward substitution */ HTS_PStream_backward_substitution(pst, m); /* backward substitution */ if (pst->gv_length > 0) HTS_PStream_gv_parmgen(pst, m); } } /* HTS_PStreamSet_initialize: initialize parameter stream set */ void HTS_PStreamSet_initialize(HTS_PStreamSet * pss) { pss->pstream = NULL; pss->nstream = 0; pss->total_frame = 0; } /* HTS_PStreamSet_create: parameter generation using GV weight */ HTS_Boolean HTS_PStreamSet_create(HTS_PStreamSet * pss, HTS_SStreamSet * sss, double *msd_threshold, double *gv_weight) { size_t i, j, k, l, m; int shift; size_t frame, msd_frame, state; HTS_PStream *pst; HTS_Boolean not_bound; if (pss->nstream != 0) { HTS_error(1, "HTS_PstreamSet_create: HTS_PStreamSet should be clear.\n"); return FALSE; } /* initialize */ pss->nstream = HTS_SStreamSet_get_nstream(sss); pss->pstream = (HTS_PStream *) HTS_calloc(pss->nstream, sizeof(HTS_PStream)); pss->total_frame = HTS_SStreamSet_get_total_frame(sss); /* create */ for (i = 0; i < pss->nstream; i++) { pst = &pss->pstream[i]; if (HTS_SStreamSet_is_msd(sss, i)) { /* for MSD */ pst->length = 0; for (state = 0; state < HTS_SStreamSet_get_total_state(sss); state++) if (HTS_SStreamSet_get_msd(sss, i, state) > msd_threshold[i]) pst->length += HTS_SStreamSet_get_duration(sss, state); pst->msd_flag = (HTS_Boolean *) HTS_calloc(pss->total_frame, sizeof(HTS_Boolean)); for (state = 0, frame = 0; state < HTS_SStreamSet_get_total_state(sss); state++) if (HTS_SStreamSet_get_msd(sss, i, state) > msd_threshold[i]) for (j = 0; j < HTS_SStreamSet_get_duration(sss, state); j++) { pst->msd_flag[frame] = TRUE; frame++; } else for (j = 0; j < HTS_SStreamSet_get_duration(sss, state); j++) { pst->msd_flag[frame] = FALSE; frame++; } } else { /* for non MSD */ pst->length = pss->total_frame; pst->msd_flag = NULL; } pst->vector_length = HTS_SStreamSet_get_vector_length(sss, i); pst->width = HTS_SStreamSet_get_window_max_width(sss, i) * 2 + 1; /* band width of R */ pst->win_size = HTS_SStreamSet_get_window_size(sss, i); pst->sm.mean = HTS_alloc_matrix(pst->length, pst->vector_length * pst->win_size); pst->sm.ivar = HTS_alloc_matrix(pst->length, pst->vector_length * pst->win_size); pst->sm.wum = (double *) HTS_calloc(pst->length, sizeof(double)); pst->sm.wuw = HTS_alloc_matrix(pst->length, pst->width); pst->sm.g = (double *) HTS_calloc(pst->length, sizeof(double)); pst->par = HTS_alloc_matrix(pst->length, pst->vector_length); /* copy dynamic window */ pst->win_l_width = (int *) HTS_calloc(pst->win_size, sizeof(int)); pst->win_r_width = (int *) HTS_calloc(pst->win_size, sizeof(int)); pst->win_coefficient = (double **) HTS_calloc(pst->win_size, sizeof(double)); for (j = 0; j < pst->win_size; j++) { pst->win_l_width[j] = HTS_SStreamSet_get_window_left_width(sss, i, j); pst->win_r_width[j] = HTS_SStreamSet_get_window_right_width(sss, i, j); if (pst->win_l_width[j] + pst->win_r_width[j] == 0) pst->win_coefficient[j] = (double *) HTS_calloc(-2 * pst->win_l_width[j] + 1, sizeof(double)); else pst->win_coefficient[j] = (double *) HTS_calloc(-2 * pst->win_l_width[j], sizeof(double)); pst->win_coefficient[j] -= pst->win_l_width[j]; for (shift = pst->win_l_width[j]; shift <= pst->win_r_width[j]; shift++) pst->win_coefficient[j][shift] = HTS_SStreamSet_get_window_coefficient(sss, i, j, shift); } /* copy GV */ if (HTS_SStreamSet_use_gv(sss, i)) { pst->gv_mean = (double *) HTS_calloc(pst->vector_length, sizeof(double)); pst->gv_vari = (double *) HTS_calloc(pst->vector_length, sizeof(double)); for (j = 0; j < pst->vector_length; j++) { pst->gv_mean[j] = HTS_SStreamSet_get_gv_mean(sss, i, j) * gv_weight[i]; pst->gv_vari[j] = HTS_SStreamSet_get_gv_vari(sss, i, j); } pst->gv_switch = (HTS_Boolean *) HTS_calloc(pst->length, sizeof(HTS_Boolean)); if (HTS_SStreamSet_is_msd(sss, i)) { /* for MSD */ for (state = 0, frame = 0, msd_frame = 0; state < HTS_SStreamSet_get_total_state(sss); state++) for (j = 0; j < HTS_SStreamSet_get_duration(sss, state); j++, frame++) if (pst->msd_flag[frame]) pst->gv_switch[msd_frame++] = HTS_SStreamSet_get_gv_switch(sss, i, state); } else { /* for non MSD */ for (state = 0, frame = 0; state < HTS_SStreamSet_get_total_state(sss); state++) for (j = 0; j < HTS_SStreamSet_get_duration(sss, state); j++) pst->gv_switch[frame++] = HTS_SStreamSet_get_gv_switch(sss, i, state); } for (j = 0, pst->gv_length = 0; j < pst->length; j++) if (pst->gv_switch[j]) pst->gv_length++; } else { pst->gv_switch = NULL; pst->gv_length = 0; pst->gv_mean = NULL; pst->gv_vari = NULL; } /* copy pdfs */ if (HTS_SStreamSet_is_msd(sss, i)) { /* for MSD */ for (state = 0, frame = 0, msd_frame = 0; state < HTS_SStreamSet_get_total_state(sss); state++) { for (j = 0; j < HTS_SStreamSet_get_duration(sss, state); j++) { if (pst->msd_flag[frame]) { /* check current frame is MSD boundary or not */ for (k = 0; k < pst->win_size; k++) { not_bound = TRUE; for (shift = pst->win_l_width[k]; shift <= pst->win_r_width[k]; shift++) if (frame + shift < 0 || pss->total_frame <= frame + shift || !pst->msd_flag[frame + shift]) { not_bound = FALSE; break; } for (l = 0; l < pst->vector_length; l++) { m = pst->vector_length * k + l; pst->sm.mean[msd_frame][m] = HTS_SStreamSet_get_mean(sss, i, state, m); if (not_bound || k == 0) pst->sm.ivar[msd_frame][m] = HTS_finv(HTS_SStreamSet_get_vari(sss, i, state, m)); else pst->sm.ivar[msd_frame][m] = 0.0; } } msd_frame++; } frame++; } } } else { /* for non MSD */ for (state = 0, frame = 0; state < HTS_SStreamSet_get_total_state(sss); state++) { for (j = 0; j < HTS_SStreamSet_get_duration(sss, state); j++) { for (k = 0; k < pst->win_size; k++) { not_bound = TRUE; for (shift = pst->win_l_width[k]; shift <= pst->win_r_width[k]; shift++) if (frame + shift < 0 || pss->total_frame <= frame + shift) { not_bound = FALSE; break; } for (l = 0; l < pst->vector_length; l++) { m = pst->vector_length * k + l; pst->sm.mean[frame][m] = HTS_SStreamSet_get_mean(sss, i, state, m); if (not_bound || k == 0) pst->sm.ivar[frame][m] = HTS_finv(HTS_SStreamSet_get_vari(sss, i, state, m)); else pst->sm.ivar[frame][m] = 0.0; } } frame++; } } } /* parameter generation */ HTS_PStream_mlpg(pst); } return TRUE; } /* HTS_PStreamSet_get_nstream: get number of stream */ size_t HTS_PStreamSet_get_nstream(HTS_PStreamSet * pss) { return pss->nstream; } /* HTS_PStreamSet_get_vector_length: get feature length */ size_t HTS_PStreamSet_get_vector_length(HTS_PStreamSet * pss, size_t stream_index) { return pss->pstream[stream_index].vector_length; } /* HTS_PStreamSet_get_total_frame: get total number of frame */ size_t HTS_PStreamSet_get_total_frame(HTS_PStreamSet * pss) { return pss->total_frame; } /* HTS_PStreamSet_get_parameter: get parameter */ double HTS_PStreamSet_get_parameter(HTS_PStreamSet * pss, size_t stream_index, size_t frame_index, size_t vector_index) { return pss->pstream[stream_index].par[frame_index][vector_index]; } /* HTS_PStreamSet_get_parameter_vector: get parameter vector*/ double *HTS_PStreamSet_get_parameter_vector(HTS_PStreamSet * pss, size_t stream_index, size_t frame_index) { return pss->pstream[stream_index].par[frame_index]; } /* HTS_PStreamSet_get_msd_flag: get generated MSD flag per frame */ HTS_Boolean HTS_PStreamSet_get_msd_flag(HTS_PStreamSet * pss, size_t stream_index, size_t frame_index) { return pss->pstream[stream_index].msd_flag[frame_index]; } /* HTS_PStreamSet_is_msd: get MSD flag */ HTS_Boolean HTS_PStreamSet_is_msd(HTS_PStreamSet * pss, size_t stream_index) { return pss->pstream[stream_index].msd_flag ? TRUE : FALSE; } /* HTS_PStreamSet_clear: free parameter stream set */ void HTS_PStreamSet_clear(HTS_PStreamSet * pss) { size_t i, j; HTS_PStream *pstream; if (pss->pstream) { for (i = 0; i < pss->nstream; i++) { pstream = &pss->pstream[i]; HTS_free(pstream->sm.wum); HTS_free(pstream->sm.g); HTS_free_matrix(pstream->sm.wuw, pstream->length); HTS_free_matrix(pstream->sm.ivar, pstream->length); HTS_free_matrix(pstream->sm.mean, pstream->length); HTS_free_matrix(pstream->par, pstream->length); if (pstream->msd_flag) HTS_free(pstream->msd_flag); for (j = 0; j < pstream->win_size; j++) { pstream->win_coefficient[j] += pstream->win_l_width[j]; HTS_free(pstream->win_coefficient[j]); } if (pstream->gv_mean) HTS_free(pstream->gv_mean); if (pstream->gv_vari) HTS_free(pstream->gv_vari); HTS_free(pstream->win_coefficient); HTS_free(pstream->win_l_width); HTS_free(pstream->win_r_width); if (pstream->gv_switch) HTS_free(pstream->gv_switch); } HTS_free(pss->pstream); } HTS_PStreamSet_initialize(pss); } HTS_PSTREAM_C_END; #endif /* !HTS_PSTREAM_C */ festival/src/modules/hts_engine/HTS_sstream.c0000644000175000017500000004513612073131327017731 0ustar awbawb/* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ #ifndef HTS_SSTREAM_C #define HTS_SSTREAM_C #ifdef __cplusplus #define HTS_SSTREAM_C_START extern "C" { #define HTS_SSTREAM_C_END } #else #define HTS_SSTREAM_C_START #define HTS_SSTREAM_C_END #endif /* __CPLUSPLUS */ HTS_SSTREAM_C_START; #include #include /* hts_engine libraries */ #include "HTS_hidden.h" /* HTS_set_duration: set duration from state duration probability distribution */ static double HTS_set_duration(size_t * duration, double *mean, double *vari, size_t size, double frame_length) { size_t i; int j; double temp1, temp2; double rho = 0.0; size_t sum = 0; size_t target_length; /* if the frame length is not specified, only the mean vector is used */ if (frame_length == 0.0) { for (i = 0; i < size; i++) { temp1 = (mean[i] + 0.5); if (temp1 < 1.0) duration[i] = 1; else duration[i] = (size_t) temp1; sum += duration[i]; } return (double) sum; } /* get the target frame length */ if (frame_length + 0.5 < 1.0) target_length = 1; else target_length = (size_t) (frame_length + 0.5); /* check the specified duration */ if (target_length <= size) { if (target_length < size) HTS_error(-1, "HTS_set_duration: Specified frame length is too short.\n"); for (i = 0; i < size; i++) duration[i] = 1; return (double) size; } /* RHO calculation */ temp1 = 0.0; temp2 = 0.0; for (i = 0; i < size; i++) { temp1 += mean[i]; temp2 += vari[i]; } rho = ((double) target_length - temp1) / temp2; /* first estimation */ for (i = 0; i < size; i++) { temp1 = mean[i] + rho * vari[i] + 0.5; if (temp1 < 1.0) duration[i] = 1; else duration[i] = (size_t) temp1; sum += duration[i]; } /* loop estimation */ while (target_length != sum) { /* sarch flexible state and modify its duration */ if (target_length > sum) { j = -1; for (i = 0; i < size; i++) { temp2 = fabs(rho - ((double) duration[i] + 1 - mean[i]) / vari[i]); if (j < 0 || temp1 > temp2) { j = i; temp1 = temp2; } } sum++; duration[j]++; } else { j = -1; for (i = 0; i < size; i++) { if (duration[i] > 1) { temp2 = fabs(rho - ((double) duration[i] - 1 - mean[i]) / vari[i]); if (j < 0 || temp1 > temp2) { j = i; temp1 = temp2; } } } sum--; duration[j]--; } } return (double) target_length; } /* HTS_SStreamSet_initialize: initialize state stream set */ void HTS_SStreamSet_initialize(HTS_SStreamSet * sss) { sss->nstream = 0; sss->nstate = 0; sss->sstream = NULL; sss->duration = NULL; sss->total_state = 0; sss->total_frame = 0; } /* HTS_SStreamSet_create: parse label and determine state duration */ HTS_Boolean HTS_SStreamSet_create(HTS_SStreamSet * sss, HTS_ModelSet * ms, HTS_Label * label, HTS_Boolean phoneme_alignment_flag, double speed, double *duration_iw, double **parameter_iw, double **gv_iw) { size_t i, j, k; double temp; int shift; size_t state; HTS_SStream *sst; double *duration_mean, *duration_vari; double frame_length; size_t next_time; size_t next_state; /* check interpolation weights */ for (i = 0, temp = 0.0; i < HTS_ModelSet_get_nvoices(ms); i++) temp += duration_iw[i]; if (temp == 0.0) { return FALSE; } else if (temp != 1.0) { for (i = 0; i < HTS_ModelSet_get_nvoices(ms); i++) if (duration_iw[i] != 0.0) duration_iw[i] /= temp; } for (i = 0; i < HTS_ModelSet_get_nstream(ms); i++) { for (j = 0, temp = 0.0; j < HTS_ModelSet_get_nvoices(ms); j++) temp += parameter_iw[i][j]; if (temp == 0.0) { return FALSE; } else if (temp != 1.0) { for (j = 0; j < HTS_ModelSet_get_nvoices(ms); j++) if (parameter_iw[i][j] != 0.0) parameter_iw[i][j] /= temp; } if (HTS_ModelSet_use_gv(ms, i)) { for (j = 0, temp = 0.0; j < HTS_ModelSet_get_nvoices(ms); j++) temp += gv_iw[i][j]; if (temp == 0.0) return FALSE; else if (temp != 1.0) for (j = 0; j < HTS_ModelSet_get_nvoices(ms); j++) if (gv_iw[i][j] != 0.0) gv_iw[i][j] /= temp; } } /* initialize state sequence */ sss->nstate = HTS_ModelSet_get_nstate(ms); sss->nstream = HTS_ModelSet_get_nstream(ms); sss->total_frame = 0; sss->total_state = HTS_Label_get_size(label) * sss->nstate; sss->duration = (size_t *) HTS_calloc(sss->total_state, sizeof(size_t)); sss->sstream = (HTS_SStream *) HTS_calloc(sss->nstream, sizeof(HTS_SStream)); for (i = 0; i < sss->nstream; i++) { sst = &sss->sstream[i]; sst->vector_length = HTS_ModelSet_get_vector_length(ms, i); sst->mean = (double **) HTS_calloc(sss->total_state, sizeof(double *)); sst->vari = (double **) HTS_calloc(sss->total_state, sizeof(double *)); if (HTS_ModelSet_is_msd(ms, i)) sst->msd = (double *) HTS_calloc(sss->total_state, sizeof(double)); else sst->msd = NULL; for (j = 0; j < sss->total_state; j++) { sst->mean[j] = (double *) HTS_calloc(sst->vector_length * HTS_ModelSet_get_window_size(ms, i), sizeof(double)); sst->vari[j] = (double *) HTS_calloc(sst->vector_length * HTS_ModelSet_get_window_size(ms, i), sizeof(double)); } sst->gv_switch = (HTS_Boolean *) HTS_calloc(sss->total_state, sizeof(HTS_Boolean)); for (j = 0; j < sss->total_state; j++) sst->gv_switch[j] = TRUE; } /* determine state duration */ duration_mean = (double *) HTS_calloc(sss->total_state, sizeof(double)); duration_vari = (double *) HTS_calloc(sss->total_state, sizeof(double)); for (i = 0; i < HTS_Label_get_size(label); i++) HTS_ModelSet_get_duration(ms, HTS_Label_get_string(label, i), duration_iw, &duration_mean[i * sss->nstate], &duration_vari[i * sss->nstate]); if (phoneme_alignment_flag == TRUE) { /* use duration set by user */ next_time = 0; next_state = 0; state = 0; for (i = 0; i < HTS_Label_get_size(label); i++) { temp = HTS_Label_get_end_frame(label, i); if (temp >= 0) { next_time += (size_t) HTS_set_duration(&sss->duration[next_state], &duration_mean[next_state], &duration_vari[next_state], state + sss->nstate - next_state, temp - next_time); next_state = state + sss->nstate; } else if (i + 1 == HTS_Label_get_size(label)) { HTS_error(-1, "HTS_SStreamSet_create: The time of final label is not specified.\n"); HTS_set_duration(&sss->duration[next_state], &duration_mean[next_state], &duration_vari[next_state], state + sss->nstate - next_state, 0.0); } state += sss->nstate; } } else { /* determine frame length */ if (speed != 1.0) { temp = 0.0; for (i = 0; i < sss->total_state; i++) { temp += duration_mean[i]; } frame_length = temp / speed; } else { frame_length = 0.0; } /* set state duration */ HTS_set_duration(sss->duration, duration_mean, duration_vari, sss->total_state, frame_length); } HTS_free(duration_mean); HTS_free(duration_vari); /* get parameter */ for (i = 0, state = 0; i < HTS_Label_get_size(label); i++) { for (j = 2; j <= sss->nstate + 1; j++) { sss->total_frame += sss->duration[state]; for (k = 0; k < sss->nstream; k++) { sst = &sss->sstream[k]; if (sst->msd) HTS_ModelSet_get_parameter(ms, k, j, HTS_Label_get_string(label, i), parameter_iw[k], sst->mean[state], sst->vari[state], &sst->msd[state]); else HTS_ModelSet_get_parameter(ms, k, j, HTS_Label_get_string(label, i), parameter_iw[k], sst->mean[state], sst->vari[state], NULL); } state++; } } /* copy dynamic window */ for (i = 0; i < sss->nstream; i++) { sst = &sss->sstream[i]; sst->win_size = HTS_ModelSet_get_window_size(ms, i); sst->win_max_width = HTS_ModelSet_get_window_max_width(ms, i); sst->win_l_width = (int *) HTS_calloc(sst->win_size, sizeof(int)); sst->win_r_width = (int *) HTS_calloc(sst->win_size, sizeof(int)); sst->win_coefficient = (double **) HTS_calloc(sst->win_size, sizeof(double)); for (j = 0; j < sst->win_size; j++) { sst->win_l_width[j] = HTS_ModelSet_get_window_left_width(ms, i, j); sst->win_r_width[j] = HTS_ModelSet_get_window_right_width(ms, i, j); if (sst->win_l_width[j] + sst->win_r_width[j] == 0) sst->win_coefficient[j] = (double *) HTS_calloc(-2 * sst->win_l_width[j] + 1, sizeof(double)); else sst->win_coefficient[j] = (double *) HTS_calloc(-2 * sst->win_l_width[j], sizeof(double)); sst->win_coefficient[j] -= sst->win_l_width[j]; for (shift = sst->win_l_width[j]; shift <= sst->win_r_width[j]; shift++) sst->win_coefficient[j][shift] = HTS_ModelSet_get_window_coefficient(ms, i, j, shift); } } /* determine GV */ for (i = 0; i < sss->nstream; i++) { sst = &sss->sstream[i]; if (HTS_ModelSet_use_gv(ms, i)) { sst->gv_mean = (double *) HTS_calloc(sst->vector_length, sizeof(double)); sst->gv_vari = (double *) HTS_calloc(sst->vector_length, sizeof(double)); HTS_ModelSet_get_gv(ms, i, HTS_Label_get_string(label, 0), gv_iw[i], sst->gv_mean, sst->gv_vari); } else { sst->gv_mean = NULL; sst->gv_vari = NULL; } } for (i = 0; i < HTS_Label_get_size(label); i++) if (HTS_ModelSet_get_gv_flag(ms, HTS_Label_get_string(label, i)) == FALSE) for (j = 0; j < sss->nstream; j++) if (HTS_ModelSet_use_gv(ms, j) == TRUE) for (k = 0; k < sss->nstate; k++) sss->sstream[j].gv_switch[i * sss->nstate + k] = FALSE; return TRUE; } /* HTS_SStreamSet_get_nstream: get number of stream */ size_t HTS_SStreamSet_get_nstream(HTS_SStreamSet * sss) { return sss->nstream; } /* HTS_SStreamSet_get_vector_length: get vector length */ size_t HTS_SStreamSet_get_vector_length(HTS_SStreamSet * sss, size_t stream_index) { return sss->sstream[stream_index].vector_length; } /* HTS_SStreamSet_is_msd: get MSD flag */ HTS_Boolean HTS_SStreamSet_is_msd(HTS_SStreamSet * sss, size_t stream_index) { return sss->sstream[stream_index].msd ? TRUE : FALSE; } /* HTS_SStreamSet_get_total_state: get total number of state */ size_t HTS_SStreamSet_get_total_state(HTS_SStreamSet * sss) { return sss->total_state; } /* HTS_SStreamSet_get_total_frame: get total number of frame */ size_t HTS_SStreamSet_get_total_frame(HTS_SStreamSet * sss) { return sss->total_frame; } /* HTS_SStreamSet_get_msd: get MSD parameter */ double HTS_SStreamSet_get_msd(HTS_SStreamSet * sss, size_t stream_index, size_t state_index) { return sss->sstream[stream_index].msd[state_index]; } /* HTS_SStreamSet_window_size: get dynamic window size */ size_t HTS_SStreamSet_get_window_size(HTS_SStreamSet * sss, size_t stream_index) { return sss->sstream[stream_index].win_size; } /* HTS_SStreamSet_get_window_left_width: get left width of dynamic window */ int HTS_SStreamSet_get_window_left_width(HTS_SStreamSet * sss, size_t stream_index, size_t window_index) { return sss->sstream[stream_index].win_l_width[window_index]; } /* HTS_SStreamSet_get_winodow_right_width: get right width of dynamic window */ int HTS_SStreamSet_get_window_right_width(HTS_SStreamSet * sss, size_t stream_index, size_t window_index) { return sss->sstream[stream_index].win_r_width[window_index]; } /* HTS_SStreamSet_get_window_coefficient: get coefficient of dynamic window */ double HTS_SStreamSet_get_window_coefficient(HTS_SStreamSet * sss, size_t stream_index, size_t window_index, int coefficient_index) { return sss->sstream[stream_index].win_coefficient[window_index][coefficient_index]; } /* HTS_SStreamSet_get_window_max_width: get max width of dynamic window */ size_t HTS_SStreamSet_get_window_max_width(HTS_SStreamSet * sss, size_t stream_index) { return sss->sstream[stream_index].win_max_width; } /* HTS_SStreamSet_use_gv: get GV flag */ HTS_Boolean HTS_SStreamSet_use_gv(HTS_SStreamSet * sss, size_t stream_index) { return sss->sstream[stream_index].gv_mean ? TRUE : FALSE; } /* HTS_SStreamSet_get_duration: get state duration */ size_t HTS_SStreamSet_get_duration(HTS_SStreamSet * sss, size_t state_index) { return sss->duration[state_index]; } /* HTS_SStreamSet_get_mean: get mean parameter */ double HTS_SStreamSet_get_mean(HTS_SStreamSet * sss, size_t stream_index, size_t state_index, size_t vector_index) { return sss->sstream[stream_index].mean[state_index][vector_index]; } /* HTS_SStreamSet_set_mean: set mean parameter */ void HTS_SStreamSet_set_mean(HTS_SStreamSet * sss, size_t stream_index, size_t state_index, size_t vector_index, double f) { sss->sstream[stream_index].mean[state_index][vector_index] = f; } /* HTS_SStreamSet_get_vari: get variance parameter */ double HTS_SStreamSet_get_vari(HTS_SStreamSet * sss, size_t stream_index, size_t state_index, size_t vector_index) { return sss->sstream[stream_index].vari[state_index][vector_index]; } /* HTS_SStreamSet_set_vari: set variance parameter */ void HTS_SStreamSet_set_vari(HTS_SStreamSet * sss, size_t stream_index, size_t state_index, size_t vector_index, double f) { sss->sstream[stream_index].vari[state_index][vector_index] = f; } /* HTS_SStreamSet_get_gv_mean: get GV mean parameter */ double HTS_SStreamSet_get_gv_mean(HTS_SStreamSet * sss, size_t stream_index, size_t vector_index) { return sss->sstream[stream_index].gv_mean[vector_index]; } /* HTS_SStreamSet_get_gv_mean: get GV variance parameter */ double HTS_SStreamSet_get_gv_vari(HTS_SStreamSet * sss, size_t stream_index, size_t vector_index) { return sss->sstream[stream_index].gv_vari[vector_index]; } /* HTS_SStreamSet_set_gv_switch: set GV switch */ void HTS_SStreamSet_set_gv_switch(HTS_SStreamSet * sss, size_t stream_index, size_t state_index, HTS_Boolean i) { sss->sstream[stream_index].gv_switch[state_index] = i; } /* HTS_SStreamSet_get_gv_switch: get GV switch */ HTS_Boolean HTS_SStreamSet_get_gv_switch(HTS_SStreamSet * sss, size_t stream_index, size_t state_index) { return sss->sstream[stream_index].gv_switch[state_index]; } /* HTS_SStreamSet_clear: free state stream set */ void HTS_SStreamSet_clear(HTS_SStreamSet * sss) { size_t i, j; HTS_SStream *sst; if (sss->sstream) { for (i = 0; i < sss->nstream; i++) { sst = &sss->sstream[i]; for (j = 0; j < sss->total_state; j++) { HTS_free(sst->mean[j]); HTS_free(sst->vari[j]); } if (sst->msd) HTS_free(sst->msd); HTS_free(sst->mean); HTS_free(sst->vari); for (j = 0; j < sst->win_size; j++) { sst->win_coefficient[j] += sst->win_l_width[j]; HTS_free(sst->win_coefficient[j]); } HTS_free(sst->win_coefficient); HTS_free(sst->win_l_width); HTS_free(sst->win_r_width); if (sst->gv_mean) HTS_free(sst->gv_mean); if (sst->gv_vari) HTS_free(sst->gv_vari); HTS_free(sst->gv_switch); } HTS_free(sss->sstream); } if (sss->duration) HTS_free(sss->duration); HTS_SStreamSet_initialize(sss); } HTS_SSTREAM_C_END; #endif /* !HTS_SSTREAM_C */ festival/src/modules/hts_engine/HTS_vocoder.c0000644000175000017500000006763412110442174017720 0ustar awbawb/* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ #ifndef HTS_VOCODER_C #define HTS_VOCODER_C #ifdef __cplusplus #define HTS_VOCODER_C_START extern "C" { #define HTS_VOCODER_C_END } #else #define HTS_VOCODER_C_START #define HTS_VOCODER_C_END #endif /* __CPLUSPLUS */ HTS_VOCODER_C_START; #include /* for sqrt(),log(),exp(),pow(),cos() */ /* hts_engine libraries */ #include "HTS_hidden.h" static const double HTS_pade[21] = { 1.00000000000, 1.00000000000, 0.00000000000, 1.00000000000, 0.00000000000, 0.00000000000, 1.00000000000, 0.00000000000, 0.00000000000, 0.00000000000, 1.00000000000, 0.49992730000, 0.10670050000, 0.01170221000, 0.00056562790, 1.00000000000, 0.49993910000, 0.11070980000, 0.01369984000, 0.00095648530, 0.00003041721 }; /* HTS_movem: move memory */ static void HTS_movem(double *a, double *b, const int nitem) { long i = (long) nitem; if (a > b) while (i--) *b++ = *a++; else { a += i; b += i; while (i--) *--b = *--a; } } /* HTS_mlsafir: sub functions for MLSA filter */ static double HTS_mlsafir(const double x, const double *b, const int m, const double a, const double aa, double *d) { double y = 0.0; int i; d[0] = x; d[1] = aa * d[0] + a * d[1]; for (i = 2; i <= m; i++) d[i] += a * (d[i + 1] - d[i - 1]); for (i = 2; i <= m; i++) y += d[i] * b[i]; for (i = m + 1; i > 1; i--) d[i] = d[i - 1]; return (y); } /* HTS_mlsadf1: sub functions for MLSA filter */ static double HTS_mlsadf1(double x, const double *b, const int m, const double a, const double aa, const int pd, double *d, const double *ppade) { double v, out = 0.0, *pt; int i; pt = &d[pd + 1]; for (i = pd; i >= 1; i--) { d[i] = aa * pt[i - 1] + a * d[i]; pt[i] = d[i] * b[1]; v = pt[i] * ppade[i]; x += (1 & i) ? v : -v; out += v; } pt[0] = x; out += x; return (out); } /* HTS_mlsadf2: sub functions for MLSA filter */ static double HTS_mlsadf2(double x, const double *b, const int m, const double a, const double aa, const int pd, double *d, const double *ppade) { double v, out = 0.0, *pt; int i; pt = &d[pd * (m + 2)]; for (i = pd; i >= 1; i--) { pt[i] = HTS_mlsafir(pt[i - 1], b, m, a, aa, &d[(i - 1) * (m + 2)]); v = pt[i] * ppade[i]; x += (1 & i) ? v : -v; out += v; } pt[0] = x; out += x; return (out); } /* HTS_mlsadf: functions for MLSA filter */ static double HTS_mlsadf(double x, const double *b, const int m, const double a, const int pd, double *d) { const double aa = 1 - a * a; const double *ppade = &(HTS_pade[pd * (pd + 1) / 2]); x = HTS_mlsadf1(x, b, m, a, aa, pd, d, ppade); x = HTS_mlsadf2(x, b, m, a, aa, pd, &d[2 * (pd + 1)], ppade); return (x); } /* HTS_rnd: functions for random noise generation */ static double HTS_rnd(unsigned long *next) { double r; *next = *next * 1103515245L + 12345; r = (*next / 65536L) % 32768L; return (r / RANDMAX); } /* HTS_nrandom: functions for gaussian random noise generation */ static double HTS_nrandom(HTS_Vocoder * v) { if (v->sw == 0) { v->sw = 1; do { v->r1 = 2 * HTS_rnd(&v->next) - 1; v->r2 = 2 * HTS_rnd(&v->next) - 1; v->s = v->r1 * v->r1 + v->r2 * v->r2; } while (v->s > 1 || v->s == 0); v->s = sqrt(-2 * log(v->s) / v->s); return (v->r1 * v->s); } else { v->sw = 0; return (v->r2 * v->s); } } /* HTS_mceq: function for M-sequence random noise generation */ static int HTS_mseq(HTS_Vocoder * v) { int x0, x28; v->x >>= 1; if (v->x & B0) x0 = 1; else x0 = -1; if (v->x & B28) x28 = 1; else x28 = -1; if (x0 + x28) v->x &= B31_; else v->x |= B31; return (x0); } /* HTS_mc2b: transform mel-cepstrum to MLSA digital fillter coefficients */ static void HTS_mc2b(double *mc, double *b, int m, const double a) { if (mc != b) { if (a != 0.0) { b[m] = mc[m]; for (m--; m >= 0; m--) b[m] = mc[m] - a * b[m + 1]; } else HTS_movem(mc, b, m + 1); } else if (a != 0.0) for (m--; m >= 0; m--) b[m] -= a * b[m + 1]; } /* HTS_b2bc: transform MLSA digital filter coefficients to mel-cepstrum */ static void HTS_b2mc(const double *b, double *mc, int m, const double a) { double d, o; d = mc[m] = b[m]; for (m--; m >= 0; m--) { o = b[m] + a * d; d = b[m]; mc[m] = o; } } /* HTS_freqt: frequency transformation */ static void HTS_freqt(HTS_Vocoder * v, const double *c1, const int m1, double *c2, const int m2, const double a) { int i, j; const double b = 1 - a * a; double *g; if (m2 > v->freqt_size) { if (v->freqt_buff != NULL) HTS_free(v->freqt_buff); v->freqt_buff = (double *) HTS_calloc(m2 + m2 + 2, sizeof(double)); v->freqt_size = m2; } g = v->freqt_buff + v->freqt_size + 1; for (i = 0; i < m2 + 1; i++) g[i] = 0.0; for (i = -m1; i <= 0; i++) { if (0 <= m2) g[0] = c1[-i] + a * (v->freqt_buff[0] = g[0]); if (1 <= m2) g[1] = b * v->freqt_buff[0] + a * (v->freqt_buff[1] = g[1]); for (j = 2; j <= m2; j++) g[j] = v->freqt_buff[j - 1] + a * ((v->freqt_buff[j] = g[j]) - g[j - 1]); } HTS_movem(g, c2, m2 + 1); } /* HTS_c2ir: The minimum phase impulse response is evaluated from the minimum phase cepstrum */ static void HTS_c2ir(const double *c, const int nc, double *h, const int leng) { int n, k, upl; double d; h[0] = exp(c[0]); for (n = 1; n < leng; n++) { d = 0; upl = (n >= nc) ? nc - 1 : n; for (k = 1; k <= upl; k++) d += k * c[k] * h[n - k]; h[n] = d / n; } } /* HTS_b2en: calculate frame energy */ static double HTS_b2en(HTS_Vocoder * v, const double *b, const int m, const double a) { int i; double en = 0.0; double *cep; double *ir; if (v->spectrum2en_size < m) { if (v->spectrum2en_buff != NULL) HTS_free(v->spectrum2en_buff); v->spectrum2en_buff = (double *) HTS_calloc((m + 1) + 2 * IRLENG, sizeof(double)); v->spectrum2en_size = m; } cep = v->spectrum2en_buff + m + 1; ir = cep + IRLENG; HTS_b2mc(b, v->spectrum2en_buff, m, a); HTS_freqt(v, v->spectrum2en_buff, m, cep, IRLENG - 1, -a); HTS_c2ir(cep, IRLENG, ir, IRLENG); for (i = 0; i < IRLENG; i++) en += ir[i] * ir[i]; return (en); } /* HTS_ignorm: inverse gain normalization */ static void HTS_ignorm(double *c1, double *c2, int m, const double g) { double k; if (g != 0.0) { k = pow(c1[0], g); for (; m >= 1; m--) c2[m] = k * c1[m]; c2[0] = (k - 1.0) / g; } else { HTS_movem(&c1[1], &c2[1], m); c2[0] = log(c1[0]); } } /* HTS_gnorm: gain normalization */ static void HTS_gnorm(double *c1, double *c2, int m, const double g) { double k; if (g != 0.0) { k = 1.0 + g * c1[0]; for (; m >= 1; m--) c2[m] = c1[m] / k; c2[0] = pow(k, 1.0 / g); } else { HTS_movem(&c1[1], &c2[1], m); c2[0] = exp(c1[0]); } } /* HTS_lsp2lpc: transform LSP to LPC */ static void HTS_lsp2lpc(HTS_Vocoder * v, double *lsp, double *a, const int m) { int i, k, mh1, mh2, flag_odd; double xx, xf, xff; double *p, *q; double *a0, *a1, *a2, *b0, *b1, *b2; flag_odd = 0; if (m % 2 == 0) mh1 = mh2 = m / 2; else { mh1 = (m + 1) / 2; mh2 = (m - 1) / 2; flag_odd = 1; } if (m > v->lsp2lpc_size) { if (v->lsp2lpc_buff != NULL) HTS_free(v->lsp2lpc_buff); v->lsp2lpc_buff = (double *) HTS_calloc(5 * m + 6, sizeof(double)); v->lsp2lpc_size = m; } p = v->lsp2lpc_buff + m; q = p + mh1; a0 = q + mh2; a1 = a0 + (mh1 + 1); a2 = a1 + (mh1 + 1); b0 = a2 + (mh1 + 1); b1 = b0 + (mh2 + 1); b2 = b1 + (mh2 + 1); HTS_movem(lsp, v->lsp2lpc_buff, m); for (i = 0; i < mh1 + 1; i++) a0[i] = 0.0; for (i = 0; i < mh1 + 1; i++) a1[i] = 0.0; for (i = 0; i < mh1 + 1; i++) a2[i] = 0.0; for (i = 0; i < mh2 + 1; i++) b0[i] = 0.0; for (i = 0; i < mh2 + 1; i++) b1[i] = 0.0; for (i = 0; i < mh2 + 1; i++) b2[i] = 0.0; /* lsp filter parameters */ for (i = k = 0; i < mh1; i++, k += 2) p[i] = -2.0 * cos(v->lsp2lpc_buff[k]); for (i = k = 0; i < mh2; i++, k += 2) q[i] = -2.0 * cos(v->lsp2lpc_buff[k + 1]); /* impulse response of analysis filter */ xx = 1.0; xf = xff = 0.0; for (k = 0; k <= m; k++) { if (flag_odd) { a0[0] = xx; b0[0] = xx - xff; xff = xf; xf = xx; } else { a0[0] = xx + xf; b0[0] = xx - xf; xf = xx; } for (i = 0; i < mh1; i++) { a0[i + 1] = a0[i] + p[i] * a1[i] + a2[i]; a2[i] = a1[i]; a1[i] = a0[i]; } for (i = 0; i < mh2; i++) { b0[i + 1] = b0[i] + q[i] * b1[i] + b2[i]; b2[i] = b1[i]; b1[i] = b0[i]; } if (k != 0) a[k - 1] = -0.5 * (a0[mh1] + b0[mh2]); xx = 0.0; } for (i = m - 1; i >= 0; i--) a[i + 1] = -a[i]; a[0] = 1.0; } /* HTS_gc2gc: generalized cepstral transformation */ static void HTS_gc2gc(HTS_Vocoder * v, double *c1, const int m1, const double g1, double *c2, const int m2, const double g2) { int i, min, k, mk; double ss1, ss2, cc; if (m1 > v->gc2gc_size) { if (v->gc2gc_buff != NULL) HTS_free(v->gc2gc_buff); v->gc2gc_buff = (double *) HTS_calloc(m1 + 1, sizeof(double)); v->gc2gc_size = m1; } HTS_movem(c1, v->gc2gc_buff, m1 + 1); c2[0] = v->gc2gc_buff[0]; for (i = 1; i <= m2; i++) { ss1 = ss2 = 0.0; min = m1 < i ? m1 : i - 1; for (k = 1; k <= min; k++) { mk = i - k; cc = v->gc2gc_buff[k] * c2[mk]; ss2 += k * cc; ss1 += mk * cc; } if (i <= m1) c2[i] = v->gc2gc_buff[i] + (g2 * ss2 - g1 * ss1) / i; else c2[i] = (g2 * ss2 - g1 * ss1) / i; } } /* HTS_mgc2mgc: frequency and generalized cepstral transformation */ static void HTS_mgc2mgc(HTS_Vocoder * v, double *c1, const int m1, const double a1, const double g1, double *c2, const int m2, const double a2, const double g2) { double a; if (a1 == a2) { HTS_gnorm(c1, c1, m1, g1); HTS_gc2gc(v, c1, m1, g1, c2, m2, g2); HTS_ignorm(c2, c2, m2, g2); } else { a = (a2 - a1) / (1 - a1 * a2); HTS_freqt(v, c1, m1, c2, m2, a); HTS_gnorm(c2, c2, m2, g1); HTS_gc2gc(v, c2, m2, g1, c2, m2, g2); HTS_ignorm(c2, c2, m2, g2); } } /* HTS_lsp2mgc: transform LSP to MGC */ static void HTS_lsp2mgc(HTS_Vocoder * v, double *lsp, double *mgc, const int m, const double alpha) { int i; /* lsp2lpc */ HTS_lsp2lpc(v, lsp + 1, mgc, m); if (v->use_log_gain) mgc[0] = exp(lsp[0]); else mgc[0] = lsp[0]; /* mgc2mgc */ if (NORMFLG1) HTS_ignorm(mgc, mgc, m, v->gamma); else if (MULGFLG1) mgc[0] = (1.0 - mgc[0]) * ((double) v->stage); if (MULGFLG1) for (i = m; i >= 1; i--) mgc[i] *= -((double) v->stage); HTS_mgc2mgc(v, mgc, m, alpha, v->gamma, mgc, m, alpha, v->gamma); if (NORMFLG2) HTS_gnorm(mgc, mgc, m, v->gamma); else if (MULGFLG2) mgc[0] = mgc[0] * v->gamma + 1.0; if (MULGFLG2) for (i = m; i >= 1; i--) mgc[i] *= v->gamma; } /* HTS_mglsadff: sub functions for MGLSA filter */ static double HTS_mglsadff(double x, const double *b, const int m, const double a, double *d) { int i; double y; y = d[0] * b[1]; for (i = 1; i < m; i++) { d[i] += a * (d[i + 1] - d[i - 1]); y += d[i] * b[i + 1]; } x -= y; for (i = m; i > 0; i--) d[i] = d[i - 1]; d[0] = a * d[0] + (1 - a * a) * x; return x; } /* HTS_mglsadf: sub functions for MGLSA filter */ static double HTS_mglsadf(double x, const double *b, const int m, const double a, const int n, double *d) { int i; for (i = 0; i < n; i++) x = HTS_mglsadff(x, b, m, a, &d[i * (m + 1)]); return x; } /* THS_check_lsp_stability: check LSP stability */ static void HTS_check_lsp_stability(double *lsp, size_t m) { size_t i, j; double tmp; double min = (CHECK_LSP_STABILITY_MIN * PI) / (m + 1); HTS_Boolean find; for (i = 0; i < CHECK_LSP_STABILITY_NUM; i++) { find = FALSE; for (j = 1; j < m; j++) { tmp = lsp[j + 1] - lsp[j]; if (tmp < min) { lsp[j] -= 0.5 * (min - tmp); lsp[j + 1] += 0.5 * (min - tmp); find = TRUE; } } if (lsp[1] < min) { lsp[1] = min; find = TRUE; } if (lsp[m] > PI - min) { lsp[m] = PI - min; find = TRUE; } if (find == FALSE) break; } } /* HTS_lsp2en: calculate frame energy */ static double HTS_lsp2en(HTS_Vocoder * v, double *lsp, size_t m, double alpha) { size_t i; double en = 0.0; double *buff; if (v->spectrum2en_size < m) { if (v->spectrum2en_buff != NULL) HTS_free(v->spectrum2en_buff); v->spectrum2en_buff = (double *) HTS_calloc(m + 1 + IRLENG, sizeof(double)); v->spectrum2en_size = m; } buff = v->spectrum2en_buff + m + 1; /* lsp2lpc */ HTS_lsp2lpc(v, lsp + 1, v->spectrum2en_buff, m); if (v->use_log_gain) v->spectrum2en_buff[0] = exp(lsp[0]); else v->spectrum2en_buff[0] = lsp[0]; /* mgc2mgc */ if (NORMFLG1) HTS_ignorm(v->spectrum2en_buff, v->spectrum2en_buff, m, v->gamma); else if (MULGFLG1) v->spectrum2en_buff[0] = (1.0 - v->spectrum2en_buff[0]) * ((double) v->stage); if (MULGFLG1) for (i = 1; i <= m; i++) v->spectrum2en_buff[i] *= -((double) v->stage); HTS_mgc2mgc(v, v->spectrum2en_buff, m, alpha, v->gamma, buff, IRLENG - 1, 0.0, 1); for (i = 0; i < IRLENG; i++) en += buff[i] * buff[i]; return en; } /* HTS_white_noise: return white noise */ static double HTS_white_noise(HTS_Vocoder * v) { if (v->gauss) return (double) HTS_nrandom(v); else return (double) HTS_mseq(v); } /* HTS_ping_pulse: ping pulse using low-pass filter */ static void HTS_ping_pulse(HTS_Vocoder * v, const int ping_place, const double p, const int nlpf, const double *lpf) { int i, j; const double power = sqrt(p); for (i = ping_place - nlpf, j = 0; i <= ping_place + nlpf; i++, j++) if (0 <= i && i < v->pulse_size) v->pulse_buff[i] += power * lpf[j]; } /* HTS_ping_noise: ping noise using low-pass filter */ static void HTS_ping_noise(HTS_Vocoder * v, const int ping_place, const int nlpf, const double *lpf) { int i, j; const double power = HTS_white_noise(v); for (i = ping_place - nlpf, j = 0; i <= ping_place + nlpf; i++, j++) if (0 <= i && i < v->pulse_size) { if (j == nlpf) v->pulse_buff[i] += power * (1.0 - lpf[j]); else v->pulse_buff[i] += power * (0.0 - lpf[j]); } } /* HTS_Vocoder_initialize_excitation: initialize excitation */ static void HTS_Vocoder_initialize_excitation(HTS_Vocoder * v, size_t nlpf) { size_t i; if (nlpf > 0) { v->pulse_size = (v->fprd > v->rate / MIN_F0) ? v->fprd + nlpf : v->rate / MIN_F0 + nlpf; v->pulse_buff = (double *) HTS_calloc(v->pulse_size, sizeof(double)); for (i = 0; i < v->pulse_size; i++) v->pulse_buff[i] = 0.0; } v->p1 = 0.0; v->pc = 0.0; v->p = 0.0; v->inc = 0.0; } /* HTS_Vocoder_start_excitation: start excitation of each frame */ static void HTS_Vocoder_start_excitation(HTS_Vocoder * v, const double pitch, const int nlpf) { if (v->p1 != 0.0 && pitch != 0.0) v->inc = (pitch - v->p1) * IPERIOD / v->fprd; else { v->inc = 0.0; if (nlpf <= 0) { v->p1 = 0.0; v->pc = pitch; } } v->p = pitch; } /* HTS_Vocoder_get_excitation: get excitation of each sample */ static double HTS_Vocoder_get_excitation(HTS_Vocoder * v, const int fprd_index, const int iprd_index, const int nlpf, const double *lpf) { double x; int i, j; if (nlpf > 0) { if (fprd_index == 0) { if (v->p1 == 0.0) { for (i = 0; i < v->fprd; i++) v->pulse_buff[i] += HTS_white_noise(v); if (v->p != 0.0) { HTS_ping_pulse(v, v->fprd + nlpf, v->p, nlpf, lpf); v->pc = v->fprd + nlpf; } } else if (v->p == 0.0) { for (i = nlpf; i < v->fprd; i++) v->pulse_buff[i] += HTS_white_noise(v); } else { for (i = 0, j = (IPERIOD + 1) / 2; i < v->fprd; i++) { if ((v->pc + v->p1) <= i) { HTS_ping_pulse(v, i, v->p1, nlpf, lpf); v->pc += v->p1; } if (!--j) { v->p1 += v->inc; j = IPERIOD; } } for (i = v->fprd; i < v->fprd + nlpf; i++) { if ((v->pc + v->p) <= i) { HTS_ping_pulse(v, i, v->p, nlpf, lpf); v->pc += v->p; } } for (i = nlpf; i < v->fprd + nlpf; i++) HTS_ping_noise(v, i, nlpf, lpf); } } x = v->pulse_buff[fprd_index]; } else { if (v->p1 == 0.0) x = HTS_white_noise(v); else { if ((v->pc += 1.0) >= v->p1) { x = sqrt(v->p1); v->pc -= v->p1; } else x = 0.0; } if (iprd_index <= 1) v->p1 += v->inc; } return x; } /* HTS_Vocoder_end_excitation: end excitation of each frame */ static void HTS_Vocoder_end_excitation(HTS_Vocoder * v, const int nlpf) { int i; if (nlpf > 0) { v->pc -= v->fprd; for (i = 0; i < v->pulse_size; i++) if (i < v->pulse_size - v->fprd) v->pulse_buff[i] = v->pulse_buff[i + v->fprd]; else v->pulse_buff[i] = 0.0; } v->p1 = v->p; } /* HTS_Vocoder_postfilter_mcp: postfilter for MCP */ static void HTS_Vocoder_postfilter_mcp(HTS_Vocoder * v, double *mcp, const int m, double alpha, double beta) { double e1, e2; int k; if (beta > 0.0 && m > 1) { if (v->postfilter_size < m) { if (v->postfilter_buff != NULL) HTS_free(v->postfilter_buff); v->postfilter_buff = (double *) HTS_calloc(m + 1, sizeof(double)); v->postfilter_size = m; } HTS_mc2b(mcp, v->postfilter_buff, m, alpha); e1 = HTS_b2en(v, v->postfilter_buff, m, alpha); v->postfilter_buff[1] -= beta * alpha * mcp[2]; for (k = 2; k <= m; k++) v->postfilter_buff[k] *= (1.0 + beta); e2 = HTS_b2en(v, v->postfilter_buff, m, alpha); v->postfilter_buff[0] += log(e1 / e2) / 2; HTS_b2mc(v->postfilter_buff, mcp, m, alpha); } } /* HTS_Vocoder_postfilter_lsp: postfilter for LSP */ static void HTS_Vocoder_postfilter_lsp(HTS_Vocoder * v, double *lsp, size_t m, double alpha, double beta) { double e1, e2; size_t i; double d1, d2; if (beta > 0.0 && m > 1) { if (v->postfilter_size < m) { if (v->postfilter_buff != NULL) HTS_free(v->postfilter_buff); v->postfilter_buff = (double *) HTS_calloc(m + 1, sizeof(double)); v->postfilter_size = m; } e1 = HTS_lsp2en(v, lsp, m, alpha); /* postfiltering */ for (i = 0; i <= m; i++) { if (i > 1 && i < m) { d1 = beta * (lsp[i + 1] - lsp[i]); d2 = beta * (lsp[i] - lsp[i - 1]); v->postfilter_buff[i] = lsp[i - 1] + d2 + (d2 * d2 * ((lsp[i + 1] - lsp[i - 1]) - (d1 + d2))) / ((d2 * d2) + (d1 * d1)); } else { v->postfilter_buff[i] = lsp[i]; } } HTS_movem(v->postfilter_buff, lsp, m + 1); e2 = HTS_lsp2en(v, lsp, m, alpha); if (e1 != e2) { if (v->use_log_gain) lsp[0] += 0.5 * log(e1 / e2); else lsp[0] *= sqrt(e1 / e2); } } } /* HTS_Vocoder_initialize: initialize vocoder */ void HTS_Vocoder_initialize(HTS_Vocoder * v, size_t m, size_t stage, HTS_Boolean use_log_gain, size_t rate, size_t fperiod) { /* set parameter */ v->stage = stage; if (stage != 0) v->gamma = -1.0 / v->stage; else v->gamma = 0.0; v->use_log_gain = use_log_gain; v->fprd = fperiod; v->next = SEED; v->gauss = GAUSS; v->rate = rate; v->p1 = -1.0; v->sw = 0; v->x = 0x55555555; /* init buffer */ v->freqt_buff = NULL; v->freqt_size = 0; v->gc2gc_buff = NULL; v->gc2gc_size = 0; v->lsp2lpc_buff = NULL; v->lsp2lpc_size = 0; v->postfilter_buff = NULL; v->postfilter_size = 0; v->spectrum2en_buff = NULL; v->spectrum2en_size = 0; if (v->stage == 0) { /* for MCP */ v->c = (double *) HTS_calloc(m * (3 + PADEORDER) + 5 * PADEORDER + 6, sizeof(double)); v->cc = v->c + m + 1; v->cinc = v->cc + m + 1; v->d1 = v->cinc + m + 1; } else { /* for LSP */ v->c = (double *) HTS_calloc((m + 1) * (v->stage + 3), sizeof(double)); v->cc = v->c + m + 1; v->cinc = v->cc + m + 1; v->d1 = v->cinc + m + 1; } v->pulse_buff = NULL; v->pulse_size = 0; } /* HTS_Vocoder_synthesize: pulse/noise excitation and MLSA/MGLSA filster based waveform synthesis */ void HTS_Vocoder_synthesize(HTS_Vocoder * v, size_t m, double lf0, double *spectrum, size_t nlpf, double *lpf, double alpha, double beta, double volume, double *rawdata, HTS_Audio * audio) { double x; int i, j; short xs; int rawidx = 0; double p; /* lf0 -> pitch */ if (lf0 == LZERO) p = 0.0; else if (lf0 <= MIN_LF0) p = v->rate / MIN_F0; else if (lf0 >= MAX_LF0) p = v->rate / MAX_F0; else p = v->rate / exp(lf0); /* first time */ if (v->p1 < 0.0) { HTS_Vocoder_initialize_excitation(v, nlpf); if (v->stage == 0) { /* for MCP */ HTS_mc2b(spectrum, v->c, m, alpha); } else { /* for LSP */ if (v->use_log_gain) v->c[0] = LZERO; else v->c[0] = ZERO; for (i = 1; i <= m; i++) v->c[i] = i * PI / (m + 1); HTS_lsp2mgc(v, v->c, v->c, m, alpha); HTS_mc2b(v->c, v->c, m, alpha); HTS_gnorm(v->c, v->c, m, v->gamma); for (i = 1; i <= m; i++) v->c[i] *= v->gamma; } } HTS_Vocoder_start_excitation(v, p, nlpf); if (v->stage == 0) { /* for MCP */ HTS_Vocoder_postfilter_mcp(v, spectrum, m, alpha, beta); HTS_mc2b(spectrum, v->cc, m, alpha); for (i = 0; i <= m; i++) v->cinc[i] = (v->cc[i] - v->c[i]) * IPERIOD / v->fprd; } else { /* for LSP */ HTS_Vocoder_postfilter_lsp(v, spectrum, m, alpha, beta); HTS_check_lsp_stability(spectrum, m); HTS_lsp2mgc(v, spectrum, v->cc, m, alpha); HTS_mc2b(v->cc, v->cc, m, alpha); HTS_gnorm(v->cc, v->cc, m, v->gamma); for (i = 1; i <= m; i++) v->cc[i] *= v->gamma; for (i = 0; i <= m; i++) v->cinc[i] = (v->cc[i] - v->c[i]) * IPERIOD / v->fprd; } for (j = 0, i = (IPERIOD + 1) / 2; j < v->fprd; j++) { x = HTS_Vocoder_get_excitation(v, j, i, nlpf, lpf); if (v->stage == 0) { /* for MCP */ if (x != 0.0) x *= exp(v->c[0]); x = HTS_mlsadf(x, v->c, m, alpha, PADEORDER, v->d1); } else { /* for LSP */ if (!NGAIN) x *= v->c[0]; x = HTS_mglsadf(x, v->c, m, alpha, v->stage, v->d1); } x *= volume; /* output */ if (rawdata) rawdata[rawidx++] = x; if (audio) { if (x > 32767.0) xs = 32767; else if (x < -32768.0) xs = -32768; else xs = (short) x; HTS_Audio_write(audio, xs); } if (!--i) { for (i = 0; i <= m; i++) v->c[i] += v->cinc[i]; i = IPERIOD; } } HTS_Vocoder_end_excitation(v, nlpf); HTS_movem(v->cc, v->c, m + 1); } /* HTS_Vocoder_clear: clear vocoder */ void HTS_Vocoder_clear(HTS_Vocoder * v) { if (v != NULL) { /* free buffer */ if (v->freqt_buff != NULL) { HTS_free(v->freqt_buff); v->freqt_buff = NULL; } v->freqt_size = 0; if (v->gc2gc_buff != NULL) { HTS_free(v->gc2gc_buff); v->gc2gc_buff = NULL; } v->gc2gc_size = 0; if (v->lsp2lpc_buff != NULL) { HTS_free(v->lsp2lpc_buff); v->lsp2lpc_buff = NULL; } v->lsp2lpc_size = 0; if (v->postfilter_buff != NULL) { HTS_free(v->postfilter_buff); v->postfilter_buff = NULL; } v->postfilter_size = 0; if (v->spectrum2en_buff != NULL) { HTS_free(v->spectrum2en_buff); v->spectrum2en_buff = NULL; } v->spectrum2en_size = 0; if (v->c != NULL) { HTS_free(v->c); v->c = NULL; } v->pulse_size = 0; if (v->pulse_buff != NULL) { HTS_free(v->pulse_buff); v->pulse_buff = NULL; } } } HTS_VOCODER_C_END; #endif /* !HTS_VOCODER_C */ festival/src/modules/hts_engine/HTS_engine.h0000644000175000017500000005054712110442173017523 0ustar awbawb/* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ #ifndef HTS_ENGINE_H #define HTS_ENGINE_H #ifdef __cplusplus #define HTS_ENGINE_H_START extern "C" { #define HTS_ENGINE_H_END } #else #define HTS_ENGINE_H_START #define HTS_ENGINE_H_END #endif /* __CPLUSPLUS */ HTS_ENGINE_H_START; #include /* common ---------------------------------------------------------- */ typedef int HTS_Boolean; #ifndef TRUE #define TRUE 1 #endif /* !TRUE */ #ifndef FALSE #define FALSE 0 #endif /* !FALSE */ #ifndef HTS_NODATA #define HTS_NODATA (-1.0e+10) #endif /* HTS_NODATA */ /* copyright ------------------------------------------------------- */ #define HTS_COPYRIGHT "The HMM-Based Speech Synthesis Engine \"hts_engine API\"\nVersion 1.07 (http://hts-engine.sourceforge.net/)\nCopyright (C) 2001-2012 Nagoya Institute of Technology\n 2001-2008 Tokyo Institute of Technology\nAll rights reserved.\n" /* audio ----------------------------------------------------------- */ /* HTS_Audio: audio output wrapper */ typedef struct _HTS_Audio { size_t sampling_frequency; /* sampling frequency */ size_t max_buff_size; /* buffer size for audio output interface */ short *buff; /* current buffer */ size_t buff_size; /* current buffer size */ void *audio_interface; /* audio interface specified in compile step */ } HTS_Audio; /* model ----------------------------------------------------------- */ /* HTS_Window: window coefficients to calculate dynamic features. */ typedef struct _HTS_Window { size_t size; /* # of windows (static + deltas) */ int *l_width; /* left width of windows */ int *r_width; /* right width of windows */ double **coefficient; /* window coefficient */ size_t max_width; /* maximum width of windows */ } HTS_Window; /* HTS_Pattern: list of patterns in a question and a tree. */ typedef struct _HTS_Pattern { char *string; /* pattern string */ struct _HTS_Pattern *next; /* pointer to the next pattern */ } HTS_Pattern; /* HTS_Question: list of questions in a tree. */ typedef struct _HTS_Question { char *string; /* name of this question */ HTS_Pattern *head; /* pointer to the head of pattern list */ struct _HTS_Question *next; /* pointer to the next question */ } HTS_Question; /* HTS_Node: list of tree nodes in a tree. */ typedef struct _HTS_Node { int index; /* index of this node */ size_t pdf; /* index of PDF for this node (leaf node only) */ struct _HTS_Node *yes; /* pointer to its child node (yes) */ struct _HTS_Node *no; /* pointer to its child node (no) */ struct _HTS_Node *next; /* pointer to the next node */ HTS_Question *quest; /* question applied at this node */ } HTS_Node; /* HTS_Tree: list of decision trees in a model. */ typedef struct _HTS_Tree { HTS_Pattern *head; /* pointer to the head of pattern list for this tree */ struct _HTS_Tree *next; /* pointer to next tree */ HTS_Node *root; /* root node of this tree */ size_t state; /* state index of this tree */ } HTS_Tree; /* HTS_Model: set of PDFs, decision trees and questions. */ typedef struct _HTS_Model { size_t vector_length; /* vector length (static features only) */ size_t num_windows; /* # of windows for delta */ HTS_Boolean is_msd; /* flag for MSD */ size_t ntree; /* # of trees */ size_t *npdf; /* # of PDFs at each tree */ float ***pdf; /* PDFs */ HTS_Tree *tree; /* pointer to the list of trees */ HTS_Question *question; /* pointer to the list of questions */ } HTS_Model; /* HTS_ModelSet: set of duration models, HMMs and GV models. */ typedef struct _HTS_ModelSet { char *hts_voice_version; /* version of HTS voice format */ size_t sampling_frequency; /* sampling frequency */ size_t frame_period; /* frame period */ size_t num_voices; /* # of HTS voices */ size_t num_states; /* # of HMM states */ size_t num_streams; /* # of streams */ char *stream_type; /* stream type */ char *fullcontext_format; /* fullcontext label format */ char *fullcontext_version; /* version of fullcontext label */ HTS_Question *gv_off_context; /* GV switch */ char **option; /* options for each stream */ HTS_Model *duration; /* duration PDFs and trees */ HTS_Window *window; /* window coefficients for delta */ HTS_Model **stream; /* parameter PDFs and trees */ HTS_Model **gv; /* GV PDFs and trees */ } HTS_ModelSet; /* label ----------------------------------------------------------- */ /* HTS_LabelString: individual label string with time information */ typedef struct _HTS_LabelString { struct _HTS_LabelString *next; /* pointer to next label string */ char *name; /* label string */ double start; /* start frame specified in the given label */ double end; /* end frame specified in the given label */ } HTS_LabelString; /* HTS_Label: list of label strings */ typedef struct _HTS_Label { HTS_LabelString *head; /* pointer to the head of label string */ size_t size; /* # of label strings */ } HTS_Label; /* sstream --------------------------------------------------------- */ /* HTS_SStream: individual state stream */ typedef struct _HTS_SStream { size_t vector_length; /* vector length (static features only) */ double **mean; /* mean vector sequence */ double **vari; /* variance vector sequence */ double *msd; /* MSD parameter sequence */ size_t win_size; /* # of windows (static + deltas) */ int *win_l_width; /* left width of windows */ int *win_r_width; /* right width of windows */ double **win_coefficient; /* window cofficients */ size_t win_max_width; /* maximum width of windows */ double *gv_mean; /* mean vector of GV */ double *gv_vari; /* variance vector of GV */ HTS_Boolean *gv_switch; /* GV flag sequence */ } HTS_SStream; /* HTS_SStreamSet: set of state stream */ typedef struct _HTS_SStreamSet { HTS_SStream *sstream; /* state streams */ size_t nstream; /* # of streams */ size_t nstate; /* # of states */ size_t *duration; /* duration sequence */ size_t total_state; /* total state */ size_t total_frame; /* total frame */ } HTS_SStreamSet; /* pstream --------------------------------------------------------- */ /* HTS_SMatrices: matrices/vectors used in the speech parameter generation algorithm. */ typedef struct _HTS_SMatrices { double **mean; /* mean vector sequence */ double **ivar; /* inverse diag variance sequence */ double *g; /* vector used in the forward substitution */ double **wuw; /* W' U^-1 W */ double *wum; /* W' U^-1 mu */ } HTS_SMatrices; /* HTS_PStream: individual PDF stream. */ typedef struct _HTS_PStream { size_t vector_length; /* vector length (static features only) */ size_t length; /* stream length */ size_t width; /* width of dynamic window */ double **par; /* output parameter vector */ HTS_SMatrices sm; /* matrices for parameter generation */ size_t win_size; /* # of windows (static + deltas) */ int *win_l_width; /* left width of windows */ int *win_r_width; /* right width of windows */ double **win_coefficient; /* window coefficients */ HTS_Boolean *msd_flag; /* Boolean sequence for MSD */ double *gv_mean; /* mean vector of GV */ double *gv_vari; /* variance vector of GV */ HTS_Boolean *gv_switch; /* GV flag sequence */ size_t gv_length; /* frame length for GV calculation */ } HTS_PStream; /* HTS_PStreamSet: set of PDF streams. */ typedef struct _HTS_PStreamSet { HTS_PStream *pstream; /* PDF streams */ size_t nstream; /* # of PDF streams */ size_t total_frame; /* total frame */ } HTS_PStreamSet; /* gstream --------------------------------------------------------- */ /* HTS_GStream: generated parameter stream. */ typedef struct _HTS_GStream { size_t vector_length; /* vector length (static features only) */ double **par; /* generated parameter */ } HTS_GStream; /* HTS_GStreamSet: set of generated parameter stream. */ typedef struct _HTS_GStreamSet { size_t total_nsample; /* total sample */ size_t total_frame; /* total frame */ size_t nstream; /* # of streams */ HTS_GStream *gstream; /* generated parameter streams */ double *gspeech; /* generated speech */ } HTS_GStreamSet; /* engine ---------------------------------------------------------- */ /* HTS_Condition: synthesis condition */ typedef struct _HTS_Condition { /* global */ size_t sampling_frequency; /* sampling frequency */ size_t fperiod; /* frame period */ size_t audio_buff_size; /* audio buffer size (for audio device) */ HTS_Boolean stop; /* stop flag */ double volume; /* volume */ double *msd_threshold; /* MSD thresholds */ double *gv_weight; /* GV weights */ /* duration */ HTS_Boolean phoneme_alignment_flag; /* flag for using phoneme alignment in label */ double speed; /* speech speed */ /* spectrum */ size_t stage; /* if stage=0 then gamma=0 else gamma=-1/stage */ HTS_Boolean use_log_gain; /* log gain flag (for LSP) */ double alpha; /* all-pass constant */ double beta; /* postfiltering coefficient */ /* log F0 */ double additional_half_tone; /* additional half tone */ /* interpolation weights */ double *duration_iw; /* weights for duration interpolation */ double **parameter_iw; /* weights for parameter interpolation */ double **gv_iw; /* weights for GV interpolation */ } HTS_Condition; /* HTS_Engine: Engine itself. */ typedef struct _HTS_Engine { HTS_Condition condition; /* synthesis condition */ HTS_Audio audio; /* audio output */ HTS_ModelSet ms; /* set of duration models, HMMs and GV models */ HTS_Label label; /* label */ HTS_SStreamSet sss; /* set of state streams */ HTS_PStreamSet pss; /* set of PDF streams */ HTS_GStreamSet gss; /* set of generated parameter streams */ } HTS_Engine; /* engine method --------------------------------------------------- */ /* HTS_Engine_initialize: initialize engine */ void HTS_Engine_initialize(HTS_Engine * engine); /* HTS_Engine_load: load HTS voices */ HTS_Boolean HTS_Engine_load(HTS_Engine * engine, char **voices, size_t num_voices); /* HTS_Engine_set_sampling_frequency: set sampling fraquency */ void HTS_Engine_set_sampling_frequency(HTS_Engine * engine, size_t i); /* HTS_Engine_get_sampling_frequency: get sampling frequency */ size_t HTS_Engine_get_sampling_frequency(HTS_Engine * engine); /* HTS_Engine_set_fperiod: set frame period */ void HTS_Engine_set_fperiod(HTS_Engine * engine, size_t i); /* HTS_Engine_get_fperiod: get frame period */ size_t HTS_Engine_get_fperiod(HTS_Engine * engine); /* HTS_Engine_set_audio_buff_size: set audio buffer size */ void HTS_Engine_set_audio_buff_size(HTS_Engine * engine, size_t i); /* HTS_Engine_get_audio_buff_size: get audio buffer size */ size_t HTS_Engine_get_audio_buff_size(HTS_Engine * engine); /* HTS_Engine_set_stop_flag: set stop flag */ void HTS_Engine_set_stop_flag(HTS_Engine * engine, HTS_Boolean b); /* HTS_Engine_get_stop_flag: get stop flag */ HTS_Boolean HTS_Engine_get_stop_flag(HTS_Engine * engine); /* HTS_Engine_set_volume: set volume in db */ void HTS_Engine_set_volume(HTS_Engine * engine, double f); /* HTS_Engine_get_volume: get volume in db */ double HTS_Engine_get_volume(HTS_Engine * engine); /* HTS_Egnine_set_msd_threshold: set MSD threshold */ void HTS_Engine_set_msd_threshold(HTS_Engine * engine, size_t stream_index, double f); /* HTS_Engine_get_msd_threshold: get MSD threshold */ double HTS_Engine_get_msd_threshold(HTS_Engine * engine, size_t stream_index); /* HTS_Engine_set_gv_weight: set GV weight */ void HTS_Engine_set_gv_weight(HTS_Engine * engine, size_t stream_index, double f); /* HTS_Engine_get_gv_weight: get GV weight */ double HTS_Engine_get_gv_weight(HTS_Engine * engine, size_t stream_index); /* HTS_Engine_set_speed: set speech speed */ void HTS_Engine_set_speed(HTS_Engine * engine, double f); /* HTS_Engine_set_phoneme_alignment_flag: set flag for using phoneme alignment in label */ void HTS_Engine_set_phoneme_alignment_flag(HTS_Engine * engine, HTS_Boolean b); /* HTS_Engine_set_alpha: set alpha */ void HTS_Engine_set_alpha(HTS_Engine * engine, double f); /* HTS_Engine_get_alpha: get alpha */ double HTS_Engine_get_alpha(HTS_Engine * engine); /* HTS_Engine_set_beta: set beta */ void HTS_Engine_set_beta(HTS_Engine * engine, double f); /* HTS_Engine_get_beta: get beta */ double HTS_Engine_get_beta(HTS_Engine * engine); /* HTS_Engine_add_half_tone: add half tone */ void HTS_Engine_add_half_tone(HTS_Engine * engine, double f); /* HTS_Engine_set_duration_interpolation_weight: set interpolation weight for duration */ void HTS_Engine_set_duration_interpolation_weight(HTS_Engine * engine, size_t voice_index, double f); /* HTS_Engine_get_duration_interpolation_weight: get interpolation weight for duration */ double HTS_Engine_get_duration_interpolation_weight(HTS_Engine * engine, size_t voice_index); /* HTS_Engine_set_parameter_interpolation_weight: set interpolation weight for parameter */ void HTS_Engine_set_parameter_interpolation_weight(HTS_Engine * engine, size_t voice_index, size_t stream_index, double f); /* HTS_Engine_get_parameter_interpolation_weight: get interpolation weight for parameter */ double HTS_Engine_get_parameter_interpolation_weight(HTS_Engine * engine, size_t voice_index, size_t stream_index); /* HTS_Engine_set_gv_interpolation_weight: set interpolation weight for GV */ void HTS_Engine_set_gv_interpolation_weight(HTS_Engine * engine, size_t voice_index, size_t stream_index, double f); /* HTS_Engine_get_gv_interpolation_weight: get interpolation weight for GV */ double HTS_Engine_get_gv_interpolation_weight(HTS_Engine * engine, size_t voice_index, size_t stream_index); /* HTS_Engine_get_total_state: get total number of state */ size_t HTS_Engine_get_total_state(HTS_Engine * engine); /* HTS_Engine_set_state_mean: set mean value of state */ void HTS_Engine_set_state_mean(HTS_Engine * engine, size_t stream_index, size_t state_index, size_t vector_index, double f); /* HTS_Engine_get_state_mean: get mean value of state */ double HTS_Engine_get_state_mean(HTS_Engine * engine, size_t stream_index, size_t state_index, size_t vector_index); /* HTS_Engine_get_state_duration: get state duration */ size_t HTS_Engine_get_state_duration(HTS_Engine * engine, size_t state_index); /* HTS_Engine_get_nvoices: get number of voices */ size_t HTS_Engine_get_nvoices(HTS_Engine * engine); /* HTS_Engine_get_nstream: get number of stream */ size_t HTS_Engine_get_nstream(HTS_Engine * engine); /* HTS_Engine_get_nstate: get number of state */ size_t HTS_Engine_get_nstate(HTS_Engine * engine); /* HTS_Engine_get_total_frame: get total number of frame */ size_t HTS_Engine_get_total_frame(HTS_Engine * engine); /* HTS_Engine_get_nsamples: get number of samples */ size_t HTS_Engine_get_nsamples(HTS_Engine * engine); /* HTS_Engine_get_generated_parameter: output generated parameter */ double HTS_Engine_get_generated_parameter(HTS_Engine * engine, size_t stream_index, size_t frame_index, size_t vector_index); /* HTS_Engine_get_generated_speech: output generated speech */ double HTS_Engine_get_generated_speech(HTS_Engine * engine, size_t index); /* HTS_Engine_synthesize_from_fn: synthesize speech from file name */ HTS_Boolean HTS_Engine_synthesize_from_fn(HTS_Engine * engine, const char *fn); /* HTS_Engine_synthesize_from_strings: synthesize speech from string list */ HTS_Boolean HTS_Engine_synthesize_from_strings(HTS_Engine * engine, char **lines, size_t num_lines); /* HTS_Engine_save_information: save trace information */ void HTS_Engine_save_information(HTS_Engine * engine, FILE * fp); /* HTS_Engine_save_label: save label with time */ void HTS_Engine_save_label(HTS_Engine * engine, FILE * fp); /* HTS_Engine_save_generated_parameter: save generated parameter */ void HTS_Engine_save_generated_parameter(HTS_Engine * engine, size_t stream_index, FILE * fp); /* HTS_Engine_save_generated_speech: save generated speech */ void HTS_Engine_save_generated_speech(HTS_Engine * engine, FILE * fp); /* HTS_Engine_save_riff: save RIFF format file */ void HTS_Engine_save_riff(HTS_Engine * engine, FILE * fp); /* HTS_Engine_refresh: free memory per one time synthesis */ void HTS_Engine_refresh(HTS_Engine * engine); /* HTS_Engine_clear: free engine */ void HTS_Engine_clear(HTS_Engine * engine); HTS_ENGINE_H_END; #endif /* !HTS_ENGINE_H */ festival/src/modules/hts_engine/HTS_hidden.h0000644000175000017500000005407712110442173017513 0ustar awbawb/* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ #ifndef HTS_HIDDEN_H #define HTS_HIDDEN_H #ifdef __cplusplus #define HTS_HIDDEN_H_START extern "C" { #define HTS_HIDDEN_H_END } #else #define HTS_HIDDEN_H_START #define HTS_HIDDEN_H_END #endif /* __CPLUSPLUS */ HTS_HIDDEN_H_START; /* hts_engine libraries */ #include "HTS_engine.h" /* common ---------------------------------------------------------- */ #define HTS_MAXBUFLEN 1024 #if !defined(WORDS_BIGENDIAN) && !defined(WORDS_LITTLEENDIAN) #define WORDS_LITTLEENDIAN #endif /* !WORDS_BIGENDIAN && !WORDS_LITTLEENDIAN */ #if defined(WORDS_BIGENDIAN) && defined(WORDS_LITTLEENDIAN) #undef WORDS_BIGENDIAN #endif /* WORDS_BIGENDIAN && WORDS_LITTLEENDIAN */ #define MAX_F0 20000.0 #define MIN_F0 20.0 #define MAX_LF0 9.9034875525361280454891979401956 /* log(20000.0) */ #define MIN_LF0 2.9957322735539909934352235761425 /* log(20.0) */ #define HALF_TONE 0.05776226504666210911810267678818 /* log(2.0) / 12.0 */ #define DB 0.11512925464970228420089957273422 /* log(10.0) / 20.0 */ /* misc ------------------------------------------------------------ */ typedef struct _HTS_File { unsigned char type; void *pointer; } HTS_File; /* HTS_fopen: wrapper for fopen */ HTS_File *HTS_fopen_from_fn(const char *name, const char *opt); /* HTS_fopen_from_fp: wrapper for fopen */ HTS_File *HTS_fopen_from_fp(HTS_File * fp, size_t size); /* HTS_fopen_from_data: wrapper for fopen */ HTS_File *HTS_fopen_from_data(void *data, size_t size); /* HTS_fclose: wrapper for fclose */ void HTS_fclose(HTS_File * fp); /* HTS_fgetc: wrapper for fgetc */ int HTS_fgetc(HTS_File * fp); /* HTS_feof: wrapper for feof */ int HTS_feof(HTS_File * fp); /* HTS_fseek: wrapper for fseek */ int HTS_fseek(HTS_File * fp, long offset, int origin); /* HTS_ftell: wrapper for ftell */ size_t HTS_ftell(HTS_File * fp); /* HTS_fread_big_endiana: fread with byteswap */ size_t HTS_fread_big_endian(void *buf, size_t size, size_t n, HTS_File * fp); /* HTS_fread_little_endiana: fread with byteswap */ size_t HTS_fread_little_endian(void *buf, size_t size, size_t n, HTS_File * fp); /* HTS_fwrite_little_endian: fwrite with byteswap */ size_t HTS_fwrite_little_endian(const void *buf, size_t size, size_t n, FILE * fp); /* HTS_get_pattern_token: get pattern token (single/double quote can be used) */ HTS_Boolean HTS_get_pattern_token(HTS_File * fp, char *buff); /* HTS_get_token: get token from file pointer (separators are space,tab,line break) */ HTS_Boolean HTS_get_token_from_fp(HTS_File * fp, char *buff); /* HTS_get_token: get token from file pointer with specified separator */ HTS_Boolean HTS_get_token_from_fp_with_separator(HTS_File * fp, char *buff, char separator); /* HTS_get_token_from_string: get token from string (separator are space,tab,line break) */ HTS_Boolean HTS_get_token_from_string(const char *string, size_t * index, char *buff); /* HTS_get_token_from_string_with_separator: get token from string with specified separator */ HTS_Boolean HTS_get_token_from_string_with_separator(const char *str, size_t * index, char *buff, char separator); /* HTS_calloc: wrapper for calloc */ void *HTS_calloc(const size_t num, const size_t size); /* HTS_strdup: wrapper for strdup */ char *HTS_strdup(const char *string); /* HTS_calloc_matrix: allocate double matrix */ double **HTS_alloc_matrix(size_t x, size_t y); /* HTS_free_matrix: free double matrix */ void HTS_free_matrix(double **p, size_t x); /* HTS_Free: wrapper for free */ void HTS_free(void *p); /* HTS_error: output error message */ void HTS_error(int error, const char *message, ...); /* audio ----------------------------------------------------------- */ /* HTS_Audio_initialize: initialize audio */ void HTS_Audio_initialize(HTS_Audio * audio); /* HTS_Audio_set_parameter: set parameters for audio */ void HTS_Audio_set_parameter(HTS_Audio * audio, size_t sampling_frequency, size_t max_buff_size); /* HTS_Audio_write: send data to audio */ void HTS_Audio_write(HTS_Audio * audio, short data); /* HTS_Audio_flush: flush remain data */ void HTS_Audio_flush(HTS_Audio * audio); /* HTS_Audio_clear: free audio */ void HTS_Audio_clear(HTS_Audio * audio); /* model ----------------------------------------------------------- */ /* HTS_ModelSet_initialize: initialize model set */ void HTS_ModelSet_initialize(HTS_ModelSet * ms); /* HTS_ModelSet_load: load HTS voices */ HTS_Boolean HTS_ModelSet_load(HTS_ModelSet * ms, char **voices, size_t num_voices); /* HTS_ModelSet_get_sampling_frequency: get sampling frequency of HTS voices */ size_t HTS_ModelSet_get_sampling_frequency(HTS_ModelSet * ms); /* HTS_ModelSet_get_fperiod: get frame period of HTS voices */ size_t HTS_ModelSet_get_fperiod(HTS_ModelSet * ms); /* HTS_ModelSet_get_fperiod: get stream option */ const char *HTS_ModelSet_get_option(HTS_ModelSet * ms, size_t stream_index); /* HTS_ModelSet_get_gv_flag: get GV flag */ HTS_Boolean HTS_ModelSet_get_gv_flag(HTS_ModelSet * ms, const char *string); /* HTS_ModelSet_get_nstate: get number of state */ size_t HTS_ModelSet_get_nstate(HTS_ModelSet * ms); /* HTS_ModelSet_get_nstream: get number of stream */ size_t HTS_ModelSet_get_nstream(HTS_ModelSet * ms); /* HTS_ModelSet_get_nvoices: get number of HTS voices */ size_t HTS_ModelSet_get_nvoices(HTS_ModelSet * ms); /* HTS_ModelSet_get_vector_length: get vector length */ size_t HTS_ModelSet_get_vector_length(HTS_ModelSet * ms, size_t stream_index); /* HTS_ModelSet_is_msd: get MSD flag */ HTS_Boolean HTS_ModelSet_is_msd(HTS_ModelSet * ms, size_t stream_index); /* HTS_ModelSet_get_window_size: get dynamic window size */ size_t HTS_ModelSet_get_window_size(HTS_ModelSet * ms, size_t stream_index); /* HTS_ModelSet_get_window_left_width: get left width of dynamic window */ int HTS_ModelSet_get_window_left_width(HTS_ModelSet * ms, size_t stream_index, size_t window_index); /* HTS_ModelSet_get_window_right_width: get right width of dynamic window */ int HTS_ModelSet_get_window_right_width(HTS_ModelSet * ms, size_t stream_index, size_t window_index); /* HTS_ModelSet_get_window_coefficient: get coefficient of dynamic window */ double HTS_ModelSet_get_window_coefficient(HTS_ModelSet * ms, size_t stream_index, size_t window_index, size_t coefficient_index); /* HTS_ModelSet_get_window_max_width: get max width of dynamic window */ size_t HTS_ModelSet_get_window_max_width(HTS_ModelSet * ms, size_t stream_index); /* HTS_ModelSet_use_gv: get GV flag */ HTS_Boolean HTS_ModelSet_use_gv(HTS_ModelSet * ms, size_t stream_index); /* HTS_ModelSet_get_duration_index: get index of duration tree and PDF */ void HTS_ModelSet_get_duration_index(HTS_ModelSet * ms, size_t voice_index, const char *string, size_t * tree_index, size_t * pdf_index); /* HTS_ModelSet_get_duration: get duration using interpolation weight */ void HTS_ModelSet_get_duration(HTS_ModelSet * ms, const char *string, const double *iw, double *mean, double *vari); /* HTS_ModelSet_get_parameter_index: get index of parameter tree and PDF */ void HTS_ModelSet_get_parameter_index(HTS_ModelSet * ms, size_t voice_index, size_t stream_index, size_t state_index, const char *string, size_t * tree_index, size_t * pdf_index); /* HTS_ModelSet_get_parameter: get parameter using interpolation weight */ void HTS_ModelSet_get_parameter(HTS_ModelSet * ms, size_t stream_index, size_t state_index, const char *string, const double *iw, double *mean, double *vari, double *msd); void HTS_ModelSet_get_gv_index(HTS_ModelSet * ms, size_t voice_index, size_t stream_index, const char *string, size_t * tree_index, size_t * pdf_index); /* HTS_ModelSet_get_gv: get GV using interpolation weight */ void HTS_ModelSet_get_gv(HTS_ModelSet * ms, size_t stream_index, const char *string, const double *iw, double *mean, double *vari); /* HTS_ModelSet_clear: free model set */ void HTS_ModelSet_clear(HTS_ModelSet * ms); /* label ----------------------------------------------------------- */ /* HTS_Label_initialize: initialize label */ void HTS_Label_initialize(HTS_Label * label); /* HTS_Label_load_from_fn: load label from file name */ void HTS_Label_load_from_fn(HTS_Label * label, size_t sampling_rate, size_t fperiod, const char *fn); /* HTS_Label_load_from_strings: load label list from string list */ void HTS_Label_load_from_strings(HTS_Label * label, size_t sampling_rate, size_t fperiod, char **lines, size_t num_lines); /* HTS_Label_get_size: get number of label string */ size_t HTS_Label_get_size(HTS_Label * label); /* HTS_Label_get_string: get label string */ const char *HTS_Label_get_string(HTS_Label * label, size_t index); /* HTS_Label_get_start_frame: get start frame */ double HTS_Label_get_start_frame(HTS_Label * label, size_t index); /* HTS_Label_get_end_frame: get end frame */ double HTS_Label_get_end_frame(HTS_Label * label, size_t index); /* HTS_Label_clear: free label */ void HTS_Label_clear(HTS_Label * label); /* sstream --------------------------------------------------------- */ /* HTS_SStreamSet_initialize: initialize state stream set */ void HTS_SStreamSet_initialize(HTS_SStreamSet * sss); /* HTS_SStreamSet_create: parse label and determine state duration */ HTS_Boolean HTS_SStreamSet_create(HTS_SStreamSet * sss, HTS_ModelSet * ms, HTS_Label * label, HTS_Boolean phoneme_alignment_flag, double speed, double *duration_iw, double **parameter_iw, double **gv_iw); /* HTS_SStreamSet_get_nstream: get number of stream */ size_t HTS_SStreamSet_get_nstream(HTS_SStreamSet * sss); /* HTS_SStreamSet_get_vector_length: get vector length */ size_t HTS_SStreamSet_get_vector_length(HTS_SStreamSet * sss, size_t stream_index); /* HTS_SStreamSet_is_msd: get MSD flag */ HTS_Boolean HTS_SStreamSet_is_msd(HTS_SStreamSet * sss, size_t stream_index); /* HTS_SStreamSet_get_total_state: get total number of state */ size_t HTS_SStreamSet_get_total_state(HTS_SStreamSet * sss); /* HTS_SStreamSet_get_total_frame: get total number of frame */ size_t HTS_SStreamSet_get_total_frame(HTS_SStreamSet * sss); /* HTS_SStreamSet_get_msd: get msd parameter */ double HTS_SStreamSet_get_msd(HTS_SStreamSet * sss, size_t stream_index, size_t state_index); /* HTS_SStreamSet_window_size: get dynamic window size */ size_t HTS_SStreamSet_get_window_size(HTS_SStreamSet * sss, size_t stream_index); /* HTS_SStreamSet_get_window_left_width: get left width of dynamic window */ int HTS_SStreamSet_get_window_left_width(HTS_SStreamSet * sss, size_t stream_index, size_t window_index); /* HTS_SStreamSet_get_window_right_width: get right width of dynamic window */ int HTS_SStreamSet_get_window_right_width(HTS_SStreamSet * sss, size_t stream_index, size_t window_index); /* HTS_SStreamSet_get_window_coefficient: get coefficient of dynamic window */ double HTS_SStreamSet_get_window_coefficient(HTS_SStreamSet * sss, size_t stream_index, size_t window_index, int coefficient_index); /* HTS_SStreamSet_get_window_max_width: get max width of dynamic window */ size_t HTS_SStreamSet_get_window_max_width(HTS_SStreamSet * sss, size_t stream_index); /* HTS_SStreamSet_use_gv: get GV flag */ HTS_Boolean HTS_SStreamSet_use_gv(HTS_SStreamSet * sss, size_t stream_index); /* HTS_SStreamSet_get_duration: get state duration */ size_t HTS_SStreamSet_get_duration(HTS_SStreamSet * sss, size_t state_index); /* HTS_SStreamSet_get_mean: get mean parameter */ double HTS_SStreamSet_get_mean(HTS_SStreamSet * sss, size_t stream_index, size_t state_index, size_t vector_index); /* HTS_SStreamSet_set_mean: set mean parameter */ void HTS_SStreamSet_set_mean(HTS_SStreamSet * sss, size_t stream_index, size_t state_index, size_t vector_index, double f); /* HTS_SStreamSet_get_vari: get variance parameter */ double HTS_SStreamSet_get_vari(HTS_SStreamSet * sss, size_t stream_index, size_t state_index, size_t vector_index); /* HTS_SStreamSet_set_vari: set variance parameter */ void HTS_SStreamSet_set_vari(HTS_SStreamSet * sss, size_t stream_index, size_t state_index, size_t vector_index, double f); /* HTS_SStreamSet_get_gv_mean: get GV mean parameter */ double HTS_SStreamSet_get_gv_mean(HTS_SStreamSet * sss, size_t stream_index, size_t vector_index); /* HTS_SStreamSet_get_gv_mean: get GV variance parameter */ double HTS_SStreamSet_get_gv_vari(HTS_SStreamSet * sss, size_t stream_index, size_t vector_index); /* HTS_SStreamSet_set_gv_switch: set GV switch */ void HTS_SStreamSet_set_gv_switch(HTS_SStreamSet * sss, size_t stream_index, size_t state_index, HTS_Boolean i); /* HTS_SStreamSet_get_gv_switch: get GV switch */ HTS_Boolean HTS_SStreamSet_get_gv_switch(HTS_SStreamSet * sss, size_t stream_index, size_t state_index); /* HTS_SStreamSet_clear: free state stream set */ void HTS_SStreamSet_clear(HTS_SStreamSet * sss); /* pstream --------------------------------------------------------- */ /* check variance in finv() */ #define INFTY ((double) 1.0e+38) #define INFTY2 ((double) 1.0e+19) #define INVINF ((double) 1.0e-38) #define INVINF2 ((double) 1.0e-19) /* GV */ #define STEPINIT 0.1 #define STEPDEC 0.5 #define STEPINC 1.2 #define W1 1.0 #define W2 1.0 #define GV_MAX_ITERATION 5 /* HTS_PStreamSet_initialize: initialize parameter stream set */ void HTS_PStreamSet_initialize(HTS_PStreamSet * pss); /* HTS_PStreamSet_create: parameter generation using GV weight */ HTS_Boolean HTS_PStreamSet_create(HTS_PStreamSet * pss, HTS_SStreamSet * sss, double *msd_threshold, double *gv_weight); /* HTS_PStreamSet_get_nstream: get number of stream */ size_t HTS_PStreamSet_get_nstream(HTS_PStreamSet * pss); /* HTS_PStreamSet_get_static_length: get features length */ size_t HTS_PStreamSet_get_vector_length(HTS_PStreamSet * pss, size_t stream_index); /* HTS_PStreamSet_get_total_frame: get total number of frame */ size_t HTS_PStreamSet_get_total_frame(HTS_PStreamSet * pss); /* HTS_PStreamSet_get_parameter: get parameter */ double HTS_PStreamSet_get_parameter(HTS_PStreamSet * pss, size_t stream_index, size_t frame_index, size_t vector_index); /* HTS_PStreamSet_get_parameter_vector: get parameter vector */ double *HTS_PStreamSet_get_parameter_vector(HTS_PStreamSet * pss, size_t stream_index, size_t frame_index); /* HTS_PStreamSet_get_msd_flag: get generated MSD flag per frame */ HTS_Boolean HTS_PStreamSet_get_msd_flag(HTS_PStreamSet * pss, size_t stream_index, size_t frame_index); /* HTS_PStreamSet_is_msd: get MSD flag */ HTS_Boolean HTS_PStreamSet_is_msd(HTS_PStreamSet * pss, size_t stream_index); /* HTS_PStreamSet_clear: free parameter stream set */ void HTS_PStreamSet_clear(HTS_PStreamSet * pss); /* gstream --------------------------------------------------------- */ /* HTS_GStreamSet_initialize: initialize generated parameter stream set */ void HTS_GStreamSet_initialize(HTS_GStreamSet * gss); /* HTS_GStreamSet_create: generate speech */ HTS_Boolean HTS_GStreamSet_create(HTS_GStreamSet * gss, HTS_PStreamSet * pss, size_t stage, HTS_Boolean use_log_gain, size_t sampling_rate, size_t fperiod, double alpha, double beta, HTS_Boolean * stop, double volume, HTS_Audio * audio); /* HTS_GStreamSet_get_total_nsamples: get total number of sample */ size_t HTS_GStreamSet_get_total_nsamples(HTS_GStreamSet * gss); /* HTS_GStreamSet_get_total_frame: get total number of frame */ size_t HTS_GStreamSet_get_total_frame(HTS_GStreamSet * gss); /* HTS_GStreamSet_get_static_length: get features length */ size_t HTS_GStreamSet_get_vector_length(HTS_GStreamSet * gss, size_t stream_index); /* HTS_GStreamSet_get_speech: get synthesized speech parameter */ double HTS_GStreamSet_get_speech(HTS_GStreamSet * gss, size_t sample_index); /* HTS_GStreamSet_get_parameter: get generated parameter */ double HTS_GStreamSet_get_parameter(HTS_GStreamSet * gss, size_t stream_index, size_t frame_index, size_t vector_index); /* HTS_GStreamSet_clear: free generated parameter stream set */ void HTS_GStreamSet_clear(HTS_GStreamSet * gss); /* vocoder --------------------------------------------------------- */ #ifndef LZERO #define LZERO (-1.0e+10) /* ~log(0) */ #endif /* !LZERO */ #ifndef ZERO #define ZERO (1.0e-10) /* ~(0) */ #endif /* !ZERO */ #ifndef PI #define PI 3.14159265358979323846 #endif /* !PI */ #ifndef PI2 #define PI2 6.28318530717958647692 #endif /* !PI2 */ #define RANDMAX 32767 #define IPERIOD 1 #define SEED 1 #define B0 0x00000001 #define B28 0x10000000 #define B31 0x80000000 #define B31_ 0x7fffffff #define Z 0x00000000 #ifdef HTS_EMBEDDED #define GAUSS FALSE #define PADEORDER 4 /* pade order (for MLSA filter) */ #define IRLENG 384 /* length of impulse response */ #else #define GAUSS TRUE #define PADEORDER 5 #define IRLENG 576 #endif /* HTS_EMBEDDED */ #define CHECK_LSP_STABILITY_MIN 0.25 #define CHECK_LSP_STABILITY_NUM 4 /* for MGLSA filter */ #define NORMFLG1 TRUE #define NORMFLG2 FALSE #define MULGFLG1 TRUE #define MULGFLG2 FALSE #define NGAIN FALSE /* HTS_Vocoder: structure for setting of vocoder */ typedef struct _HTS_Vocoder { size_t stage; /* Gamma=-1/stage: if stage=0 then Gamma=0 */ double gamma; /* Gamma */ HTS_Boolean use_log_gain; /* log gain flag (for LSP) */ size_t fprd; /* frame shift */ unsigned long next; /* temporary variable for random generator */ HTS_Boolean gauss; /* flag to use Gaussian noise */ double rate; /* sampling rate */ double p1; /* used in excitation generation */ double pc; /* used in excitation generation */ double p; /* used in excitation generation */ double inc; /* used in excitation generation */ double *pulse_buff; /* used in excitation generation */ size_t pulse_size; /* used in excitation generation */ unsigned char sw; /* switch used in random generator */ int x; /* excitation signal */ double *freqt_buff; /* used in freqt */ size_t freqt_size; /* buffer size for freqt */ double *spectrum2en_buff; /* used in spectrum2en */ size_t spectrum2en_size; /* buffer size for spectrum2en */ double r1, r2, s; /* used in random generator */ double *postfilter_buff; /* used in postfiltering */ size_t postfilter_size; /* buffer size for postfiltering */ double *c, *cc, *cinc, *d1; /* used in the MLSA/MGLSA filter */ double *lsp2lpc_buff; /* used in lsp2lpc */ size_t lsp2lpc_size; /* buffer size of lsp2lpc */ double *gc2gc_buff; /* used in gc2gc */ size_t gc2gc_size; /* buffer size for gc2gc */ } HTS_Vocoder; /* HTS_Vocoder_initialize: initialize vocoder */ void HTS_Vocoder_initialize(HTS_Vocoder * v, size_t m, size_t stage, HTS_Boolean use_log_gain, size_t rate, size_t fperiod); /* HTS_Vocoder_synthesize: pulse/noise excitation and MLSA/MGLSA filster based waveform synthesis */ void HTS_Vocoder_synthesize(HTS_Vocoder * v, size_t m, double lf0, double *spectrum, size_t nlpf, double *lpf, double alpha, double beta, double volume, double *rawdata, HTS_Audio * audio); /* HTS_Vocoder_clear: clear vocoder */ void HTS_Vocoder_clear(HTS_Vocoder * v); HTS_HIDDEN_H_END; #endif /* !HTS_HIDDEN_H */ festival/src/modules/hts_engine/README0000644000175000017500000001544112073131305016241 0ustar awbawb=============================================================================== The HMM-Based Speech Synthesis Engine "hts_engine API" version 1.07 release December 25, 2012 The hts_engine API is an API version of hts_engine which has been released since HTS version 1.1. It has been being developed by the HTS working group (see "Who we are" below) and some graduate students in Nagoya Institute of Technology (see "AUTHORS" in the same directory). ******************************************************************************* Copying ******************************************************************************* The hts_engine API is released under the Modified BSD license (see http://www.opensource.org/). Using and distributing this software is free (without restriction including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of this work, and to permit persons to whom this work is furnished to do so) subject to the conditions in the following license: /* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ Although this software is free, we still offer no warranties and no maintenance. We will continue to endeavor to fix bugs and answer queries when can, but are not in a position to guarantee it. We will consider consultancy if desired, please contacts us for details. If you are using the hts_engine API in commercial environments, even though no license is required, we would be grateful if you let us know as it helps justify ourselves to our various sponsors. We also strongly encourage you to * refer to the use of hts_engine API in any publications that use this software * report bugs, where possible with bug fixes, that are found See also "COPYING" file in the current directory for details. ******************************************************************************* Installation ******************************************************************************* See "INSTALL" in the same directory for details. ******************************************************************************* Documentation ******************************************************************************* Reference manual of hts_engine API is available at http://hts-engine.sourceforge.net/ ******************************************************************************* Acknowledgements ******************************************************************************* Keiichi Tokuda Shinji Sako Heiga Zen Keiichiro Oura Kazuhiro Nakamura Keijiro Saino ******************************************************************************* Who we are ******************************************************************************* The HTS working group is a voluntary group for developing the HMM-Based Speech Synthesis System. Current members are Keiichi Tokuda http://www.sp.nitech.ac.jp/~tokuda/ (Produce and Design) Keiichiro Oura http://www.sp.nitech.ac.jp/~uratec/ (Design and Development, Main Maintainer) Kei Hashimoto http://www.sp.nitech.ac.jp/~bonanza/ Sayaka Shiota http://www.sp.nitech.ac.jp/~sayaka/ Shinji Takaki http://www.sp.nitech.ac.jp/~k-prr44/ Heiga Zen Junichi Yamagishi http://homepages.inf.ed.ac.uk/jyamagis/ Tomoki Toda http://spalab.naist.jp/~tomoki/index_e.html Takashi Nose Shinji Sako http://www.mmsp.nitech.ac.jp/~sako/ Alan W. Black http://www.cs.cmu.edu/~awb/ and the members are dynamically changing. The current formal contact address of HTS working group and a mailing list for HTS users can be found at http://hts.sp.nitech.ac.jp/ =============================================================================== festival/src/modules/hts_engine/COPYING0000644000175000017500000000603012073131302016403 0ustar awbawb/* ----------------------------------------------------------------- */ /* The HMM-Based Speech Synthesis Engine "hts_engine API" */ /* developed by HTS Working Group */ /* http://hts-engine.sourceforge.net/ */ /* ----------------------------------------------------------------- */ /* */ /* Copyright (c) 2001-2012 Nagoya Institute of Technology */ /* Department of Computer Science */ /* */ /* 2001-2008 Tokyo Institute of Technology */ /* Interdisciplinary Graduate School of */ /* Science and Engineering */ /* */ /* All rights reserved. */ /* */ /* Redistribution and use in source and binary forms, with or */ /* without modification, are permitted provided that the following */ /* conditions are met: */ /* */ /* - Redistributions of source code must retain the above copyright */ /* notice, this list of conditions and the following disclaimer. */ /* - Redistributions in binary form must reproduce the above */ /* copyright notice, this list of conditions and the following */ /* disclaimer in the documentation and/or other materials provided */ /* with the distribution. */ /* - Neither the name of the HTS working group nor the names of its */ /* contributors may be used to endorse or promote products derived */ /* from this software without specific prior written permission. */ /* */ /* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND */ /* CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, */ /* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */ /* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */ /* DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS */ /* BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, */ /* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED */ /* TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, */ /* DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON */ /* ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, */ /* OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY */ /* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */ /* POSSIBILITY OF SUCH DAMAGE. */ /* ----------------------------------------------------------------- */ festival/src/modules/hts_engine/INSTALL0000644000175000017500000000050312073131304016402 0ustar awbawbInstallation Instructions ************************* 1. After unpacking the tar.gz file, cd to the hts_engine API directory. 2. Run configure script with appropriate options. % ./configure For detail, please see. % ./configure --help 3. Run make. % make 4. Install library and binary. % make install festival/src/modules/hts_engine/AUTHORS0000644000175000017500000000120712073131302016421 0ustar awbawbThe hts_engine API is software to synthesize speech waveform from HMMs trained by the HMM-based speech synthesis system (HTS). This software is released under the Modified BSD license. See the COPYING file in the same directory as this file for the license. The hts_engine_API has been developed by several members of HTS working group and some graduate students in Nagoya Institute of Technology: Keiichi Tokuda http://www.sp.nitech.ac.jp/~tokuda/ (Produce and Design) Keiichiro Oura http://www.sp.nitech.ac.jp/~uratec/ (Design and Development, Main Maintainer) Heiga Zen Shinji Sako http://www.mmsp.nitech.ac.jp/~sako/ festival/src/main/Makefile0000644000175000017500000000702512131543117014154 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../.. DIRNAME=src/main SRCS = festival_main.cc audsp.cc festival_client.cc OBJS = $(SRCS:.cc=.o) FILES=Makefile $(SRCS) LOCAL_CLEAN = $(ETCDIR)/audsp $(ETCDIR)/.made ETCDIR=$(TOP)/lib/etc/$(SYSTEM_TYPE) ALL_EXECS = festival festival_client ALL = $(ALL_EXECS) make_audiosp include $(TOP)/config/common_make_rules include $(EST)/config/rules/bin_process.mak ETCDIR=$(TOP)/lib/etc/$(SYSTEM_TYPE) ifeq (SIOD_PYTHON, $(findstring SIOD_PYTHON, $(INCLUDE_MODULES))) LIBS += $(shell python3-config --ldflags) -ldl -lutil -lm endif festival: festival_main.o $(LIBDEPS) $(LINK_COMMAND) -o festival festival_main.o $(LIBS) festival_client: festival_client.o $(REQUIRED_LIBDEPS) $(LINK_COMMAND) -o festival_client festival_client.o $(LIBS) $(ETCDIR)/audsp: $(ETCDIR)/.made audsp.o $(LIBDEPS) $(LINK_COMMAND) -o $(ETCDIR)/audsp audsp.o $(LIBS) # Can't just rely on the dir as it gets updated with new files # check for the data of a file created in etcdir make_audiosp: $(ETCDIR)/audsp @$(DO_NOTHING) $(ETCDIR)/.made: @ if [ ! -d $(ETCDIR) ] ; \ then mkdir -p $(ETCDIR); fi @ if [ ! -f $(ETCDIR)/.made ] ; \ then touch $(ETCDIR)/.made ; fi festival/src/main/festival_main.cc0000644000175000017500000002170012431672642015651 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996-1998 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black, Paul Taylor, Richard Caley */ /* and others */ /* Date : April 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Top level file for synthesizer */ /* */ /*=======================================================================*/ #include using namespace std; #include "festival.h" static void festival_main(int argc, char **argv); static int festival_check_script_mode(int argc, char **argv); static void festival_script_mode(int argc, char **argv); void awb_free_diph_index(); extern "C" { void mtrace(); void muntrace(); } int main(int argc, char **argv) { /* putenv("MALLOC_TRACE=mallfile"); mtrace(); */ festival_main(argc,argv); /* awb_free_diph_index(); muntrace(); */ return 0; } static void festival_main(int argc, char **argv) { EST_Option al; int stdin_input,interactive; EST_Litem *p; EST_StrList files; int real_number_of_files = 0; int heap_size = FESTIVAL_HEAP_SIZE; if (festival_check_script_mode(argc,argv) == TRUE) { // Need to check this directly as in script mode args are // passed for analysis in the script itself return; } parse_command_line(argc, argv, EST_String("Usage:\n")+ "festival ...\n"+ "In evaluation mode \"filenames\" starting with ( are evaluated inline\n"+ "Festival Speech Synthesis System: "+ festival_version +"\n"+ "-q Load no default setup files\n"+ "--libdir \n"+ " Set library directory pathname\n"+ "-b Run in batch mode (no interaction)\n"+ "--batch Run in batch mode (no interaction)\n"+ "--tts Synthesize text in files as speech\n"+ " no files means read from stdin\n"+ " (implies no interaction by default)\n"+ "-i Run in interactive mode (default)\n"+ "--interactive\n"+ " Run in interactive mode (default)\n"+ "--pipe Run in pipe mode, reading commands from\n"+ " stdin, but no prompt or return values\n"+ " are printed (default if stdin not a tty)\n"+ "--language \n"+ " Run in named language, default is\n"+ " english, spanish and welsh are available\n"+ "--server Run in server mode waiting for clients\n"+ " of server_port (1314)\n"+ "--script \n"+ " Used in #! scripts, runs in batch mode on\n"+ " file and passes all other args to Scheme\n"+ "--heap {10000000}\n"+ " Set size of Lisp heap, should not normally need\n"+ " to be changed from its default\n"+ "-v Display version number and exit\n"+ "--version Display version number and exit\n", files, al); if ((al.present("-v")) || (al.present("--version"))) { printf("%s: Festival Speech Synthesis System: %s\n", argv[0],festival_version); exit(0); } if (al.present("--libdir")) festival_libdir = wstrdup(al.val("--libdir")); else if (getenv("FESTLIBDIR") != 0) festival_libdir = getenv("FESTLIBDIR"); if (al.present("--heap")) heap_size = al.ival("--heap"); festival_initialize(!al.present("-q"),heap_size); if (al.present("--language")) festival_init_lang(al.val("--language")); // File processing for (p=files.head(); p != 0; p=p->next()) { if (files(p) == "-") // paul thinks I want the "-" -- I don't continue; real_number_of_files++; if (al.present("--tts")) { if (!festival_say_file(files(p))) festival_error(); } else if (files(p).matches(make_regex("^(.*"))) { if (!festival_eval_command(files(p))) festival_error(); // fail if it fails } else if (!festival_load_file(files(p))) festival_error(); } // What to do about standard input and producing prompts etc. if ((al.present("-i")) || (al.present("--interactive"))) { interactive = TRUE; stdin_input = TRUE; } else if ((al.present("--pipe"))) { interactive=FALSE; stdin_input = TRUE; } else if ((al.present("-b")) || (al.present("--batch")) || (al.present("--tts"))) { interactive=FALSE; stdin_input=FALSE; } else if (isatty(0)) // if stdin is a terminal assume interactive { interactive = TRUE; stdin_input = TRUE; } else // else assume pipe mode { interactive = FALSE; stdin_input = TRUE; } if (al.present("--server")) festival_server_mode(); // server mode else if ((al.present("--tts")) && (real_number_of_files == 0)) festival_say_file("-"); // text to speech from files else if (stdin_input) festival_repl(interactive); // expect input from stdin if (al.present("--tts")) festival_wait_for_spooler(); // wait for end of audio output return; } static int festival_check_script_mode(int argc, char **argv) { // Checks if we are in script mode, i.e. if --script exists // which may be possibily be preceeded by --libdir (and heap ?) // if (argc == 0) return FALSE; else if ((argc > 2) && (streq("--script",argv[1]))) { if (getenv("FESTLIBDIR") != 0) festival_libdir = getenv("FESTLIBDIR"); festival_script_mode(argc,argv); return TRUE; } else if ((argc > 4) && (streq("--script",argv[3])) && (streq("--libdir",argv[1]))) { festival_libdir = wstrdup(argv[2]); festival_script_mode(argc,argv); return TRUE; } else return FALSE; } static void festival_script_mode(int argc, char **argv) { // In script mode the first file arg after -script is interpreted and // the remainder are set in the variable argv so the script // itself may do what ever it wants LISP args; const char *siodheapsize; int i; if (argc < 2) { cerr << "festival: script_mode has no file to interpret" << endl; return; } // initialize without loading init files siodheapsize = getenv("SIODHEAPSIZE"); if (siodheapsize) festival_initialize(FALSE,atoi(siodheapsize)); else festival_initialize(FALSE,FESTIVAL_HEAP_SIZE); for (args=NIL,i=3; i < argc; i++) args = cons(rintern(argv[i]),args); siod_set_lval("argv",reverse(args)); siod_set_lval("argc",flocons(argc)); festival_load_file(argv[2]); return; } festival/src/main/audsp.cc0000644000175000017500000002632310126474330014143 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : September 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* An audio file spooler, like lpd. Reads in commands about files to */ /* to play, and queues them until any previous requests are finished. */ /* This allows the synthesizer to get on with synthesizing the next */ /* utterance. */ /* */ /* Actually this doesn't use anything in Festival, only the speech_tools */ /*=======================================================================*/ #include #include #include #include using namespace std; #include "EST.h" #include "EST_unix.h" #ifdef NO_SPOOLER int main(int argc, char **argv) { printf("Audio spooler not supported\n"); return 0; } #else class Command { private: EST_String p_file; int p_rate; public: Command(const EST_String &f, int rate) { p_file=f; p_rate=rate; } int rate(void) const { return p_rate; } const EST_String &file(void) const { return p_file; } }; class CQueue_Item { public: Command *c; CQueue_Item *next; CQueue_Item(Command *com) { c=com; next=0; } ~CQueue_Item() { delete c; if (next != 0) delete next; } }; class CQueue { private: CQueue_Item *head; CQueue_Item *tail; public: CQueue() { head = tail = 0; } ~CQueue() { delete head; } void push(Command *c); Command *pop(void); void display(void) const; int length(void) const; void clear(void); }; static void auspl_main(int argc, char **argv); static void check_new_input(void); static char *read_a_line(void); static void process_command(char *line); static void check_new_output(void); static int execute_command(Command *c); static void load_play_file(Command *c); static int sp_terminate(void); static void tidy_up(void); void CQueue::push(Command *c) { // Put this item on tail CQueue_Item *n = new CQueue_Item(c); if (head == 0) { // first one head = n; tail = n; } else { tail->next = n; tail = n; } } Command *CQueue::pop(void) { // Pop top from the queue if (head == 0) return 0; else { Command *c = head->c; CQueue_Item *h; h = head; h->c = 0; head = head->next; h->next = 0; delete h; return c; } } void CQueue::display(void) const { CQueue_Item *t; int i; cerr << "Command_queue: " << length() << endl; for (i=0,t=head; t != 0; t=t->next,i++) cerr << " " << i << ": " << t->c->file() << endl; } int CQueue::length(void) const { // length of queue CQueue_Item *t; int i; for (i=0,t=head; t != 0; t=t->next) i++; return i; } void CQueue::clear(void) { // Remove all memebers in the queue CQueue_Item *t; // Somebody has to do it ... for (t=head; t != 0; t=t->next) unlink(t->c->file()); delete head; head = 0; tail = 0; } static int no_more_input = FALSE; static CQueue command_queue; static int child_pid = 0; static EST_String current_file; static EST_Option play_wave_options; static int maxqueue = 5; static int pending_close = FALSE; static int kids = 0; int main(int argc, char **argv) { auspl_main(argc,argv); return 0; } static void auspl_main(int argc, char **argv) { EST_Option al; EST_StrList files; parse_command_line(argc, argv, EST_String("Usage: audio spooler \n")+ "auspl ...\n"+ "--method audio play method\n"+ "--command Unix command to play file, used when\n"+ " method is audio_command\n"+ "--maxqueue {5} Maximum number of files in queue\n", files, al); if (al.present("--method")) play_wave_options.add_item("-p",al.val("--method")); if (al.present("--command")) play_wave_options.add_item("-command",al.val("--command")); play_wave_options.add_item("-quality","HIGH"); if (al.present("--maxqueue")) maxqueue = al.ival("--maxqueue"); while (!sp_terminate()) { check_new_input(); check_new_output(); } tidy_up(); } static int sp_terminate(void) { // I'm never very sure of all the conditions necessary to terminate if (no_more_input && (command_queue.length() == 0)) return TRUE; else return FALSE; } static void tidy_up(void) { // should wait for any remaining children if I've been // requested to. int pid; int statusp; if (pending_close == TRUE) { while (kids > 0) { pid = waitpid(0,&statusp,0); kids--; } fprintf(stdout,"OK\n"); // give an acknowledgement fflush(stdout); } return; } static void check_new_input(void) { // Do a select on stdin to find out if there is any new // commands to process fd_set inset; fd_set outset; fd_set exset; struct timeval t; int sv; t.tv_sec = 0; t.tv_usec = 1000; // 0.1 seconds FD_ZERO(&inset); FD_ZERO(&outset); FD_ZERO(&exset); if ((command_queue.length() >= maxqueue) || no_more_input) { // wait a bit for the queue to go down a bit // not we're selecting on no fds at all, just for the delay sv = select(0,&inset,&outset,&exset,&t); return; } FD_SET(0,&inset); sv = select(1,&inset,&outset,&exset,&t); if (sv == 1) process_command(read_a_line()); else if (sv == -1) no_more_input = TRUE; } static int getc_unbuffered(int fd) { // An attempted to get rid of the buffering char c; int n; n = read(fd,&c,1); if (n == 0) return EOF; else return c; } static char *read_a_line(void) { // read upto \n on stdin -- wonder if I should read instead int maxsize = 1024; char *line = walloc(char,maxsize+2); int i,c; for (i=0; (((c=getc_unbuffered(0)) != '\n') && (c != EOF)); i++) { if (i == maxsize) { char *nline = walloc(char,maxsize*2); memcpy(nline,line,maxsize); maxsize = maxsize*2; wfree(line); line = nline; } line[i] = c; } line[i] = '\n'; line[i+1] = '\0'; if (c == EOF) no_more_input = TRUE; if (strncmp(line,"close",5) != 0) { fprintf(stdout,"OK\n"); // give an acknowledgement fflush(stdout); } return line; } static void process_command(char *line) { // Process command, some are immediate EST_TokenStream ts; ts.open_string(line); EST_String comm = ts.get().string(); if ((comm == "quit") || (comm == "")) { no_more_input = TRUE; } else if (comm == "play") { EST_String file = ts.get().string(); int rate = atoi(ts.get().string()); Command *c = new Command(file,rate); command_queue.push(c); } else if (comm == "method") { play_wave_options.add_item("-p",ts.get().string()); } else if (comm == "command") { play_wave_options.add_item("-command",ts.get_upto_eoln().string()); } else if (comm == "rate") { play_wave_options.add_item("-rate",ts.get().string()); } else if (comm == "otype") { play_wave_options.add_item("-otype",ts.get().string()); } else if (comm == "device") { play_wave_options.add_item("-audiodevice",ts.get().string()); } else if (comm == "close") { pending_close = TRUE; no_more_input = TRUE; } else if (comm == "shutup") { // clear queue and kill and child currently playing command_queue.clear(); if (child_pid != 0) { kill(child_pid,SIGKILL); unlink(current_file); } } else if (comm == "query") command_queue.display(); else if (comm != "") { cerr << "audsp: unknown command \"" << comm << "\"\n"; } ts.close(); wfree(line); } static void check_new_output(void) { // If we are not waiting on any children lauch next command int pid; int statusp; if (kids > 0) { pid = waitpid(0,&statusp,WNOHANG); if (pid != 0) { kids--; child_pid = 0; } } else if (command_queue.length() != 0) { Command *c = command_queue.pop(); if (execute_command(c) == 0) kids++; delete c; } // else do nothing } static int execute_command(Command *c) { // Execute the command as a child process int pid; current_file = c->file(); if ((pid=fork()) == 0) { // child process load_play_file(c); _exit(0); // don't close any files on exit return 0; // can't get here } else if (pid > 0) { // parent process child_pid = pid; return 0; } else { cerr << "auspd: fork failed, \"" << c->file() << "\"\n"; return -1; } } static void load_play_file(Command *c) { // Load in wave file and play it EST_Wave w; w.load(c->file()); play_wave(w,play_wave_options); unlink(c->file()); // delete it afterwards } #endif festival/src/main/festival_client.cc0000644000175000017500000003314610126474330016203 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1996,1997 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black */ /* Date : December 1996 */ /*-----------------------------------------------------------------------*/ /* */ /* Client program used to send comands/data to a festival server */ /* */ /*=======================================================================*/ #include using namespace std; #include "EST_unix.h" #include "festival.h" #ifdef WIN32 typedef HANDLE SERVER_FD; #else typedef FILE *SERVER_FD; #endif static void festival_client_main(int argc, char **argv); static void copy_to_server(FILE *fdin,SERVER_FD serverfd); static void ttw_file(SERVER_FD serverfd, const EST_String &file); static void client_accept_waveform(SERVER_FD fd); static void client_accept_s_expr(SERVER_FD fd); static void new_state(int c, int &state, int &bdepth); static EST_String output_filename = "-"; static EST_String output_type = "riff"; static EST_String tts_mode = "nil"; static EST_String prolog = ""; static int withlisp = FALSE; static EST_String aucommand = ""; static int async_mode = FALSE; // So that festival_error works (and I don't need the whole of libFestival.a) void festival_tidy_up() { return; } int main(int argc, char **argv) { festival_client_main(argc,argv); return 0; } static void festival_client_main(int argc, char **argv) { EST_Option al; EST_StrList files; EST_String server; int port; FILE *infd; parse_command_line(argc, argv, EST_String("Usage:\n")+ "festival_client ...\n"+ "Access to festival server process\n"+ "--server hostname (or IP number) of server\n"+ "--port {1314} port number of server process (1314)\n"+ "--output file to save output waveform to\n"+ "--otype {riff}\n" " output type for waveform\n"+ "--passwd server passwd in plain text (optional)\n"+ "--prolog filename containing commands to be sent\n"+ " to the server before standard commands\n"+ " (useful when using --ttw)\n"+ "--async Asynchronous mode, server may send back\n"+ " multiple waveforms per text file\n"+ "--ttw Text to waveform: take text from first\n"+ " arg or stdin get server to return\n"+ " waveform(s) stored in output or operated\n"+ " on by aucommand.\n"+ "--withlisp Output lisp replies from server.\n"+ "--tts_mode TTS mode for file (default is fundamental).\n"+ "--aucommand \n"+ " command to be applied to each\n"+ " waveform retruned from server. Use $FILE\n"+ " in string to refer to waveform file\n", files, al); if (al.present("--server")) server = al.val("--server"); else server = "localhost"; if (al.present("--port")) port = al.ival("--port"); else port = FESTIVAL_DEFAULT_PORT; if (al.present("--tts_mode")) tts_mode = al.val("--tts_mode"); if (al.present("--output")) output_filename = al.val("--output"); if (al.present("--otype")) { output_type = al.val("--otype"); if (!output_type.matches(RXalphanum)) { cerr << "festival_client: invalid output type \"" << output_type << "\"" << endl; exit(-1); } } // Specify what to do with received waveform if (al.present("--aucommand")) aucommand = al.val("--aucommand"); if (al.present("--withlisp")) withlisp = TRUE; if (al.present("--async")) async_mode = TRUE; else async_mode = FALSE; int fd = festival_socket_client(server,port); #ifdef WIN32 HANDLE serverfd = (HANDLE)fd; #else FILE *serverfd = fdopen(fd,"wb"); #endif if (al.present("--passwd")) #ifdef WIN32 { DWORD bytes_written; DWORD passwdlen = (DWORD)strlen(al.val("--passwd"))+1; char *buffer = new char[passwdlen]; sprintf(buffer,"%s\n",al.val("--passwd")); bytes_written = send((SOCKET)serverfd,buffer,passwdlen,0); if (SOCKET_ERROR == bytes_written || bytes_written < passwdlen) { GetLastError(); // at least get the error cerr << "festival_client: can't send password to server\n"; } delete [] buffer; } #else fprintf(serverfd,"%s\n",(const char *)al.val("--passwd")); #endif if (al.present("--prolog")) { FILE *pfd = fopen(al.val("--prolog"),"rb"); if (pfd == NULL) { cerr << "festival_client: can't open prolog file \"" << al.val("--prolog") << "\"" << endl; exit(-1); } copy_to_server(pfd,serverfd); fclose(pfd); } if (al.present("--ttw")) ttw_file(serverfd,files.nth(0)); else { if ((files.length() == 0) || (files.nth(0) == "-")) copy_to_server(stdin,serverfd); else { if ((infd=fopen(files.nth(0),"rb")) == NULL) { cerr << "festival_client: can't open \"" << files.nth(0) << "\"\n"; exit(-1); } copy_to_server(infd,serverfd); } } return; } static void ttw_file(SERVER_FD serverfd, const EST_String &file) { // text to waveform file. This includes the tts wraparounds for // the text in file and outputs a waveform in output_filename // This is done as *one* waveform. This is designed for short // dialog type examples. If you need spooling this isn't the // way to do it EST_String tmpfile = make_tmp_filename(); FILE *fd, *tfd; int c; if ((fd=fopen(tmpfile,"wb")) == NULL) { cerr << "festival_client: can't open tmpfile \"" << tmpfile << "\"\n"; exit(-1); } // Here we ask for NIST because its a byte order aware headered format // the eventual desired format might be unheadered and if we asked the // the server for that we wouldn't know if it required byte swap or // not. The returned wave data from the server is actually saved // to a file and read in by EST_Wave::load so NIST is a safe option // Of course when the wave is saved by the client the requested // format is respected. fprintf(fd,"(Parameter.set 'Wavefiletype 'nist)\n"); if (async_mode) { // In async mode we need to set up tts_hooks to send back the waves fprintf(fd,"(tts_return_to_client)\n"); fprintf(fd,"(tts_text \"\n"); } else // do it in one go fprintf(fd,"(tts_textall \"\n"); if (file == "-") tfd = stdin; else if ((tfd=fopen(file,"rb")) == NULL) { cerr << "festival_client: can't open text file \"" << file << "\"\n"; exit(-1); } while ((c=getc(tfd)) != EOF) { if ((c == '"') || (c == '\\')) putc('\\',fd); putc(c,fd); } if (file != "-") fclose(tfd); fprintf(fd,"\" \"%s\")\n",(const char *)tts_mode); fclose(fd); // Now send the file to the server if ((fd=fopen(tmpfile,"rb")) == NULL) { cerr << "festival_client: tmpfile \"" << tmpfile << "\" mysteriously disappeared\n"; exit(-1); } copy_to_server(fd,serverfd); fclose(fd); unlink(tmpfile); } static void copy_to_server(FILE *fdin,SERVER_FD serverfd) { // Open a connection and copy everything from stdin to // server int c,n; int state=0; int bdepth=0; char ack[4]; while((c=getc(fdin)) != EOF) { #ifdef WIN32 DWORD bytes_pending; n = send((SOCKET)serverfd,(const char *)&c,1,0); if (SOCKET_ERROR == n || 0 == n) { if (SOCKET_ERROR == n) GetLastError(); cerr << "festival_client: couldn't copy to server\n"; } #else putc(c,serverfd); #endif new_state(c,state,bdepth); if (state == 1) { state = 0; #ifndef WIN32 fflush(serverfd); #endif do { #ifdef WIN32 { for (n=0; n < 3; ) { int bytes_read = recv((SOCKET)serverfd,ack+n,3-n,0); if (SOCKET_ERROR == bytes_read) { GetLastError(); cerr << "festival_client: error reading from server\n"; } else n+= bytes_read; } } #else for (n=0; n < 3; ) n += read(fileno(serverfd),ack+n,3-n); #endif ack[3] = '\0'; if (streq(ack,"WV\n")) // I've been sent a waveform client_accept_waveform(serverfd); else if (streq(ack,"LP\n")) // I've been sent an s-expr { client_accept_s_expr(serverfd); } else if (streq(ack,"ER\n")) { cerr << "festival server error: reset to top level\n"; break; } } while (!streq(ack,"OK\n")); } } } static void new_state(int c, int &state, int &bdepth) { // FSM (plus depth) to detect end of s-expr if (state == 0) { if ((c == ' ') || (c == '\t') || (c == '\n') || (c == '\r')) state = 0; else if (c == '\\') // escaped character state = 2; else if (c == ';') state = 3; // comment else if (c == '"') state = 4; // quoted string else if (c == '(') { bdepth++; state = 5; } else state = 5; // in s-expr } else if (state == 2) state = 5; // escaped character else if (state == 3) { if (c == '\n') state = 5; else state = 3; } else if (state == 4) { if (c == '\\') state = 6; else if (c == '"') state = 5; else state = 4; } else if (state == 6) state = 4; else if (state == 5) { if ((c == ' ') || (c == '\t') || (c == '\n') || (c == '\r')) { if (bdepth == 0) state = 1; else state = 5; } else if (c == '\\') // escaped character state = 2; else if (c == ';') state = 3; // comment else if (c == '"') state = 4; // quoted string else if (c == '(') { bdepth++; state = 5; } else if (c == ')') { bdepth--; state = 5; } else state = 5; // in s-expr } else // shouldn't get here state = 5; } static void client_accept_waveform(SERVER_FD fd) { // Read a waveform from fd. The waveform will be passed // using EST_String tmpfile = make_tmp_filename(); EST_Wave sig; // Have to copy this to a temporary file, then load it. #ifdef WIN32 socket_receive_file((SOCKET)fd,tmpfile); #else socket_receive_file(fileno(fd),tmpfile); #endif sig.load(tmpfile); if (aucommand != "") { // apply the command to this file EST_String tmpfile2 = make_tmp_filename(); sig.save(tmpfile2,output_type); char *command = walloc(char,1024+tmpfile2.length()+aucommand.length()); sprintf(command,"FILE=\"%s\"; %s",(const char *)tmpfile2, (const char *)aucommand); system(command); unlink(tmpfile2); } else if (output_filename == "") cerr << "festival_client: ignoring received waveform, no output file" << endl; else sig.save(output_filename,output_type); unlink(tmpfile); } static void client_accept_s_expr(SERVER_FD fd) { // Read an s-expression. Inefficeintly into a file EST_String tmpfile = make_tmp_filename(); FILE *tf; int c; // Have to copy this to a temporary file, then load it. #ifdef WIN32 socket_receive_file((SOCKET)fd,tmpfile); #else socket_receive_file(fileno(fd),tmpfile); #endif if (withlisp) { if (( tf = fopen(tmpfile,"rb")) == NULL) { cerr << "festival_client: lost an s_expr tmp file" << endl; } else { while ((c=getc(tf)) != EOF) putc(c,stdout); fclose(tf); } } unlink(tmpfile); } festival/src/lib/Makefile0000644000175000017500000000477307262605650014017 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../.. DIRNAME=src/lib ALL_LIBS = FILES=Makefile LOCAL_CLEAN = libFestival.a include $(TOP)/config/common_make_rules festival/src/scripts/Makefile0000644000175000017500000000631407262605650014731 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Makfile for script directory ## ## --------------------------------------------------------------------- ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## ## ########################################################################### TOP=../.. DIRNAME=src/scripts SCRIPTS= festival_server.sh festival_server_control.sh EXTRA_SCRIPTS = jsapi_example.sh festival_client_java.sh FILES = Makefile shared_setup_sh shared_setup_prl shared_script $(SCRIPTS) $(EXTRA_SCRIPTS) INSTALL = ALL = $(SCRIPTS) include $(TOP)/config/common_make_rules include $(EST)/config/rules/bin_process.mak festival/src/scripts/shared_setup_sh0000644000175000017500000000036407262605650016373 0ustar awbawb # festival shared setup extend() { var="$1" extra="$2" eval "val=\$$var" if [ -n "$val" ] then val="$extra:$val" else val="$extra" fi eval "$var='$val'" eval "export $var" } extend LD_LIBRARY_PATH "__EST__/lib:__LDPATH__" festival/src/scripts/shared_setup_prl0000644000175000017500000000025107262605650016551 0ustar awbawb if (defined($ENV{LD_LIBRARY_PATH})) { $ENV{LD_LIBRARY_PATH} = "__TOP__/lib:__LDPATH__:$ENV{LD_LIBRARY_PATH}"; } else { $ENV{LD_LIBRARY_PATH} = "__TOP__/lib"; } festival/src/scripts/shared_script0000644000175000017500000000044011337203617016033 0ustar awbawb#!/bin/sh # Festival shared script extend() { var="$1" extra="$2" eval "val=\$$var" if [ -n "$val" ] then val="$extra:$val" else val="$extra" fi eval "$var='$val'" eval "export $var" } extend __LDVAR__ "__EST__/lib:__LDPATH__" exec __MAIN__/__PROGRAM__ "$@" exit 0 festival/src/scripts/festival_server.sh0000644000175000017500000001566712114677707017047 0ustar awbawb#!/bin/sh ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### # # # Run a festival server. # # # ########################################################################### TOP=__TOP__ EST=__EST__ #__SHARED_SETUP__ useage() { cat <&3 } handle_hup() { log_mess "got SIGHUP" respawn_wait=0; if [ $festival_pid = 0 ] then : else kill -9 $festival_pid festival_pid=0 fi } handle_term() { log_mess "got SIGTERM" respawn=false; respawn_wait=0; if [ $festival_pid = 0 ] then : else kill -9 $festival_pid festival_pid=0 fi } create_server_startup() { local port server_log su port="$1" server_log="$2" server_startup="$3" { cat < $server_startup } port=1314 festival=festival logdir="." config="" show=false while [ $# -gt 0 ] do case "$1" in -h* | -\? ) useage 0;; -p* ) port=$2; shift 2;; -l* ) logdir=$2; shift 2;; -c* ) config=$2; shift 2;; -show* ) show=true; shift;; -* ) useage 1;; * ) break;; esac done case $# in 1 ) festival=$1;; 0 ) : ;; * ) useage 2;; esac respawn=true normal_respawn_wait=10 festival_pid=0 if [ $port = "1314" ] then pext="" else pext=".$port" fi server_log="$logdir/festival_server$pext.log" # hangup trap "handle_hup" 1 # int trap "handle_term" 2 # term (ie default kill signal) trap "handle_term" 15 # exit from shell trap "handle_term" 0 if $show then create_server_startup $port $server_log /tmp/$$ 3>/dev/null fl=false while read l do if $fl ; then echo $l ; fi if [ "$l" = ";---" ] ; then fl=true ; fi done "$logdir/festival_wrapper_pid$pext" while $respawn do respawn_wait=$normal_respawn_wait if [ $festival_pid = 0 ] then : else kill -9 $festival_pid festival_pid=0 fi log_mess "run $festival port=$port" create_server_startup $port $server_log "$logdir/festival_server$pext.scm" $festival --server $server_startup & x=$! if [ -n "$x" ] then festival_pid="$x" log_mess "waiting" while [ $festival_pid -ne 0 ] && kill -0 $festival_pid do sleep 60& echo $! > "$logdir/festival_sleep_pid$pext" wait $! done if [ "$festival_pid" = 0 ] then : else wait $festival_pid fi log_mess "festival exited $?" else log_mess "can't run $festival" fi if $respawn then log_mess "respawn wait $respawn_wait seconds" sleep $respawn_wait fi done } >$server_log 3>$server_log exit 0 festival/src/scripts/festival_server_control.sh0000644000175000017500000000775607262605650020603 0ustar awbawb#!/bin/sh ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### # # # Run a festival server. # # # ########################################################################### TOP=__TOP__ EST=__EST__ #__SHARED_SETUP__ useage() { cat </dev/null then case "$action" in restart ) kill -HUP $wrapper_pid;; kill|quit|exit ) kill -TERM $wrapper_pid;; esac kill -TERM $sleep_pid else echo "Server wrapper not running for port $port" fi } exit 0 festival/src/scripts/jsapi_example.sh0000644000175000017500000000735107262605650016450 0ustar awbawb#!/bin/sh ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=__TOP__ EST=__EST__ CLASSPATH=__CLASSPATH__ JAVA_HOME=__JAVA_HOME__ #__SHARED_SETUP__ useage() { cat < $* @chmod a-w $* festival/lib/VCLocalRules0000644000175000017500000000023007262605650013771 0ustar awbawbSIOD = siod.scm web.scm cstr.scm fringe.scm .copy_from_est: $(SIOD) $(SIOD) : @echo 'Copy $@ from EST/lib/siod' -del $@ copy $(EST)\lib\siod\$@ $@festival/lib/mrpa_phones.scm0000644000175000017500000001145207262605650014577 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; A definition of the mrpa phone set ;; (defPhoneSet mrpa ;;; Phone Features (;; vowel or consonant (vc + -) ;; vowel length: short long dipthong schwa (vlng s l d a 0) ;; vowel height: high mid low (vheight 1 2 3 0) ;; vowel frontness: front mid back (vfront 1 2 3 0) ;; lip rounding (vrnd + - 0) ;; consonant type: stop fricative affricate nasal lateral approximant (ctype s f a n l r 0) ;; place of articulation: labial alveolar palatal labio-dental ;; dental velar glottal (cplace l a p b d v g 0) ;; consonant voicing (cvox + - 0) ) ;; Phone set members ( (uh + s 2 3 - 0 0 0) (e + s 2 1 - 0 0 0) (a + s 3 1 - 0 0 0) (o + s 2 3 + 0 0 0) (i + s 1 1 - 0 0 0) (u + s 1 3 + 0 0 0) (ii + l 1 1 - 0 0 0) (uu + l 1 3 + 0 0 0) (oo + l 3 3 + 0 0 0) (aa + l 3 3 - 0 0 0) (@@ + l 2 2 - 0 0 0) (ai + d 3 2 - 0 0 0) (ei + d 2 1 - 0 0 0) (oi + d 3 3 + 0 0 0) (au + d 3 2 + 0 0 0) (ou + d 2 2 - 0 0 0) (e@ + d 2 1 - 0 0 0) (i@ + d 1 1 - 0 0 0) (u@ + d 3 1 + 0 0 0) (@ + a 2 2 - 0 0 0) (p - 0 0 0 0 s l -) (t - 0 0 0 0 s a -) (k - 0 0 0 0 s v -) (b - 0 0 0 0 s l +) (d - 0 0 0 0 s a +) (g - 0 0 0 0 s v +) (s - 0 0 0 0 f a -) (z - 0 0 0 0 f a +) (sh - 0 0 0 0 f p -) (zh - 0 0 0 0 f p +) (f - 0 0 0 0 f b -) (v - 0 0 0 0 f b +) (th - 0 0 0 0 f d -) (dh - 0 0 0 0 f d +) (ch - 0 0 0 0 a p -) (jh - 0 0 0 0 a p +) (h - 0 0 0 0 f g -) (m - 0 0 0 0 n l +) (n - 0 0 0 0 n a +) (ng - 0 0 0 0 n v +) (l - 0 0 0 0 l a +) (y - 0 0 0 0 r p +) (r - 0 0 0 0 r a +) (w - 0 0 0 0 r l +) (# - 0 0 0 0 0 0 -) ) ) (PhoneSet.silences '(#)) (provide 'mrpa_phones) festival/lib/mrpa_allophones.scm0000644000175000017500000001154507262605650015452 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; A definition of the extended mrpa phone set used for some diphone sets ;; (defPhoneSet mrpa_allophones ;;; Phone Features (;; vowel or consonant (vc + -) ;; vowel length: short long dipthong schwa (vlng s l d a 0) ;; vowel height: high mid low (vheight 1 2 3 -) ;; vowel frontness: front mid back (vfront 1 2 3 -) ;; lip rounding (vrnd + -) ;; consonant type: stop fricative affricative nasal liquid (ctype s f a n l 0) ;; place of articulation: labial alveolar palatal labio-dental ;; dental velar (cplace l a p b d v 0) ;; consonant voicing (cvox + -) ) ;; Phone set members ( (uh + s 2 3 - 0 0 +) (e + s 2 1 - 0 0 +) (a + s 3 1 - 0 0 +) (o + s 3 3 - 0 0 +) (i + s 1 1 - 0 0 +) (u + s 1 3 + 0 0 +) (ii + l 1 1 - 0 0 +) (uu + l 2 3 + 0 0 +) (oo + l 3 2 - 0 0 +) (aa + l 3 1 - 0 0 +) (@@ + l 2 2 - 0 0 +) (ai + d 3 1 - 0 0 +) (ei + d 2 1 - 0 0 +) (oi + d 3 3 - 0 0 +) (au + d 3 3 + 0 0 +) (ou + d 3 3 + 0 0 +) (e@ + d 2 1 - 0 0 +) (i@ + d 1 1 - 0 0 +) (u@ + d 3 1 - 0 0 +) (@ + a - - - 0 0 +) (p - 0 - - + s l -) (t - 0 - - + s a -) (k - 0 - - + s p -) (b - 0 - - + s l +) (d - 0 - - + s a +) (g - 0 - - + s p +) (s - 0 - - + f a -) (z - 0 - - + f a +) (sh - 0 - - + f p -) (zh - 0 - - + f p +) (f - 0 - - + f b -) (v - 0 - - + f b +) (th - 0 - - + f d -) (dh - 0 - - + f d +) (ch - 0 - - + a a -) (jh - 0 - - + a a +) (h - 0 - - + a v -) (m - 0 - - + n l +) (n - 0 - - + n d +) (ng - 0 - - + n v +) (l - 0 - - + l d +) (ll - 0 - - + l d +) (y - 0 - - + l a +) (r - 0 - - + l p +) (w - 0 - - + l l +) (# - 0 - - - 0 0 -) ) ) (PhoneSet.silences '(#)) (provide 'mrpa_allophones) festival/lib/radio_phones.scm0000644000175000017500000001256607262605650014745 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; A definition of the radio phone set used in the BU RADIO FM ;;; corpus, some people call this the darpa set. This one ;;; has the closures removed ;;; (defPhoneSet radio ;;; Phone Features (;; vowel or consonant (vc + -) ;; vowel length: short long dipthong schwa (vlng s l d a 0) ;; vowel height: high mid low (vheight 1 2 3 0) ;; vowel frontness: front mid back (vfront 1 2 3 0) ;; lip rounding (vrnd + - 0) ;; consonant type: stop fricative affricate nasal lateral approximant (ctype s f a n l r 0) ;; place of articulation: labial alveolar palatal labio-dental ;; dental velar glottal (cplace l a p b d v g 0) ;; consonant voicing (cvox + - 0) ) ;; Phone set members ( ;; Note these features were set by awb so they are wrong !!! (aa + l 3 3 - 0 0 0) ;; father (ae + s 3 1 - 0 0 0) ;; fat (ah + s 2 2 - 0 0 0) ;; but (ao + l 3 3 + 0 0 0) ;; lawn (aw + d 3 2 - 0 0 0) ;; how (ax + a 2 2 - 0 0 0) ;; about (axr + a 2 2 - r a +) (ay + d 3 2 - 0 0 0) ;; hide (b - 0 0 0 0 s l +) (ch - 0 0 0 0 a p -) (d - 0 0 0 0 s a +) (dh - 0 0 0 0 f d +) (dx - a 0 0 0 s a +) ;; ?? (eh + s 2 1 - 0 0 0) ;; get (el + s 0 0 0 l a +) (em + s 0 0 0 n l +) (en + s 0 0 0 n a +) (er + a 2 2 - r 0 0) ;; always followed by r (er-r == axr) (ey + d 2 1 - 0 0 0) ;; gate (f - 0 0 0 0 f b -) (g - 0 0 0 0 s v +) (hh - 0 0 0 0 f g -) (hv - 0 0 0 0 f g +) (ih + s 1 1 - 0 0 0) ;; bit (iy + l 1 1 - 0 0 0) ;; beet (jh - 0 0 0 0 a p +) (k - 0 0 0 0 s v -) (l - 0 0 0 0 l a +) (m - 0 0 0 0 n l +) (n - 0 0 0 0 n a +) (nx - 0 0 0 0 n d +) ;; ??? (ng - 0 0 0 0 n v +) (ow + d 2 3 + 0 0 0) ;; lone (oy + d 2 3 + 0 0 0) ;; toy (p - 0 0 0 0 s l -) (r - 0 0 0 0 r a +) (s - 0 0 0 0 f a -) (sh - 0 0 0 0 f p -) (t - 0 0 0 0 s a -) (th - 0 0 0 0 f d -) (uh + s 1 3 + 0 0 0) ;; full (uw + l 1 3 + 0 0 0) ;; fool (v - 0 0 0 0 f b +) (w - 0 0 0 0 r l +) (y - 0 0 0 0 r p +) (z - 0 0 0 0 f a +) (zh - 0 0 0 0 f p +) (pau - 0 0 0 0 0 0 -) (h# - 0 0 0 0 0 0 -) (brth - 0 0 0 0 0 0 -) ) ) (PhoneSet.silences '(pau h# brth)) (provide 'radio_phones) festival/lib/holmes_phones.scm0000644000175000017500000001250707262605650015131 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; A definition of the Holmes phone set used by the Donovan LPC ;; diphone synthesizer, the rest of the synthesis process will ;; typically use mrpa phones and map to these. ;; ;; Hmm not sure I've got the right mapping (as usual) (defPhoneSet holmes ;;; Phone Features (;; vowel or consonant (vc + -) ;; vowel length: short long dipthong schwa (vlng s l d a 0) ;; vowel height: high mid low (vheight 1 2 3 - 0) ;; vowel frontness: front mid back (vfront 1 2 3 - 0) ;; lip rounding (vrnd + - 0) ;; consonant type: stop fricative affricative nasal lateral approximant (ctype s f a n l r 0) ;; place of articulation: labial alveolar palatal labio-dental ;; dental velar glottal (cplace l a p b d v g 0) ;; consonant voicing (cvox + - 0) ) ;; Phone set members ( ;; Note these features were set by awb so they are wrong !!! (ee + l 1 1 - 0 0 0) ;; beet (i + s 1 1 - 0 0 0) ;; bit (ai + d 2 1 - 0 0 0) ;; gate (e + s 2 1 - 0 0 0) ;; get (aa + s 3 1 - 0 0 0) ;; fat (ar + l 3 3 - 0 0 0) ;; father (aw + l 3 3 + 0 0 0) ;; lawn (oa + d 2 2 - 0 0 0) ;; lone (oo + s 1 3 + 0 0 0) ;; full (uu + l 1 3 + 0 0 0) ;; fool (o + s 2 3 + 0 0 0) (er + l 2 2 - 0 0 0) ;; murder (a + a 2 2 - 0 0 0) ;; about (u + s 2 3 - 0 0 0) ;; but (ie + d 3 2 - 0 0 0) ;; hide (ou + d 3 2 + 0 0 0) ;; how (oi + d 3 3 + 0 0 0) ;; toy (eer + d 2 1 - 0 0 0) (air + d 1 1 - 0 0 0) (oor + d 3 1 + 0 0 0) ;; (yu + l 2 3 + 0 0 +) ;; you ??? (p - 0 0 0 0 s l -) (b - 0 0 0 0 s l +) (t - 0 0 0 0 s a -) (d - 0 0 0 0 s a +) (k - 0 0 0 0 s v -) (g - 0 0 0 0 s v +) (f - 0 0 0 0 f b -) (v - 0 0 0 0 f b +) (th - 0 0 0 0 f d -) (dh - 0 0 0 0 f d +) (s - 0 0 0 0 f a -) (z - 0 0 0 0 f a +) (sh - 0 0 0 0 f p -) (zh - 0 0 0 0 f p +) (h - 0 0 0 0 f g -) (m - 0 0 0 0 n l +) (n - 0 0 0 0 n a +) (ng - 0 0 0 0 n v +) (ch - 0 0 0 0 a p -) (j - 0 0 0 0 a p +) (l - 0 0 0 0 l a +) (w - 0 0 0 0 r l +) (y - 0 0 0 0 r p +) (r - 0 0 0 0 r a +) ;; (wh - 0 - - + l l -) ;; ?? ;; (wh - 0 - - + l l +) ;; map to w (# - 0 0 0 0 0 0 -) ) ) (PhoneSet.silences '(#)) (provide 'holmes_phones) festival/lib/darpa_phones.scm0000644000175000017500000001232107262605650014723 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1999 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: April 1999 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; (yet another) darpa definition ;;; (require 'phoneset) (set! darpa_fs (cadr (defPhoneSet darpa (Features (vowel (syllabic + -) (length long short diphthong schwa) (height high mid low) (front front mid back) (round + -)) (consonant (syllabic + -) (manner stop affricate fricative approximant nasal) (place alveolar dental labial palatal velar) (voicing + -)) (silence (syllabic -))) (Phones ;; type syl length height front round (aa vowel + long low back -) (ae vowel + short low front -) (ah vowel + short mid mid -) (ao vowel + long low front +) (aw vowel + diphthong low mid -) (ax vowel + schwa mid mid -) (axr vowel + schwa mid mid -) (ay vowel + diphthong low mid -) (eh vowel + short mid front -) (ey vowel + diphthong mid front -) (ih vowel + short high front -) (iy vowel + long high front -) (ow vowel + diphthong mid back +) (oy vowel + diphthong mid back +) (uh vowel + short high back +) (uw vowel + long high back +) ;; type syl manner place voicing (b consonant - stop labial +) (ch consonant - affricate alveolar -) (d consonant - stop alveolar +) (dh consonant - fricative dental +) (dx consonant - stop alveolar +) (el consonant + approximant alveolar +) (em consonant + nasal labial +) (en consonant + stop alveolar +) (er consonant + approximant alveolar +) (f consonant - fricative labial -) (g consonant - stop velar +) (hh consonant - fricative velar -) (jh consonant - affricate alveolar +) (k consonant - stop velar -) (l consonant - approximant alveolar +) (m consonant - nasal labial +) (n consonant - nasal alveolar +) (nx consonant - nasal alveolar +) (ng consonant - nasal velar +) (p consonant - stop labial -) (r consonant - approximant alveolar +) (s consonant - fricative alveolar -) (sh consonant - fricative palatal -) (t consonant - stop alveolar -) (th consonant - fricative dental -) (v consonant - fricative labial +) (w consonant - approximant velar +) (y consonant - approximant palatal +) (z consonant - fricative alveolar +) (zh consonant - fricative palatal +) (pau silence -) ; (sil silence -) )))) (provide 'darpa_phones) festival/lib/phoneset.scm0000644000175000017500000001234107262605650014107 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1999 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: April 1999 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Support code for phone set definitions ;;; (defmac (defPhoneSet form) (list 'defPhoneSet_real (list 'quote (cadr form)) (list 'quote (car (cddr form))) (list 'quote (cadr (cddr form))))) (define (defPhoneSet_real name featdefs phones) "(defPhoneSet NAME FEATTYPES PHONES) Define a phone set with given name, feature types and list of phones. This also selects name as the current phoneset." (let (info) (if (not (eq? 'Features (car featdefs))) (begin ;; Old format that has the same number of phone features for ;; all phones (set! info (mapcar (lambda (ph) (let ((fvs (mapcar list (mapcar car featdefs) (cdr ph)))) (ps_check_fvals (cons (car ph) (cons (list 'type t) fvs)) (cons t fvs)) (list (car ph) fvs))) phones))) ;; else ;; New format where types are specified so phones may have ;; different features (set! info (mapcar (lambda (ph) (let ((fvs (cons (list 'type (cadr ph)) (mapcar list (mapcar car (cdr (assoc (cadr ph) (cdr featdefs)))) (cddr ph))))) (ps_check_fvals (cons (car ph) fvs) (assoc (cadr ph) (cdr featdefs))) (list (car ph) fvs))) (cdr phones)))) (Param.set (string-append "phonesets." name) info) (PhoneSet.select name) (list name info))) (define (ps_check_fvals fvs featdefs) "(ps_check_fvals fvs featdefs) Check that feature values in a phone definition are in the defined set of possibles." (mapcar (lambda (fp) (let ((def (cdr (assoc (car fp) (cdr featdefs))))) (cond ((not def) (error "Phoneset definition: phone has no defined type" fvs)) ((not (member_string (car (cdr fp)) def)) (error (format nil "Phoneset definition: phone feature %l is undefined" fp) fvs))))) (cdr (cdr fvs)))) (define (PhoneSet.select name) "(PhoneSet.select name) Select named phonset as current." (if (feats.present Param (string-append "phonesets." name)) (Param.set "phoneset" (Param.get (string-append "phonesets." name))) (error "no phoneset defined: " name))) (define (PhoneSet.description name) "(PhoneSet.description) Return (lisp) representation of current phoneset." (feats.tolisp (Param.get "phoneset"))) (define (PhoneSet.list) "(PhoneSet.list) List of the names of the currently defined phonesets." ;; This isn't a particularly efficient way to get the answer (mapcar car (feats.tolisp (Param.get "phonesets")))) (define (PhoneSet.silences sils) "(PhoneSet.silences SILLIST) Define the silence phones for the currently selected phoneset." (Param.set "phoneset.silences" sils)) (provide 'phoneset) festival/lib/cmusphinx2_phones.scm0000644000175000017500000001222707275000671015735 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; ;;; Carnegie Mellon University ;;; ;;; and Alan W Black and Kevin Lenzo ;;; ;;; Copyright (c) 1998-2000 ;;; ;;; All Rights Reserved. ;;; ;;; ;;; ;;; Permission is hereby granted, free of charge, to use and distribute ;;; ;;; this software and its documentation without restriction, including ;;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;;; ;;; permit persons to whom this work is furnished to do so, subject to ;;; ;;; the following conditions: ;;; ;;; 1. The code must retain the above copyright notice, this list of ;;; ;;; conditions and the following disclaimer. ;;; ;;; 2. Any modifications must be clearly marked as such. ;;; ;;; 3. Original authors' names are not deleted. ;;; ;;; 4. The authors' names are not used to endorse or promote products ;;; ;;; derived from this software without specific prior written ;;; ;;; permission. ;;; ;;; ;;; ;;; CARNEGIE MELLON UNIVERSITY AND THE CONTRIBUTORS TO THIS WORK ;;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;;; ;;; SHALL CARNEGIE MELLON UNIVERSITY NOR THE CONTRIBUTORS BE LIABLE ;;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;;; ;;; THIS SOFTWARE. ;;; ;;; ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; A definition of the cmusphinx2 phone set used in the BU RADIO FM ;;; corpus, some people call this the darpa set. This one ;;; has the closures removed ;;; (defPhoneSet cmusphinx2 ;;; Phone Features (;; vowel or consonant (vc + -) ;; vowel length: short long dipthong schwa (vlng s l d a 0) ;; vowel height: high mid low (vheight 1 2 3 0) ;; vowel frontness: front mid back (vfront 1 2 3 0) ;; lip rounding (vrnd + - 0) ;; consonant type: stop fricative affricate nasal lateral approximant (ctype s f a n l r 0) ;; place of articulation: labial alveolar palatal labio-dental ;; dental velar glottal (cplace l a p b d v g 0) ;; consonant voicing (cvox + - 0) ) ;; Phone set members ( ;; Note these features were set by awb so they are wrong !!! ; phone vc vl vh vf vr ct cp cv (AA + l 3 3 - 0 0 0) ;; father (AE + s 3 1 - 0 0 0) ;; fat (AH + s 2 2 - 0 0 0) ;; but (AO + l 3 3 + 0 0 0) ;; lawn (AW + d 3 2 - 0 0 0) ;; how (AX + a 2 2 - 0 0 0) ;; about (AXR + a 2 2 - r a +) (AY + d 3 2 - 0 0 0) ;; hide (B - 0 0 0 0 s l +) (CH - 0 0 0 0 a p -) (D - 0 0 0 0 s a +) (DH - 0 0 0 0 f d +) (DX - 0 0 0 0 s a +) (EH + s 2 1 - 0 0 0) ;; get (ER + a 2 2 - r 0 0) (EY + d 2 1 - 0 0 0) ;; gate (F - 0 0 0 0 f b -) (G - 0 0 0 0 s v +) (HH - 0 0 0 0 f g -) (IH + s 1 1 - 0 0 0) ;; bit (IY + l 1 1 - 0 0 0) ;; beet (JH - 0 0 0 0 a p +) (K - 0 0 0 0 s v -) (L - 0 0 0 0 l a +) (M - 0 0 0 0 n l +) (N - 0 0 0 0 n a +) (NG - 0 0 0 0 n v +) (OW + d 2 3 + 0 0 0) ;; lone (OY + d 2 3 + 0 0 0) ;; toy (P - 0 0 0 0 s l -) (R - 0 0 0 0 r a +) (S - 0 0 0 0 f a -) (SH - 0 0 0 0 f p -) (T - 0 0 0 0 s a -) (TH - 0 0 0 0 f d -) (UH + s 1 3 + 0 0 0) ;; full (UW + l 1 3 + 0 0 0) ;; fool (V - 0 0 0 0 f b +) (W - 0 0 0 0 r l +) (Y - 0 0 0 0 r p +) (Z - 0 0 0 0 f a +) (ZH - 0 0 0 0 f p +) (SIL - 0 0 0 0 0 0 -) ; added ) ) (PhoneSet.silences '(SIL)) (provide 'cmusphinx2_phones) festival/lib/unilex_phones.scm0000644000175000017500000001735210231477243015144 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 2003, 2004 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; unilex phoneset ;;; (defPhoneSet unilex ;;; Phone Features (;; vowel or consonant (vc + -) ;; vowel length: short long dipthong schwa (vlng s l d a 0) ;; vowel height: high mid low (vheight 1 2 3 0) ;; vowel frontness: front mid back (vfront 1 2 3 0) ;; lip rounding (vrnd + - 0) ;; consonant type: stop fricative affricative nasal liquid approximant (ctype s f a n l t r 0) ;; place of articulation: labial alveolar palatal labio-dental ;; dental velar glottal (cplace l a p b d v g 0) ;; consonant voicing (cvox + - 0) ) ( (SIL - 0 0 0 0 0 0 -) ;; slience ... (# - 0 0 0 0 0 0 -) ;; slience ... (B_10 - 0 0 0 0 0 0 -) ;; Pauses (B_20 - 0 0 0 0 0 0 -) ;; Pauses (B_30 - 0 0 0 0 0 0 -) ;; Pauses (B_40 - 0 0 0 0 0 0 -) ;; Pauses (B_50 - 0 0 0 0 0 0 -) ;; Pauses (B_100 - 0 0 0 0 0 0 -) ;; Pauses (B_150 - 0 0 0 0 0 0 -) ;; Pauses (B_200 - 0 0 0 0 0 0 -) ;; Pauses (B_250 - 0 0 0 0 0 0 -) ;; Pauses (B_300 - 0 0 0 0 0 0 -) ;; Pauses (B_400 - 0 0 0 0 0 0 -) ;; Pauses (IGNORE - 0 0 0 0 0 0 -) ;; Pauses ;; insert the phones here, see examples in ;; festival/lib/*_phones.scm ;(name vc vling vheight vfront vrnd ctype cplace cvox) ;;; Rob guesed these values for Edinburgh English ;;; Not to be taken too seriously. (p - 0 0 0 0 s l -) (t - 0 0 0 0 s a -) (? - 0 0 0 0 s g +) ;;; ??? (t^ - 0 0 0 0 t a +) ;;; ??? (k - 0 0 0 0 s v -) (x - 0 0 0 0 f v -) (b - 0 0 0 0 s l +) (d - 0 0 0 0 s a +) (g - 0 0 0 0 s v +) (ch - 0 0 0 0 a p -) (jh - 0 0 0 0 a p +) (s - 0 0 0 0 f a -) (z - 0 0 0 0 f a +) (sh - 0 0 0 0 f p -) (zh - 0 0 0 0 f p +) (f - 0 0 0 0 f b -) (v - 0 0 0 0 f b +) (th - 0 0 0 0 f d -) (dh - 0 0 0 0 f d +) (h - 0 0 0 0 f 0 -) ;;; ??? (m - 0 0 0 0 n l +) (m! - 0 0 0 0 n l +) (n - 0 0 0 0 n a +) (n! - 0 0 0 0 n a +) (ng - 0 0 0 0 n v +) (l - 0 0 0 0 r a +) (ll - 0 0 0 0 r a +) (lw - 0 0 0 0 r a +) (l! - 0 0 0 0 r a +) (r - 0 0 0 0 r a +) (y - 0 0 0 0 l p +) (w - 0 0 0 0 l l +) (hw - 0 0 0 0 l l +) (e + s 2 1 - 0 0 0) (ao + s 3 1 - 0 0 0) (a + s 3 1 - 0 0 0) (ah + s 3 1 - 0 0 0) (oa + s 3 1 - 0 0 0) (aa + s 3 1 - 0 0 0) (ar + s 3 1 - 0 0 0) (eh + s 3 1 - 0 0 0) ;;; ? (oul + d 2 3 + 0 0 0) ;;; ? (ou + d 2 3 + 0 0 0) (ouw + d 2 3 + 0 0 0) (oou + l 3 3 + 0 0 0) (o + l 3 3 + 0 0 0) (au + l 3 3 + 0 0 0) (oo + l 3 3 + 0 0 0) (or + l 3 3 + 0 0 0) (our + d 2 3 + 0 0 0) (ii + l 1 1 - 0 0 0) (ihr + s 1 1 - 0 0 0) (iy + l 1 1 - 0 0 0) (i + s 1 1 - 0 0 0) (ie + l 1 1 - 0 0 0) ;;; ? (iii + s 1 1 - 0 0 0) ;;; was ii; (@r + a 2 2 - r a +) (@ + a 2 2 - 0 0 0) (uh + s 2 2 - 0 0 0) (uhr + s 2 2 - 0 0 0) (u + l 1 3 + 0 0 0) (uu + l 1 3 + 0 0 0) (iu + l 1 3 + 0 0 0) (uuu + l 1 3 + 0 0 0) ;;; was uu; (uw + l 1 3 + 0 0 0) ;;; ??? (uul + l 1 3 + 0 0 0) ;;; ??? (ei + d 2 1 - 0 0 0) (ee + d 2 1 - 0 0 0) (ai + d 3 2 - 0 0 0) ;;; ??? (ae + d 3 2 - 0 0 0) ;;; ??? (aer + d 3 2 - 0 0 0) ;;; ??? (aai + d 3 2 - 0 0 0) ;;; ??? (oi + d 2 3 + 0 0 0) ;;; ??? (oir + d 2 3 + 0 0 0) ;;; ??? (ow + d 3 2 - 0 0 0) (owr + d 3 2 - 0 0 0) ;;; ??? (oow + d 3 2 - 0 0 0) ;;; ??? (i@ + l 1 1 - 0 0 0) ;;; iy + @ ? (ir + s 1 1 - 0 0 0) (irr + s 1 1 - 0 0 0) ;;; was ir; (iir + s 1 1 - 0 0 0) (@@r + a 2 2 - 0 0 0) (er + s 2 1 - 0 0 0) (eir + s 2 1 - 0 0 0) ;;; ??? (ur + s 1 3 + 0 0 0) ;;; ??? (urr + s 1 3 + 0 0 0) ;;; ??? (iur + s 1 3 + 0 0 0) ;;; ??? ) ) (PhoneSet.silences '( # SIL)) (define (unilex::select_phoneset) "(unilex::select_phoneset) Set up phone set for unilex" (Parameter.set 'PhoneSet 'unilex) (PhoneSet.select 'unilex) ) (define (unilex::reset_phoneset) "(unilex::reset_phoneset) Reset phone set for unilex." t ) (provide 'unilex_phones) festival/lib/mrpa_durs.scm0000644000175000017500000000754707262605650014272 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; mrpa average phoneme durations from gsw 450 ;;; (set! phoneme_durations '( (u 0.067) (i@ 0.146) (h 0.067) (uu 0.105) (uh 0.090) (v 0.053) (oo 0.145) (i 0.060) (jh 0.097) (ii 0.095) (w 0.066) (k 0.088) (+ 0.036) (y 0.051) (l 0.067) (zh 0.080) (ng 0.072) (m 0.070) (z 0.079) (## 0.256) (au 0.162) (a 0.118) (n 0.065) (o 0.102) (ai 0.156) (b 0.071) (ou 0.129) (ch 0.119) (p 0.094) (oi 0.165) (# 0.040) (e@ 0.131) (d 0.052) (dh 0.032) (e 0.091) (r 0.062) (sh 0.101) (@@ 0.149) (ei 0.131) (f 0.091) (s 0.093) (g 0.066) (u@ 0.120) (aa 0.173) (t 0.073) (th 0.080) (@ 0.054) )) (set! gsw_durs '( (# 0.200 0.100) (h 0.061 0.028) (i@ 0.141 0.061) (u 0.067 0.024) (uu 0.107 0.044) (uh 0.087 0.025) (v 0.051 0.019) (oo 0.138 0.046) (i 0.058 0.023) (ii 0.092 0.035) (w 0.054 0.023) (jh 0.094 0.024) (k 0.089 0.034) (y 0.048 0.025) (l 0.056 0.026) (zh 0.077 0.030) (ng 0.064 0.024) (m 0.063 0.021) (z 0.072 0.029) (a 0.120 0.036) (au 0.171 0.046) (n 0.059 0.025) (ou 0.134 0.039) (b 0.073 0.021) (o 0.094 0.037) (ai 0.137 0.047) (ch 0.128 0.039) (oi 0.183 0.050) (p 0.101 0.032) (e@ 0.144 0.061) (d 0.048 0.021) (dh 0.031 0.016) (e 0.092 0.035) (r 0.053 0.025) (sh 0.108 0.031) (f 0.095 0.033) (@@ 0.147 0.035) (ei 0.130 0.042) (s 0.102 0.037) (u@ 0.140 0.057) (th 0.093 0.050) (g 0.064 0.021) (aa 0.155 0.045) (t 0.070 0.034) (@ 0.046 0.020) )) (provide 'mrpa_durs) festival/lib/klatt_durs.scm0000644000175000017500000000631107262605650014436 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Phone duration info for Klatt rules, for mrpa phone set (set! duration_klatt_params '( (a 230.0 80.0) (aa 240.0 100.0) (@ 120.0 60.0) (@@ 180.0 80.0) (ai 250.0 150.0) (au 240.0 100.0) (b 85.0 60.0) (ch 70.0 50.0) (d 75.0 50.0) (dh 50.0 30.0) (e 150.0 70.0) (e@ 270.0 130.0) (ei 180.0 100.0) (f 100.0 80.0) (g 80.0 60.0) (h 80.0 20.0) (i 135.0 40.0) (i@ 230.0 100.0) (ii 155.0 55) (jh 70.0 50.0) (k 80.0 60.0) (l 80.0 40.0) (m 70.0 60.0) (n 60.0 50.0) (ng 95.0 60.0) (o 240.0 130.0) (oi 280.0 150.0) (oo 240.0 130.0) (ou 220.0 80.0) (p 90.0 50.0) (r 80.0 30.0) (s 105.0 60.0) (sh 105.0 80.0) (t 75.0 50.0) (th 90.0 60.0) (u 210.0 70.0) (u@ 230.0 110.0) (uh 160.0 60.0) (uu 230.0 150.0) (v 60.0 40.0) (w 80.0 60.0) (y 80.0 40.0) (z 75.0 40.0) (zh 70.0 40.0) (# 100.0 100.0) )) (provide 'klatt_durs) festival/lib/gswdurtreeZ.scm0000644000175000017500000010506107262605650014611 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; A tree to predict zcore durations build from gsw 450 (timit) ;;; doesn't use actual phonemes so it can have better generalizations ;;; ;; pre Sue's changes to mrpa_phones (on traing data) ;; RMSE 0.79102 Correlation is 0.610184 Mean (abs) Error 0.605081 (0.509517) ;; Post with balance ;; train test spit --stop 19 --balance 16 ;; RMSE 0.841861 Correlation is 0.526064 Mean (abs) Error 0.646614 (0.539288) ;; on training data ;; RMSE 0.784032 Correlation is 0.619165 Mean (abs) Error 0.602819 (0.501332) ;; ;; Oct 29th 1997 ;; stepwise (but its over trained) ;; RMSE 0.8322 Correlation is 0.5286 Mean (abs) Error 0.6375 (0.5350) ;; ;; May 11th 1998 ;; new architecture, full new train on f2b on test data ;; in zscore domain ;; RMSE 0.8076 Correlation is 0.5307 Mean (abs) Error 0.6113 (0.5278) ;; in absolute domain ;; RMSE 0.0276 Correlation 0.7468 Mean (abs) error 0.0203 (0.0187) ;; ;; May 18th 1998 ;; various corrections f2bdur.bbz.H0.S50.tree no names zscore ;; in zscore domain ;; RMSE 0.8049 Correlation is 0.6003 Mean (abs) Error 0.6008 (0.5357) ;; in absolute domain ;; RMSE 0.0268 Correlation 0.7766 Mean (abs) error 0.0196 (0.0183) (set! gsw_duration_cart_tree ' ((name is #) ((emph_sil is +) ((0.0 -0.5)) ((p.R:SylStructure.parent.parent.pbreak is BB) ((0.0 2.0)) ((0.0 0.0)))) ((R:SylStructure.parent.accented is 0) ((n.ph_ctype is 0) ((p.ph_vlng is 0) ((R:SylStructure.parent.syl_codasize < 1.5) ((p.ph_ctype is n) ((ph_ctype is f) ((0.559208 -0.783163)) ((1.05215 -0.222704))) ((ph_ctype is s) ((R:SylStructure.parent.syl_break is 2) ((0.589948 0.764459)) ((R:SylStructure.parent.asyl_in < 0.7) ((1.06385 0.567944)) ((0.691943 0.0530272)))) ((ph_vlng is l) ((pp.ph_vfront is 1) ((1.06991 0.766486)) ((R:SylStructure.parent.syl_break is 1) ((0.69665 0.279248)) ((0.670353 0.0567774)))) ((p.ph_ctype is s) ((seg_onsetcoda is coda) ((0.828638 -0.038356)) ((ph_ctype is f) ((0.7631 -0.545853)) ((0.49329 -0.765994)))) ((R:SylStructure.parent.parent.gpos is det) ((R:SylStructure.parent.last_accent < 0.3) ((R:SylStructure.parent.sub_phrases < 1) ((0.811686 0.160195)) ((0.799015 0.713958))) ((0.731599 -0.215472))) ((ph_ctype is r) ((0.673487 0.092772)) ((R:SylStructure.parent.asyl_in < 1) ((0.745273 0.00132813)) ((0.75457 -0.334898))))))))) ((pos_in_syl < 0.5) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((0.902446 -0.041618)) ((R:SylStructure.parent.sub_phrases < 2.3) ((0.900629 0.262952)) ((1.18474 0.594794)))) ((seg_onset_stop is 0) ((R:SylStructure.parent.position_type is mid) ((0.512323 -0.760444)) ((R:SylStructure.parent.syl_out < 6.8) ((pp.ph_vlng is a) ((0.640575 -0.450449)) ((ph_ctype is f) ((R:SylStructure.parent.sub_phrases < 1.3) ((0.862876 -0.296956)) ((R:SylStructure.parent.syl_out < 2.4) ((0.803215 0.0422868)) ((0.877856 -0.154465)))) ((R:SylStructure.parent.syl_out < 3.6) ((R:SylStructure.parent.syl_out < 1.2) ((0.567081 -0.264199)) ((0.598043 -0.541738))) ((0.676843 -0.166623))))) ((0.691678 -0.57173)))) ((R:SylStructure.parent.parent.gpos is cc) ((1.15995 0.313289)) ((pp.ph_vfront is 1) ((0.555993 0.0695819)) ((R:SylStructure.parent.asyl_in < 1.2) ((R:SylStructure.parent.sub_phrases < 2.7) ((0.721635 -0.367088)) ((0.71919 -0.194887))) ((0.547052 -0.0637491))))))) ((ph_ctype is s) ((R:SylStructure.parent.syl_break is 0) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((0.650007 -0.333421)) ((0.846301 -0.165383))) ((0.527756 -0.516332))) ((R:SylStructure.parent.syl_break is 0) ((p.ph_ctype is s) ((0.504414 -0.779112)) ((0.812498 -0.337611))) ((pos_in_syl < 1.4) ((0.513041 -0.745807)) ((p.ph_ctype is s) ((0.350582 -1.04907)) ((0.362 -0.914974)))))))) ((R:SylStructure.parent.syl_break is 0) ((ph_ctype is n) ((R:SylStructure.parent.position_type is initial) ((pos_in_syl < 1.2) ((0.580485 0.172658)) ((0.630973 -0.101423))) ((0.577937 -0.360092))) ((R:SylStructure.parent.syl_out < 2.9) ((R:SylStructure.parent.syl_out < 1.1) ((R:SylStructure.parent.position_type is initial) ((0.896092 0.764189)) ((R:SylStructure.parent.sub_phrases < 3.6) ((ph_ctype is s) ((0.877362 0.555132)) ((0.604511 0.369882))) ((0.799982 0.666966)))) ((seg_onsetcoda is coda) ((p.ph_vlng is a) ((R:SylStructure.parent.last_accent < 0.4) ((0.800736 0.240634)) ((0.720606 0.486176))) ((1.18173 0.573811))) ((0.607147 0.194468)))) ((ph_ctype is r) ((0.88377 0.499383)) ((R:SylStructure.parent.last_accent < 0.5) ((R:SylStructure.parent.position_type is initial) ((R:SylStructure.parent.parent.word_numsyls < 2.4) ((0.62798 0.0737318)) ((0.787334 0.331014))) ((ph_ctype is s) ((0.808368 0.0929299)) ((0.527948 -0.0443271)))) ((seg_coda_fric is 0) ((p.ph_vlng is a) ((0.679745 0.517681)) ((R:SylStructure.parent.sub_phrases < 1.1) ((0.759979 0.128316)) ((0.775233 0.361383)))) ((R:SylStructure.parent.last_accent < 1.3) ((0.696255 0.054136)) ((0.632425 0.246742)))))))) ((pos_in_syl < 0.3) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((0.847602 0.621547)) ((ph_ctype is s) ((0.880645 0.501679)) ((R:SylStructure.parent.sub_phrases < 3.3) ((R:SylStructure.parent.sub_phrases < 0.3) ((0.901014 -0.042049)) ((0.657493 0.183226))) ((0.680126 0.284799))))) ((ph_ctype is s) ((p.ph_vlng is s) ((0.670033 -0.820934)) ((0.863306 -0.348735))) ((ph_ctype is n) ((R:SylStructure.parent.asyl_in < 1.2) ((0.656966 -0.40092)) ((0.530966 -0.639366))) ((seg_coda_fric is 0) ((1.04153 0.364857)) ((pos_in_syl < 1.2) ((R:SylStructure.parent.syl_out < 3.4) ((0.81503 -0.00768613)) ((0.602665 -0.197753))) ((0.601844 -0.394632))))))))) ((n.ph_ctype is f) ((pos_in_syl < 1.5) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((pos_in_syl < 0.1) ((1.63863 0.938841)) ((R:SylStructure.parent.position_type is initial) ((0.897722 -0.0796637)) ((nn.ph_vheight is 0) ((0.781081 0.480026)) ((0.779711 0.127175))))) ((ph_ctype is r) ((p.ph_ctype is s) ((0.581329 -0.708767)) ((0.564366 -0.236212))) ((ph_vlng is a) ((p.ph_ctype is r) ((0.70992 -0.273389)) ((R:SylStructure.parent.parent.gpos is in) ((0.764696 0.0581338)) ((nn.ph_vheight is 0) ((0.977737 0.721904)) ((R:SylStructure.parent.sub_phrases < 2.2) ((pp.ph_vfront is 0) ((0.586708 0.0161206)) ((0.619949 0.227372))) ((0.707285 0.445569)))))) ((ph_ctype is n) ((R:SylStructure.parent.syl_break is 1) ((nn.ph_vfront is 2) ((0.430295 -0.120097)) ((0.741371 0.219042))) ((0.587492 0.321245))) ((p.ph_ctype is n) ((0.871586 0.134075)) ((p.ph_ctype is r) ((0.490751 -0.466418)) ((R:SylStructure.parent.syl_codasize < 1.3) ((R:SylStructure.parent.sub_phrases < 2.2) ((p.ph_ctype is s) ((0.407452 -0.425925)) ((0.644771 -0.542809))) ((0.688772 -0.201899))) ((ph_vheight is 1) ((nn.ph_vheight is 0) ((0.692018 0.209018)) ((0.751345 -0.178136))) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.3) ((R:SylStructure.parent.asyl_in < 1.5) ((0.599633 -0.235593)) ((0.60042 0.126118))) ((p.ph_vlng is a) ((0.7148 -0.174812)) ((R:SylStructure.parent.parent.gpos is content) ((0.761296 -0.231509)) ((0.813081 -0.536405))))))))))))) ((ph_ctype is n) ((0.898844 0.163343)) ((p.ph_vlng is s) ((seg_coda_fric is 0) ((0.752921 -0.45528)) ((0.890079 -0.0998025))) ((ph_ctype is f) ((0.729376 -0.930547)) ((ph_ctype is s) ((R:SylStructure.parent.R:Syllable.p.syl_break is 0) ((0.745052 -0.634119)) ((0.521502 -0.760176))) ((R:SylStructure.parent.syl_break is 1) ((0.766575 -0.121355)) ((0.795616 -0.557509)))))))) ((p.ph_vlng is 0) ((p.ph_ctype is r) ((ph_vlng is 0) ((0.733659 -0.402734)) ((R:SylStructure.parent.sub_phrases < 1.5) ((ph_vlng is s) ((0.326176 -0.988478)) ((n.ph_ctype is s) ((0.276471 -0.802536)) ((0.438283 -0.900628)))) ((nn.ph_vheight is 0) ((ph_vheight is 2) ((0.521 -0.768992)) ((0.615436 -0.574918))) ((ph_vheight is 1) ((0.387376 -0.756359)) ((pos_in_syl < 0.3) ((0.417235 -0.808937)) ((0.384043 -0.93315))))))) ((ph_vlng is a) ((ph_ctype is 0) ((n.ph_ctype is s) ((p.ph_ctype is f) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((0.415908 -0.428493)) ((pos_in_syl < 0.1) ((0.790441 0.0211071)) ((0.452465 -0.254485)))) ((p.ph_ctype is s) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((0.582447 -0.389966)) ((0.757648 0.185781))) ((R:SylStructure.parent.sub_phrases < 1.4) ((0.628965 0.422551)) ((0.713613 0.145576))))) ((seg_onset_stop is 0) ((R:SylStructure.parent.R:Syllable.p.syl_break is 0) ((pp.ph_vfront is 1) ((0.412363 -0.62319)) ((R:SylStructure.parent.syl_out < 3.6) ((0.729259 -0.317324)) ((0.441633 -0.591051)))) ((R:SylStructure.parent.syl_break is 1) ((R:SylStructure.parent.sub_phrases < 2.7) ((0.457728 -0.405607)) ((0.532411 -0.313148))) ((R:SylStructure.parent.last_accent < 0.3) ((1.14175 0.159416)) ((0.616396 -0.254651))))) ((R:SylStructure.parent.position_type is initial) ((0.264181 -0.799896)) ((0.439801 -0.551309))))) ((R:SylStructure.parent.position_type is final) ((0.552027 -0.707084)) ((0.585661 -0.901874)))) ((ph_ctype is s) ((pos_in_syl < 1.2) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((pp.ph_vfront is 1) ((0.607449 0.196466)) ((0.599662 0.00382414))) ((0.64109 -0.12859))) ((pp.ph_vfront is 1) ((0.720484 -0.219339)) ((0.688707 -0.516734)))) ((ph_vlng is s) ((n.ph_ctype is s) ((R:SylStructure.parent.parent.gpos is content) ((R:SylStructure.parent.position_type is single) ((0.659206 0.159445)) ((R:SylStructure.parent.parent.word_numsyls < 3.5) ((R:SylStructure.parent.sub_phrases < 2) ((0.447186 -0.419103)) ((0.631822 -0.0928561))) ((0.451623 -0.576116)))) ((ph_vheight is 3) ((0.578626 -0.64583)) ((0.56636 -0.4665)))) ((R:SylStructure.parent.parent.gpos is in) ((0.771516 -0.217292)) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((0.688571 -0.304382)) ((R:SylStructure.parent.parent.gpos is content) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((n.ph_ctype is n) ((0.556085 -0.572203)) ((0.820173 -0.240338))) ((R:SylStructure.parent.parent.word_numsyls < 2.2) ((0.595398 -0.588171)) ((0.524737 -0.95797)))) ((R:SylStructure.parent.sub_phrases < 3.9) ((0.371492 -0.959427)) ((0.440479 -0.845747))))))) ((R:SylStructure.parent.R:Syllable.p.syl_break is 0) ((p.ph_ctype is f) ((0.524088 -0.482247)) ((nn.ph_vheight is 1) ((0.587666 -0.632362)) ((ph_vlng is l) ((R:SylStructure.parent.position_type is final) ((0.513286 -0.713117)) ((0.604613 -0.924308))) ((R:SylStructure.parent.syl_codasize < 2.2) ((0.577997 -0.891342)) ((0.659804 -1.15252)))))) ((pp.ph_vlng is s) ((ph_ctype is f) ((0.813383 -0.599624)) ((0.984027 -0.0771909))) ((p.ph_ctype is f) ((R:SylStructure.parent.parent.gpos is in) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((0.313572 -1.03242)) ((0.525854 -0.542799))) ((R:SylStructure.parent.syl_out < 2.8) ((0.613007 -0.423979)) ((0.570258 -0.766379)))) ((R:SylStructure.parent.syl_break is 1) ((R:SylStructure.parent.parent.gpos is to) ((0.364585 -0.792895)) ((ph_vlng is l) ((0.69143 -0.276816)) ((0.65673 -0.523721)))) ((R:SylStructure.parent.syl_out < 3.6) ((R:SylStructure.parent.position_type is initial) ((0.682096 -0.488102)) ((0.406364 -0.731758))) ((0.584694 -0.822229))))))))))) ((n.ph_ctype is r) ((R:SylStructure.parent.position_type is initial) ((p.ph_vlng is a) ((0.797058 1.02334)) ((ph_ctype is s) ((1.0548 0.536277)) ((0.817253 0.138201)))) ((R:SylStructure.parent.sub_phrases < 1.1) ((R:SylStructure.parent.syl_out < 3.3) ((0.884574 -0.23471)) ((0.772063 -0.525292))) ((nn.ph_vfront is 1) ((1.25254 0.417485)) ((0.955557 -0.0781996))))) ((pp.ph_vfront is 0) ((ph_ctype is f) ((n.ph_ctype is s) ((R:SylStructure.parent.parent.gpos is content) ((R:SylStructure.parent.R:Syllable.p.syl_break is 0) ((0.583506 -0.56941)) ((0.525949 -0.289362))) ((0.749316 -0.0921038))) ((p.ph_vlng is s) ((0.734234 0.139463)) ((0.680119 -0.0708717)))) ((ph_vlng is s) ((ph_vheight is 1) ((0.908712 -0.618971)) ((0.55344 -0.840495))) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 1.2) ((pos_in_syl < 1.2) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((0.838715 0.00913392)) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((ph_vheight is 2) ((0.555513 -0.512523)) ((R:SylStructure.parent.position_type is initial) ((0.758711 0.121704)) ((0.737555 -0.25637)))) ((R:SylStructure.parent.syl_out < 3.1) ((n.ph_ctype is s) ((0.611756 -0.474522)) ((1.05437 -0.247206))) ((R:SylStructure.parent.syl_codasize < 2.2) ((R:SylStructure.parent.position_type is final) ((0.567761 -0.597866)) ((0.785599 -0.407765))) ((0.575598 -0.741256)))))) ((ph_ctype is s) ((n.ph_ctype is s) ((0.661069 -1.08426)) ((0.783184 -0.39789))) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((R:SylStructure.parent.sub_phrases < 2.6) ((0.511323 -0.666011)) ((0.691878 -0.499492))) ((ph_ctype is r) ((0.482131 -0.253186)) ((0.852955 -0.372832)))))) ((0.854447 -0.0936489))))) ((R:SylStructure.parent.position_type is final) ((0.685939 -0.249982)) ((R:SylStructure.parent.syl_out < 3.2) ((0.989843 0.18086)) ((0.686805 -0.0402908))))))))) ((R:SylStructure.parent.syl_out < 2.4) ((R:SylStructure.parent.syl_out < 0.2) ((seg_onsetcoda is coda) ((ph_ctype is s) ((R:SylStructure.parent.syl_break is 4) ((pp.ph_vlng is 0) ((0.959737 1.63203)) ((1.20714 0.994933))) ((n.ph_ctype is 0) ((R:SylStructure.parent.syl_break is 2) ((0.864809 0.214457)) ((0.874278 0.730381))) ((pp.ph_vfront is 0) ((seg_coda_fric is 0) ((1.20844 -0.336221)) ((1.01357 0.468302))) ((0.658106 -0.799121))))) ((n.ph_ctype is f) ((ph_ctype is f) ((1.26332 0.0300613)) ((ph_vlng is d) ((1.02719 1.1649)) ((ph_ctype is 0) ((R:SylStructure.parent.asyl_in < 1.2) ((1.14048 2.2668)) ((ph_vheight is 1) ((1.15528 1.50375)) ((1.42406 2.07927)))) ((R:SylStructure.parent.sub_phrases < 1.1) ((0.955892 1.10243)) ((R:SylStructure.parent.syl_break is 2) ((1.32682 1.8432)) ((1.27582 1.59853))))))) ((n.ph_ctype is 0) ((ph_ctype is n) ((R:SylStructure.parent.syl_break is 2) ((1.45399 1.12927)) ((1.05543 0.442376))) ((R:SylStructure.parent.syl_break is 4) ((R:SylStructure.parent.position_type is final) ((ph_ctype is f) ((1.46434 1.76508)) ((0.978055 0.7486))) ((1.2395 2.30826))) ((ph_ctype is 0) ((0.935325 1.69917)) ((nn.ph_vfront is 1) ((1.20456 1.31128)) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((nn.ph_vheight is 0) ((1.16907 0.212421)) ((0.952091 0.653094))) ((p.ph_ctype is 0) ((1.05502 1.25802)) ((0.818731 0.777568)))))))) ((ph_ctype is f) ((p.ph_ctype is 0) ((1.03918 0.163941)) ((0.737545 -0.167063))) ((R:SylStructure.parent.position_type is final) ((n.ph_ctype is n) ((R:SylStructure.parent.last_accent < 0.5) ((R:SylStructure.parent.sub_phrases < 2.8) ((0.826207 -0.000859005)) ((0.871119 0.273433))) ((R:SylStructure.parent.parent.word_numsyls < 2.4) ((1.17405 1.05694)) ((0.858394 0.244916)))) ((R:SylStructure.parent.syl_codasize < 2.2) ((p.ph_ctype is 0) ((1.14092 1.21187)) ((R:SylStructure.parent.syl_break is 2) ((1.02653 0.59865)) ((0.94248 1.1634)))) ((seg_coda_fric is 0) ((1.07441 0.292935)) ((1.15736 0.92574))))) ((ph_vlng is s) ((R:SylStructure.parent.syl_break is 2) ((1.34638 1.23484)) ((0.951514 2.02008))) ((ph_ctype is 0) ((p.ph_ctype is r) ((0.806106 0.697089)) ((R:SylStructure.parent.syl_break is 2) ((1.10891 0.992197)) ((1.04657 1.51093)))) ((1.18165 0.520952))))))))) ((p.ph_vlng is 0) ((pos_in_syl < 0.7) ((R:SylStructure.parent.position_type is final) ((ph_ctype is r) ((0.966357 0.185827)) ((ph_ctype is s) ((0.647163 0.0332298)) ((0.692972 -0.534917)))) ((ph_ctype is s) ((0.881521 0.575107)) ((p.ph_ctype is f) ((0.8223 -0.111275)) ((R:SylStructure.parent.last_accent < 0.3) ((0.969188 0.09447)) ((0.894438 0.381947)))))) ((p.ph_ctype is f) ((0.479748 -0.490108)) ((0.813125 -0.201268)))) ((ph_ctype is s) ((0.908566 1.20397)) ((R:SylStructure.parent.last_accent < 1.2) ((0.88078 0.636568)) ((0.978087 1.07763)))))) ((pos_in_syl < 1.3) ((R:SylStructure.parent.syl_break is 0) ((pos_in_syl < 0.1) ((R:SylStructure.parent.position_type is initial) ((p.ph_ctype is n) ((0.801651 -0.0163359)) ((ph_ctype is s) ((n.ph_ctype is r) ((0.893307 1.07253)) ((p.ph_vlng is 0) ((0.92651 0.525806)) ((0.652444 0.952792)))) ((p.ph_vlng is 0) ((seg_onsetcoda is coda) ((0.820151 0.469117)) ((p.ph_ctype is f) ((0.747972 -0.0716448)) ((ph_ctype is f) ((0.770882 0.457137)) ((0.840905 0.102492))))) ((R:SylStructure.parent.syl_out < 1.1) ((0.667824 0.697337)) ((0.737967 0.375114)))))) ((ph_vheight is 1) ((0.624353 0.410671)) ((R:SylStructure.parent.asyl_in < 0.8) ((0.647905 -0.331055)) ((p.ph_ctype is s) ((0.629039 -0.240616)) ((0.749277 -0.0191273)))))) ((ph_vheight is 3) ((p.ph_ctype is s) ((0.626922 0.556537)) ((0.789357 0.153892))) ((seg_onsetcoda is coda) ((n.ph_ctype is 0) ((R:SylStructure.parent.parent.word_numsyls < 3.4) ((0.744714 0.123242)) ((0.742039 0.295753))) ((seg_coda_fric is 0) ((R:SylStructure.parent.parent.word_numsyls < 2.4) ((ph_vheight is 1) ((0.549715 -0.341018)) ((0.573641 -0.00893114))) ((nn.ph_vfront is 2) ((0.67099 -0.744625)) ((0.664438 -0.302803)))) ((p.ph_vlng is 0) ((0.630028 0.113815)) ((0.632794 -0.128733))))) ((ph_ctype is r) ((0.367169 -0.854509)) ((0.94334 -0.216179)))))) ((n.ph_ctype is f) ((ph_vlng is 0) ((1.3089 0.46195)) ((R:SylStructure.parent.syl_codasize < 1.3) ((1.07673 0.657169)) ((pp.ph_vlng is 0) ((0.972319 1.08222)) ((1.00038 1.46257))))) ((p.ph_vlng is l) ((1.03617 0.785204)) ((p.ph_vlng is a) ((R:SylStructure.parent.position_type is final) ((1.00681 0.321168)) ((0.928115 0.950834))) ((ph_vlng is 0) ((pos_in_syl < 0.1) ((R:SylStructure.parent.position_type is final) ((0.863682 -0.167374)) ((nn.ph_vheight is 0) ((p.ph_ctype is f) ((0.773591 -0.00374425)) ((R:SylStructure.parent.syl_out < 1.1) ((0.951802 0.228448)) ((1.02282 0.504252)))) ((1.09721 0.736476)))) ((R:SylStructure.parent.position_type is final) ((1.04302 0.0590974)) ((0.589208 -0.431535)))) ((n.ph_ctype is 0) ((1.27879 1.00642)) ((ph_vlng is s) ((R:SylStructure.parent.asyl_in < 1.4) ((0.935787 0.481652)) ((0.9887 0.749861))) ((R:SylStructure.parent.syl_out < 1.1) ((R:SylStructure.parent.position_type is final) ((0.921307 0.0696307)) ((0.83675 0.552212))) ((0.810076 -0.0479225)))))))))) ((ph_ctype is s) ((n.ph_ctype is s) ((0.706959 -1.0609)) ((p.ph_ctype is n) ((0.850614 -0.59933)) ((n.ph_ctype is r) ((0.665947 0.00698725)) ((n.ph_ctype is 0) ((R:SylStructure.parent.position_type is initial) ((0.762889 -0.0649044)) ((0.723956 -0.248899))) ((R:SylStructure.parent.sub_phrases < 1.4) ((0.632957 -0.601987)) ((0.889114 -0.302401))))))) ((ph_ctype is f) ((R:SylStructure.parent.syl_codasize < 2.2) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((R:SylStructure.parent.syl_out < 1.1) ((0.865267 0.164636)) ((0.581827 -0.0989051))) ((nn.ph_vfront is 2) ((0.684459 -0.316836)) ((0.778854 -0.0961191)))) ((R:SylStructure.parent.syl_out < 1.1) ((p.ph_ctype is s) ((0.837964 -0.429437)) ((0.875304 -0.0652743))) ((0.611071 -0.635089)))) ((p.ph_ctype is r) ((R:SylStructure.parent.syl_out < 1.1) ((0.762012 0.0139361)) ((0.567983 -0.454845))) ((R:SylStructure.parent.syl_codasize < 2.2) ((ph_ctype is l) ((1.18845 0.809091)) ((R:SylStructure.parent.position_type is initial) ((ph_ctype is n) ((0.773548 -0.277092)) ((1.01586 0.281001))) ((p.ph_ctype is 0) ((1.06831 0.699145)) ((0.924189 0.241873))))) ((R:SylStructure.parent.syl_break is 0) ((ph_ctype is n) ((0.592321 -0.470784)) ((0.778688 -0.072112))) ((n.ph_ctype is s) ((1.08848 0.0733489)) ((1.25674 0.608371)))))))))) ((pos_in_syl < 0.7) ((p.ph_vlng is 0) ((R:SylStructure.parent.position_type is mid) ((ph_ctype is 0) ((ph_vheight is 2) ((0.456225 -0.293282)) ((0.561529 -0.0816115))) ((0.6537 -0.504024))) ((ph_ctype is s) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((1.31586 0.98395)) ((R:SylStructure.parent.position_type is single) ((0.816869 0.634789)) ((R:SylStructure.parent.syl_out < 4.4) ((1.05578 0.479029)) ((R:SylStructure.parent.asyl_in < 0.4) ((1.11813 0.143214)) ((0.87178 0.406834)))))) ((n.ph_ctype is n) ((R:SylStructure.parent.last_accent < 0.6) ((0.838154 -0.415599)) ((0.924024 0.110288))) ((seg_onsetcoda is coda) ((nn.ph_vfront is 2) ((0.670096 0.0314187)) ((n.ph_ctype is f) ((1.00363 0.693893)) ((R:SylStructure.parent.syl_out < 6) ((0.772363 0.215675)) ((0.920313 0.574068))))) ((R:SylStructure.parent.position_type is final) ((0.673837 -0.458142)) ((R:SylStructure.parent.sub_phrases < 2.8) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((0.894817 0.304628)) ((ph_ctype is n) ((0.787302 -0.23094)) ((R:SylStructure.parent.asyl_in < 1.2) ((ph_ctype is f) ((R:SylStructure.parent.last_accent < 0.5) ((1.12278 0.326954)) ((0.802236 -0.100616))) ((0.791255 -0.0919132))) ((0.95233 0.219053))))) ((R:SylStructure.parent.position_type is initial) ((ph_ctype is f) ((1.0616 0.216118)) ((0.703216 -0.00834086))) ((ph_ctype is f) ((1.22277 0.761763)) ((0.904811 0.332721)))))))))) ((ph_vheight is 0) ((p.ph_vlng is s) ((0.873379 0.217178)) ((n.ph_ctype is r) ((0.723915 1.29451)) ((n.ph_ctype is 0) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((R:SylStructure.parent.sub_phrases < 4) ((seg_coda_fric is 0) ((p.ph_vlng is l) ((0.849154 0.945261)) ((0.633261 0.687498))) ((0.728546 0.403076))) ((0.850962 1.00255))) ((0.957999 1.09113))) ((0.85771 0.209045))))) ((ph_vheight is 2) ((0.803401 -0.0544067)) ((0.681353 0.256045))))) ((n.ph_ctype is f) ((ph_ctype is s) ((p.ph_vlng is 0) ((0.479307 -0.9673)) ((0.700477 -0.351397))) ((ph_ctype is f) ((0.73467 -0.6233)) ((R:SylStructure.parent.syl_break is 0) ((p.ph_ctype is s) ((0.56282 0.266234)) ((p.ph_ctype is r) ((0.446203 -0.302281)) ((R:SylStructure.parent.sub_phrases < 2.7) ((ph_ctype is 0) ((0.572016 -0.0102436)) ((0.497358 -0.274514))) ((0.545477 0.0482177))))) ((ph_vlng is s) ((0.805269 0.888495)) ((ph_ctype is n) ((0.869854 0.653018)) ((R:SylStructure.parent.sub_phrases < 2.2) ((0.735031 0.0612886)) ((0.771859 0.346637)))))))) ((R:SylStructure.parent.syl_codasize < 1.4) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.3) ((R:SylStructure.parent.position_type is initial) ((0.743458 0.0411808)) ((1.13068 0.613305))) ((pos_in_syl < 1.2) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((1.11481 0.175467)) ((0.937893 -0.276407))) ((0.74264 -0.550878)))) ((pos_in_syl < 3.4) ((seg_onsetcoda is coda) ((ph_ctype is r) ((n.ph_ctype is s) ((0.714319 -0.240328)) ((p.ph_ctype is 0) ((0.976987 0.330352)) ((1.1781 -0.0816682)))) ((ph_ctype is l) ((n.ph_ctype is 0) ((1.39137 0.383533)) ((0.725585 -0.324515))) ((ph_vheight is 3) ((ph_vlng is d) ((0.802626 -0.62487)) ((n.ph_ctype is r) ((0.661091 -0.513869)) ((R:SylStructure.parent.position_type is initial) ((R:SylStructure.parent.parent.word_numsyls < 2.4) ((0.482285 0.207874)) ((0.401601 -0.0204711))) ((0.733755 0.397372))))) ((n.ph_ctype is r) ((p.ph_ctype is 0) ((pos_in_syl < 1.2) ((0.666325 0.271734)) ((nn.ph_vheight is 0) ((0.642401 -0.261466)) ((0.783684 -0.00956571)))) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((0.692225 -0.381895)) ((0.741921 -0.0898767)))) ((nn.ph_vfront is 2) ((ph_ctype is s) ((0.697527 -1.12626)) ((n.ph_ctype is s) ((ph_vlng is 0) ((R:SylStructure.parent.sub_phrases < 2.4) ((0.498719 -0.906926)) ((0.635342 -0.625651))) ((0.45886 -0.385089))) ((0.848596 -0.359702)))) ((p.ph_vlng is a) ((p.ph_ctype is 0) ((0.947278 0.216904)) ((0.637933 -0.394349))) ((p.ph_ctype is r) ((R:SylStructure.parent.syl_break is 0) ((0.529903 -0.860573)) ((0.581378 -0.510488))) ((ph_vlng is 0) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((seg_onset_stop is 0) ((R:SylStructure.parent.syl_break is 0) ((p.ph_vlng is d) ((0.768363 0.0108428)) ((ph_ctype is s) ((0.835756 -0.035054)) ((ph_ctype is f) ((p.ph_vlng is s) ((0.602016 -0.179727)) ((0.640126 -0.297341))) ((0.674628 -0.542602))))) ((ph_ctype is s) ((0.662261 -0.60496)) ((0.662088 -0.432058)))) ((R:SylStructure.parent.syl_out < 4.4) ((0.582448 -0.389079)) ((ph_ctype is s) ((0.60413 -0.73564)) ((0.567153 -0.605444))))) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((0.761115 -0.827377)) ((ph_ctype is n) ((0.855183 -0.275338)) ((R:SylStructure.parent.syl_break is 0) ((0.788288 -0.802801)) ((R:SylStructure.parent.syl_codasize < 2.2) ((0.686134 -0.371234)) ((0.840184 -0.772883))))))) ((pos_in_syl < 1.2) ((R:SylStructure.parent.syl_break is 0) ((n.ph_ctype is n) ((0.423592 -0.655006)) ((R:SylStructure.parent.syl_out < 4.4) ((0.595269 -0.303751)) ((0.478433 -0.456882)))) ((0.688133 -0.133182))) ((seg_onset_stop is 0) ((1.27464 0.114442)) ((0.406837 -0.167545)))))))))))) ((ph_ctype is r) ((0.462874 -0.87695)) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((0.645442 -0.640572)) ((0.673717 -0.321322))))) ((0.61008 -0.925472)))))))) ;; RMSE 0.8085 Correlation is 0.5899 Mean (abs) Error 0.6024 (0.5393) )) (provide 'gswdurtreeZ) festival/lib/f2bdurtreeZ.scm0000644000175000017500000010255307262605650014465 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; First attempt at a tree to learn durations. Although ;;; it was trained from F2B and the radio phone set should ;;; work for others that are decalred with the same phone ;;; features ;;; ;; in ancient items (not on independent data) ;; RMSE 0.821086 Correlation is 0.573693 Mean (abs) Error 0.612327 (0.547034) ;; on independent test data ;; RMSE 0.8054 Correlation is 0.5327 Mean (abs) Error 0.6073 (0.5290) (set! f2b_duration_cart_tree ' ((name is #) ((emph_sil is +) ((0.0 -0.5)) ((R:Segment.p.R:SylStructure.parent.parent.pbreak is BB) ((0.0 2.0)) ((0.0 0.0)))) ((R:SylStructure.parent.accented is 0) ((R:Segment.p.ph_ctype is 0) ((R:Segment.n.ph_cplace is 0) ((ph_ctype is n) ((R:SylStructure.parent.position_type is initial) ((ph_cplace is a) ((0.675606 -0.068741)) ((0.674321 0.204279))) ((ph_cplace is l) ((0.688993 -0.124997)) ((R:SylStructure.parent.syl_out < 10) ((0.610881 -0.394451)) ((0.664504 -0.603196))))) ((ph_ctype is r) ((lisp_onset_glide is 0) ((R:SylStructure.parent.R:Syllable.n.syl_break is 0) ((0.949991 0.619256)) ((1.05066 0.979668))) ((0.858728 0.457972))) ((R:SylStructure.parent.position_type is single) ((syl_initial is 0) ((ph_ctype is s) ((0.692981 -0.788933)) ((0.834878 -0.116988))) ((R:SylStructure.parent.syl_out < 9.4) ((0.777932 0.357818)) ((0.852909 0.115478)))) ((R:Segment.n.ph_vrnd is +) ((ph_ctype is s) ((0.81305 0.87399)) ((0.65978 0.418928))) ((R:SylStructure.parent.position_type is final) ((R:SylStructure.parent.parent.word_numsyls < 2.3) ((0.71613 -0.2888)) ((0.642029 0.0624649))) ((R:Segment.nn.ph_cplace is a) ((R:SylStructure.parent.R:Syllable.n.syl_break is 1) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 1) ((R:SylStructure.parent.position_type is initial) ((0.854092 0.384456)) ((0.769274 0.10705))) ((lisp_coda_stop is 0) ((0.571763 0.0755348)) ((0.632928 -0.11117)))) ((lisp_coda_stop is 0) ((R:SylStructure.parent.syl_out < 8.6) ((0.555092 0.30006)) ((0.552673 -0.0263481))) ((0.903186 0.519185)))) ((R:Segment.nn.ph_cplace is p) ((0.563915 0.204967)) ((R:Segment.nn.ph_cvox is -) ((ph_ctype is s) ((0.67653 0.227681)) ((0.550623 0.435079))) ((R:SylStructure.parent.position_type is initial) ((0.93428 0.732003)) ((0.84114 0.423214))))))))))) ((R:Segment.n.ph_ctype is s) ((ph_ctype is s) ((0.693376 -1.02719)) ((R:Segment.n.ph_cplace is v) ((ph_ctype is r) ((0.539799 -0.344524)) ((0.858576 0.154275))) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 1.2) ((lisp_onset_glide is 0) ((R:SylStructure.parent.R:Syllable.n.syl_break is 1) ((ph_ctype is n) ((R:Segment.nn.ph_cplace is a) ((0.64604 -0.643797)) ((0.739746 -0.450649))) ((ph_ctype is f) ((0.657043 -0.462107)) ((0.798438 -0.19569)))) ((R:SylStructure.parent.syl_out < 8.4) ((lisp_coda_stop is 0) ((0.766789 -0.0484781)) ((0.717203 -0.322113))) ((R:SylStructure.parent.position_type is single) ((0.508168 -0.412874)) ((0.703458 -0.291121))))) ((0.574827 -0.65022))) ((0.801765 -0.120813))))) ((ph_ctype is n) ((R:Segment.n.ph_ctype is f) ((R:Segment.n.ph_cplace is b) ((0.797652 0.623764)) ((R:Segment.n.ph_cplace is a) ((R:Segment.n.seg_onsetcoda is coda) ((0.675567 0.288251)) ((0.854197 0.626272))) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 1) ((0.660394 -0.225466)) ((0.65275 0.0487195))))) ((R:Segment.n.ph_ctype is n) ((0.685613 -0.512227)) ((0.736366 -0.104066)))) ((R:Segment.n.ph_ctype is r) ((R:SylStructure.parent.position_type is initial) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.1) ((0.98185 0.152471)) ((0.851907 0.788208))) ((ph_ctype is f) ((0.76106 0.406474)) ((R:Segment.n.ph_cplace is a) ((1.01348 -0.0422549)) ((0.786777 -0.714839))))) ((ph_cplace is b) ((R:SylStructure.parent.syl_out < 10.4) ((0.799025 0.0992277)) ((0.851068 -0.115896))) ((R:Segment.n.ph_cplace is p) ((0.669855 -0.655488)) ((ph_ctype is r) ((R:Segment.n.ph_cplace is a) ((1.00772 0.130892)) ((0.635981 -0.35826))) ((R:Segment.n.ph_ctype is l) ((R:SylStructure.parent.R:Syllable.n.syl_break is 1) ((0.746089 -0.286007)) ((0.89158 0.154432))) ((R:Segment.n.ph_cplace is b) ((1.04971 -0.0449782)) ((R:SylStructure.parent.syl_out < 9.8) ((R:Segment.n.ph_ctype is f) ((R:Segment.n.seg_onsetcoda is coda) ((1.4144 0.143658)) ((0.781116 -0.281483))) ((ph_vlng is 0) ((0.755959 -0.33462)) ((0.81024 -0.615287)))) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.3) ((0.7426 -0.24342)) ((R:Segment.n.ph_ctype is f) ((R:Segment.n.ph_cplace is a) ((R:SylStructure.parent.position_type is single) ((0.578639 -0.322097)) ((0.55826 -0.663238))) ((0.616575 -0.713688))) ((0.759572 -0.314116)))))))))))))) ((R:Segment.n.ph_ctype is f) ((ph_ctype is 0) ((R:Segment.p.ph_ctype is r) ((R:SylStructure.parent.parent.word_numsyls < 2.2) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 1) ((0.733193 -0.180968)) ((0.563111 -0.467934))) ((0.426244 -0.758137))) ((ph_vlng is a) ((R:Segment.n.ph_cplace is b) ((R:Segment.nn.ph_cvox is +) ((0.680234 0.059855)) ((R:SylStructure.parent.position_type is single) ((0.980851 0.443893)) ((0.715307 0.112865)))) ((R:Segment.p.ph_cplace is a) ((0.851224 0.695863)) ((R:Segment.nn.ph_cvox is -) ((0.75892 0.195772)) ((0.630633 0.478738))))) ((R:Segment.n.seg_onsetcoda is coda) ((R:Segment.n.ph_cplace is b) ((R:Segment.nn.ph_cplace is 0) ((0.815979 -0.477579)) ((0.851491 -0.168622))) ((R:SylStructure.parent.position_type is single) ((R:Segment.nn.ph_cvox is +) ((1.14265 0.717697)) ((0.814726 0.291482))) ((R:SylStructure.parent.R:Syllable.n.syl_break is 0) ((0.512322 -0.0749096)) ((0.488216 0.112774))))) ((R:SylStructure.parent.position_type is final) ((0.693071 -0.200708)) ((R:Segment.p.ph_cvox is +) ((0.489147 -0.378728)) ((0.695396 -0.525028))))))) ((ph_vlng is s) ((0.464234 -0.162706)) ((R:Segment.p.ph_cvox is +) ((R:SylStructure.parent.parent.word_numsyls < 2.2) ((0.566845 -0.616918)) ((0.92747 -0.26777))) ((0.632833 -0.858295))))) ((R:Segment.n.ph_vrnd is 0) ((R:Segment.p.ph_ctype is r) ((ph_vlng is 0) ((0.845308 -0.23426)) ((R:SylStructure.parent.syl_out < 4.8) ((R:Segment.n.ph_ctype is n) ((0.484602 -0.850587)) ((0.535398 -0.586652))) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.3) ((ph_vlng is a) ((0.368898 -0.799533)) ((lisp_coda_stop is 0) ((0.387923 -1.11431)) ((0.407377 -0.859849)))) ((R:Segment.n.ph_cplace is a) ((ph_vlng is a) ((0.382367 -0.787669)) ((0.522121 -0.687376))) ((0.361185 -0.853639)))))) ((ph_vlng is a) ((ph_ctype is 0) ((R:Segment.n.ph_ctype is s) ((R:Segment.p.ph_cvox is +) ((R:Segment.p.ph_cplace is d) ((0.502849 -0.232866)) ((R:SylStructure.parent.position_type is initial) ((0.641714 -0.0545426)) ((R:SylStructure.parent.parent.word_numsyls < 2.6) ((0.613913 0.373746)) ((R:Segment.n.ph_cplace is v) ((0.581158 0.310101)) ((0.628758 -0.068165)))))) ((R:SylStructure.parent.position_type is mid) ((0.459281 -0.553794)) ((0.728208 -0.138806)))) ((R:Segment.p.ph_cplace is v) ((0.32179 -0.728364)) ((R:Segment.p.ph_cplace is l) ((0.562971 -0.550272)) ((R:SylStructure.parent.position_type is initial) ((0.937298 -0.0246324)) ((R:Segment.p.ph_cvox is +) ((R:Segment.n.ph_ctype is n) ((R:Segment.n.ph_cplace is a) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 0) ((0.434029 -0.404793)) ((1.05548 -0.103717))) ((0.408372 -0.556145))) ((0.712335 -0.118776))) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.3) ((0.379593 -0.658075)) ((0.549207 -0.494876)))))))) ((R:SylStructure.parent.position_type is final) ((0.597124 -0.649729)) ((0.628822 -1.03743)))) ((ph_ctype is s) ((R:Segment.n.ph_ctype is r) ((R:SylStructure.parent.syl_out < 8.4) ((0.760328 0.31651)) ((0.738363 -0.0177161))) ((R:Segment.n.ph_ctype is l) ((0.649328 -0.108791)) ((0.594945 -0.712753)))) ((ph_vlng is s) ((R:Segment.n.ph_ctype is s) ((R:Segment.n.ph_cplace is v) ((R:Segment.nn.ph_cplace is a) ((0.583211 0.0724331)) ((0.434605 -0.229857))) ((R:Segment.p.ph_cplace is a) ((R:SylStructure.parent.position_type is single) ((0.785502 -0.00061573)) ((0.544995 -0.432984))) ((R:Segment.nn.ph_cplace is 0) ((0.507071 -0.715041)) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 0) ((0.506404 -0.573733)) ((0.62466 -0.3356)))))) ((R:Segment.p.ph_cplace is l) ((0.571756 -0.819693)) ((lisp_coda_stop is 0) ((R:SylStructure.parent.position_type is initial) ((0.906891 -0.352911)) ((R:Segment.n.ph_ctype is r) ((0.620335 -0.445714)) ((R:SylStructure.parent.parent.word_numsyls < 2.5) ((R:Segment.p.ph_cvox is +) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 0) ((0.484057 -0.781483)) ((0.653917 -0.615429))) ((0.754814 -0.531845))) ((0.493988 -0.881596))))) ((0.792979 -0.32648))))) ((R:Segment.p.ph_cvox is +) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.3) ((lisp_coda_stop is 0) ((0.913526 -0.195111)) ((0.56564 -0.64867))) ((R:SylStructure.parent.position_type is single) ((R:Segment.n.ph_cplace is a) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 1) ((0.790882 -0.488954)) ((0.780221 -0.185138))) ((0.487794 -0.691338))) ((R:Segment.p.ph_ctype is n) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 1) ((0.595729 -0.771698)) ((0.57908 -1.06592))) ((R:Segment.pp.ph_vfront is 0) ((0.591417 -0.784735)) ((0.486298 -0.436971)))))) ((ph_vlng is 0) ((0.629869 -0.960652)) ((R:Segment.n.ph_ctype is r) ((R:Segment.nn.ph_cplace is 0) ((0.591783 -0.671576)) ((R:Segment.nn.ph_cvox is +) ((0.365135 -0.822844)) ((0.428573 -0.988434)))) ((lisp_coda_stop is 0) ((R:Segment.p.ph_cplace is a) ((R:Segment.n.ph_cplace is a) ((0.428189 -0.730057)) ((0.337443 -0.861764))) ((0.57354 -0.494602))) ((0.497606 -0.414451)))))))))) ((ph_vlng is l) ((R:Segment.pp.ph_vfront is 1) ((0.937199 0.833877)) ((R:SylStructure.parent.syl_out < 12.7) ((0.729202 0.344121)) ((0.71086 0.101855)))) ((syl_initial is 0) ((R:Segment.p.ph_ctype is r) ((R:Segment.nn.ph_cplace is a) ((0.844815 0.175273)) ((0.662523 -0.297527))) ((ph_vlng is 0) ((R:Segment.p.ph_ctype is s) ((R:SylStructure.parent.syl_out < 14.6) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 0) ((0.665332 -0.610529)) ((0.42276 -0.848942))) ((0.427946 -0.980726))) ((R:SylStructure.parent.position_type is single) ((R:SylStructure.parent.R:Syllable.n.syl_break is 1) ((0.523367 -0.825038)) ((0.635654 -0.535303))) ((R:SylStructure.parent.position_type is final) ((0.515996 -0.707614)) ((ph_cplace is a) ((lisp_coda_stop is 0) ((0.689738 0.0446601)) ((0.698347 -0.268593))) ((R:Segment.nn.ph_cplace is a) ((0.706504 -0.659172)) ((0.775589 -0.201769))))))) ((0.79472 -0.0539192)))) ((ph_ctype is s) ((R:SylStructure.parent.position_type is single) ((R:Segment.p.ph_ctype is f) ((0.641302 0.532411)) ((R:Segment.n.ph_vrnd is +) ((0.800655 0.325651)) ((0.894711 0.0487864)))) ((R:SylStructure.parent.position_type is initial) ((R:Segment.nn.ph_cplace is a) ((0.618082 -0.0190591)) ((0.733637 0.156329))) ((ph_cplace is a) ((R:SylStructure.parent.parent.word_numsyls < 2.3) ((0.372869 -0.0827845)) ((0.494988 0.0882778))) ((0.593526 -0.335404))))) ((R:Segment.p.ph_cvox is +) ((R:Segment.p.ph_ctype is n) ((R:SylStructure.parent.syl_out < 5.4) ((1.0207 -0.152517)) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((0.711277 -0.513467)) ((0.509207 -0.726794)))) ((ph_cplace is g) ((0.545188 -0.568352)) ((R:Segment.p.ph_cplace is a) ((ph_ctype is n) ((0.61149 -0.325094)) ((R:SylStructure.parent.position_type is single) ((R:Segment.p.ph_ctype is r) ((0.525282 0.395446)) ((R:SylStructure.parent.R:Syllable.n.syl_break is 1) ((0.85778 0.0760293)) ((0.704055 0.290369)))) ((R:Segment.pp.ph_vfront is 0) ((0.590093 0.136983)) ((0.734563 -0.0570759))))) ((R:Segment.pp.ph_vfront is 2) ((0.519485 -0.477174)) ((0.707546 -0.13584)))))) ((R:SylStructure.parent.position_type is single) ((R:Segment.p.ph_ctype is f) ((0.797877 0.00462775)) ((R:Segment.pp.ph_vfront is 1) ((0.852184 -0.259914)) ((0.65313 -0.492506)))) ((R:SylStructure.parent.position_type is initial) ((0.662516 -0.45585)) ((lisp_onset_glide is 0) ((0.652534 -0.652428)) ((0.482818 -0.885728)))))))))))) ((syl_initial is 0) ((ph_cplace is 0) ((R:SylStructure.parent.position_type is single) ((R:Segment.n.ph_ctype is f) ((R:Segment.p.ph_cplace is a) ((R:Segment.n.ph_cplace is a) ((R:Segment.pp.ph_vfront is 0) ((1.06157 1.30945)) ((1.12041 1.85843))) ((1.05622 0.921414))) ((R:Segment.nn.ph_cvox is -) ((1.03073 0.916168)) ((1.06857 0.452851)))) ((R:Segment.p.ph_ctype is r) ((R:Segment.n.ph_cplace is v) ((1.22144 0.672433)) ((R:Segment.p.ph_cplace is l) ((0.859749 -0.315152)) ((R:Segment.nn.ph_cvox is -) ((0.89862 0.131037)) ((0.760033 -0.121252))))) ((R:SylStructure.parent.syl_out < 8.8) ((R:SylStructure.parent.syl_out < 0.8) ((1.06821 1.63716)) ((R:Segment.n.ph_cplace is a) ((R:Segment.p.ph_cvox is +) ((1.04477 0.581686)) ((R:Segment.nn.ph_cvox is +) ((0.769059 0.301576)) ((0.953428 0.0764058)))) ((R:Segment.p.ph_cplace is a) ((1.01367 0.507761)) ((1.2827 0.945031))))) ((R:Segment.n.ph_cplace is l) ((0.618397 -0.0873608)) ((R:SylStructure.parent.R:Syllable.n.syl_break is 0) ((R:Segment.p.ph_cvox is +) ((0.817182 0.477262)) ((0.792181 -0.0592145))) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((R:SylStructure.parent.syl_out < 16) ((0.995411 0.497843)) ((0.784087 0.152266))) ((1.11816 0.716352)))))))) ((R:Segment.n.ph_ctype is f) ((R:SylStructure.parent.position_type is final) ((1.35724 1.06028)) ((R:Segment.p.ph_ctype is r) ((R:SylStructure.parent.syl_out < 8.6) ((0.511716 -0.0833005)) ((0.492142 -0.30212))) ((R:Segment.n.ph_cplace is b) ((0.53059 0.00266551)) ((R:SylStructure.parent.parent.word_numsyls < 2.3) ((ph_vlng is l) ((0.433396 0.821463)) ((0.66915 0.415614))) ((0.501369 0.154721)))))) ((R:SylStructure.parent.position_type is final) ((R:Segment.n.ph_ctype is s) ((1.03896 0.524706)) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((1.15147 0.428386)) ((R:Segment.p.ph_cplace is a) ((0.919929 0.0314637)) ((0.716168 -0.366629))))) ((R:SylStructure.parent.R:Syllable.n.syl_break is 4) ((0.816778 0.408786)) ((lisp_onset_glide is 0) ((R:Segment.p.ph_ctype is n) ((R:Segment.n.ph_ctype is s) ((0.532911 -0.153851)) ((0.633518 -0.762353))) ((R:Segment.p.ph_cvox is -) ((R:Segment.p.ph_cplace is g) ((0.618376 -0.593197)) ((R:SylStructure.parent.R:Syllable.n.syl_break is 1) ((R:Segment.pp.ph_vfront is 0) ((R:Segment.n.ph_ctype is n) ((0.554085 -0.058903)) ((R:Segment.p.ph_cplace is a) ((0.59842 -0.174458)) ((0.585539 -0.349335)))) ((0.500857 -0.416613))) ((R:SylStructure.parent.syl_out < 7) ((0.616683 -0.00213272)) ((0.631444 -0.141773))))) ((R:SylStructure.parent.R:Syllable.n.syl_break is 0) ((0.5198 -0.151901)) ((ph_vlng is s) ((0.677428 0.203522)) ((0.780789 0.375429)))))) ((R:Segment.nn.ph_cplace is a) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((0.594604 -0.27832)) ((0.736114 -0.422756))) ((R:Segment.p.ph_cplace is a) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 1) ((0.512186 -0.732785)) ((0.550759 -0.506471))) ((0.47297 -0.791841))))))))) ((R:Segment.p.ph_ctype is 0) ((R:SylStructure.parent.position_type is final) ((lisp_coda_stop is 0) ((ph_ctype is f) ((R:Segment.nn.ph_cplace is 0) ((1.00978 0.366105)) ((0.80682 -0.0827529))) ((R:Segment.n.ph_cplace is a) ((R:Segment.nn.ph_cvox is -) ((1.07097 1.77503)) ((1.14864 1.14754))) ((R:Segment.n.ph_vrnd is -) ((0.883474 0.286471)) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 1) ((1.22264 0.884142)) ((1.03401 0.658192)))))) ((ph_cplace is a) ((R:SylStructure.parent.syl_out < 6.4) ((R:SylStructure.parent.syl_out < 0.6) ((1.07956 0.602849)) ((1.12301 0.0555897))) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((0.898888 -0.17527)) ((0.940932 0.274301)))) ((1.10093 -0.68098)))) ((R:Segment.n.ph_ctype is s) ((ph_cplace is v) ((0.639932 -1.33353)) ((R:SylStructure.parent.position_type is single) ((R:SylStructure.parent.R:Syllable.n.syl_break is 0) ((lisp_coda_stop is 0) ((0.822882 -0.131692)) ((0.971957 -0.385365))) ((R:Segment.nn.ph_cvox is -) ((1.06611 0.183678)) ((lisp_coda_stop is 0) ((0.967183 0.0925019)) ((0.876026 -0.230108))))) ((ph_ctype is f) ((R:SylStructure.parent.syl_out < 13) ((0.589198 -0.655594)) ((0.476651 -0.926625))) ((R:SylStructure.parent.syl_out < 5) ((0.682936 -0.227662)) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((R:Segment.nn.ph_cplace is a) ((0.447309 -0.700998)) ((0.626113 -0.468853))) ((0.657893 -0.383607))))))) ((ph_ctype is r) ((R:Segment.nn.ph_cvox is -) ((1.15158 1.15233)) ((R:Segment.n.ph_vrnd is -) ((1.05554 0.533749)) ((0.955478 0.0841894)))) ((ph_ctype is l) ((R:Segment.n.ph_ctype is 0) ((R:Segment.nn.ph_cplace is a) ((0.766431 0.28943)) ((1.48633 1.09574))) ((R:SylStructure.parent.position_type is single) ((1.01777 0.474653)) ((0.545859 -0.402743)))) ((R:SylStructure.parent.syl_out < 4.8) ((R:Segment.n.ph_vc is +) ((ph_ctype is n) ((0.776645 -0.433859)) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 0) ((0.776179 0.23435)) ((R:SylStructure.parent.parent.word_numsyls < 2.2) ((0.744272 -0.0859672)) ((0.782605 0.115647)))) ((0.626541 -0.167615)))) ((R:Segment.n.seg_onsetcoda is coda) ((1.28499 0.864144)) ((ph_cplace is a) ((0.926103 0.0435837)) ((0.839172 -0.189514))))) ((R:Segment.n.ph_ctype is n) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.1) ((0.973489 -0.203415)) ((0.777589 -0.849733))) ((ph_ctype is n) ((R:SylStructure.parent.position_type is initial) ((R:Segment.n.ph_vc is +) ((0.743482 -0.53384)) ((0.619309 -0.0987861))) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((1.15555 0.0786295)) ((1.06689 0.681662)))) ((R:Segment.n.ph_ctype is r) ((R:SylStructure.parent.syl_out < 8.9) ((0.752079 -0.237421)) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((0.664182 -0.041521)) ((0.772712 0.103499)))) ((R:Segment.n.seg_onsetcoda is coda) ((R:SylStructure.parent.position_type is mid) ((R:SylStructure.parent.parent.word_numsyls < 3.3) ((0.715944 -0.275113)) ((0.675729 0.202848))) ((R:Segment.n.ph_vrnd is -) ((R:SylStructure.parent.syl_out < 8.3) ((ph_ctype is s) ((0.82747 -0.116723)) ((0.689586 -0.303909))) ((R:SylStructure.parent.syl_out < 17.7) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 0) ((0.659686 -0.621268)) ((ph_cplace is a) ((0.861741 -0.285324)) ((0.507102 -0.444082)))) ((0.850664 -0.269084)))) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 0) ((0.878643 -0.255833)) ((0.98882 0.115252))))) ((ph_cplace is a) ((R:SylStructure.parent.syl_out < 13) ((0.850625 -0.289333)) ((0.788154 -0.44844))) ((0.70482 -0.630276)))))))))))) ((R:Segment.p.ph_ctype is l) ((R:SylStructure.parent.position_type is single) ((0.873748 -0.21639)) ((lisp_coda_stop is 0) ((0.71002 0.428132)) ((0.703501 0.015833)))) ((ph_vlng is 0) ((R:Segment.p.ph_ctype is r) ((R:SylStructure.parent.position_type is initial) ((0.907151 -0.494409)) ((ph_ctype is s) ((0.782539 -0.398555)) ((R:Segment.p.ph_cplace is 0) ((0.767435 -0.298857)) ((0.767046 0.151217))))) ((ph_cplace is a) ((R:Segment.n.ph_ctype is r) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((0.689367 0.0195991)) ((0.64446 -0.256648))) ((R:Segment.n.ph_vc is +) ((ph_ctype is s) ((R:Segment.nn.ph_cvox is +) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 1) ((0.59482 -0.214443)) ((0.745691 0.0292177))) ((0.523103 -0.391245))) ((R:Segment.p.ph_cvox is +) ((R:Segment.p.ph_cplace is a) ((0.524304 -0.428306)) ((0.605117 -0.165604))) ((R:Segment.p.ph_ctype is f) ((0.491251 -0.455353)) ((lisp_coda_stop is 0) ((R:SylStructure.parent.R:Syllable.n.syl_break is 1) ((0.175021 -1.02136)) ((0.264113 -0.976809))) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.3) ((0.704803 -0.716976)) ((0.300317 -0.924727))))))) ((ph_ctype is f) ((R:SylStructure.parent.syl_out < 13) ((R:Segment.n.ph_ctype is s) ((0.731994 -0.711044)) ((0.768008 -0.415076))) ((0.691821 -0.803284))) ((R:Segment.nn.ph_cplace is 0) ((R:Segment.n.ph_cplace is a) ((0.569567 -0.993506)) ((0.689849 -0.761696))) ((0.386818 -1.14744)))))) ((R:Segment.p.seg_onsetcoda is coda) ((R:Segment.p.ph_cplace is a) ((0.746337 -0.866206)) ((0.532751 -1.22185))) ((ph_cplace is l) ((0.74942 -0.820648)) ((0.685988 -0.298146)))))) ((0.812766 0.17291)))))) ((R:SylStructure.parent.position_type is mid) ((ph_ctype is r) ((0.577775 -0.54714)) ((R:Segment.n.ph_ctype is f) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 0) ((0.370448 0.00076407)) ((0.460385 0.20631))) ((R:Segment.p.ph_cvox is -) ((ph_vlng is 0) ((0.615959 -0.57434)) ((0.50852 -0.197814))) ((R:Segment.n.ph_ctype is 0) ((1.34281 0.477163)) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 1) ((0.59975 -0.1342)) ((0.640294 -0.32653))))))) ((R:Segment.n.ph_ctype is f) ((R:SylStructure.parent.position_type is initial) ((0.758739 0.311943)) ((R:Segment.n.seg_onsetcoda is coda) ((R:Segment.p.ph_ctype is f) ((1.28746 1.99771)) ((R:Segment.pp.ph_vfront is 1) ((1.42474 1.76925)) ((R:SylStructure.parent.R:Syllable.n.syl_break is 1) ((0.979414 1.37583)) ((1.00321 1.06671))))) ((1.15222 0.852004)))) ((R:Segment.p.ph_ctype is 0) ((R:Segment.n.ph_ctype is s) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 1) ((0.664807 -0.0880262)) ((0.573589 0.217234))) ((ph_ctype is s) ((ph_cplace is l) ((0.800348 0.66579)) ((ph_cplace is a) ((0.859133 1.46854)) ((R:SylStructure.parent.position_type is single) ((0.692229 1.23671)) ((0.552426 0.923928))))) ((R:SylStructure.parent.syl_out < 9.2) ((R:SylStructure.parent.position_type is single) ((R:SylStructure.parent.syl_out < 3.6) ((1.01673 1.26824)) ((0.848274 0.92375))) ((R:SylStructure.parent.R:Syllable.n.syl_break is 1) ((R:Segment.nn.ph_cplace is a) ((0.788163 0.818855)) ((0.822028 1.01227))) ((0.8365 0.483313)))) ((lisp_coda_stop is 0) ((R:Segment.nn.ph_cvox is +) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((0.807795 0.670829)) ((0.773774 0.435486))) ((0.849529 0.103561))) ((0.858848 0.763836)))))) ((R:Segment.n.ph_vrnd is -) ((ph_vlng is 0) ((R:SylStructure.parent.position_type is final) ((ph_cplace is a) ((R:Segment.nn.ph_cvox is -) ((0.691915 -0.42124)) ((R:Segment.p.ph_cplace is a) ((0.773696 0.354001)) ((0.65495 -0.14321)))) ((0.610433 -0.479739))) ((R:Segment.p.ph_ctype is r) ((R:SylStructure.parent.R:Syllable.n.syl_break is 0) ((0.560921 0.384674)) ((0.895267 0.746476))) ((R:Segment.p.ph_ctype is l) ((0.704694 0.568012)) ((R:Segment.p.ph_cplace is b) ((1.34739 0.539049)) ((R:Segment.p.ph_ctype is s) ((R:SylStructure.parent.syl_out < 12.9) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((0.807285 0.151429)) ((0.988033 0.383763))) ((0.878655 0.102291))) ((ph_ctype is n) ((0.759582 -0.315096)) ((R:SylStructure.parent.syl_out < 8.8) ((R:Segment.pp.ph_vfront is 0) ((0.846546 0.000647117)) ((R:Segment.pp.ph_vfront is 1) ((0.586216 0.150701)) ((0.793898 0.379041)))) ((lisp_coda_stop is 0) ((ph_ctype is f) ((0.74736 -0.31103)) ((0.715751 -0.00576581))) ((0.914486 0.17528)))))))))) ((1.24204 0.908819))) ((ph_ctype is s) ((ph_cplace is a) ((0.864408 1.35528)) ((R:Segment.n.seg_onsetcoda is coda) ((0.85602 0.344576)) ((0.869622 0.659223)))) ((R:Segment.nn.ph_cvox is -) ((R:Segment.n.ph_ctype is s) ((R:Segment.nn.ph_cplace is 0) ((0.942964 1.27475)) ((0.978218 0.650268))) ((R:SylStructure.parent.syl_out < 3.9) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((1.32463 1.05026)) ((0.896966 0.417727))) ((R:Segment.p.ph_cplace is a) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 0) ((0.776698 0.195369)) ((0.969518 0.432394))) ((0.799096 -0.0203318))))) ((ph_cplace is a) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((0.680861 -0.315846)) ((R:SylStructure.parent.R:Syllable.nn.syl_break is 1) ((0.954393 0.0965487)) ((0.884928 0.372884)))) ((lisp_coda_stop is 0) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((R:SylStructure.parent.position_type is final) ((1.03696 0.565834)) ((0.906661 0.277961))) ((R:SylStructure.parent.position_type is final) ((0.778429 -0.0967381)) ((0.863993 0.314023)))) ((R:Segment.p.ph_cplace is a) ((R:SylStructure.parent.R:Syllable.p.stress is 0) ((0.898898 0.571009)) ((0.830278 0.787486))) ((1.1101 0.333888))))))))))))) ;; RMSE 0.7726 Correlation is 0.5943 Mean (abs) Error 0.5752 (0.5160) )) (provide 'f2bdurtreeZ) festival/lib/tobi.scm0000644000175000017500000005545607262605650013235 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; A CART tree for predicting ToBI accents (learned from f2b) ;;; punctuation and minimal pos ;;; ; NON !H* L+H L*+ ; NONE10265 583 66 40 0 0 10954 [10265/10954] 93.710 ; H* 650 1805 61 57 0 0 2573 [1805/2573] 70.152 ; !H* 317 241 125 42 0 0 725 [125/725] 17.241 ; L+H* 457 486 76 80 0 0 1099 [80/1099] 7.279 ; L* 45 113 14 4 0 0 176 [0/176] 0.000 ; L*+H 6 6 0 1 0 0 13 [0/13] 0.000 ; 11740 3234 342 224 0 0 ;total 15540 correct 12275.000 78.990% (set! f2b_int_accent_cart_tree ' ;; these first few lines are hand written to deal with emphasis (from ssml) ((R:SylStructure.parent.R:Token.parent.EMPH is 1) (((NONE 0.0) (H* 1) (!H* 0.0) (L+H* 0) (L* 0) (L*+H 0) H*)) ((n.R:SylStructure.parent.R:Token.parent.EMPH is 1) (((NONE 1.0) (H* 0) (!H* 0.0) (L+H* 0) (L* 0) (L*+H 0) NONE)) ((p.R:SylStructure.parent.R:Token.parent.EMPH is 1) (((NONE 1.0) (H* 0) (!H* 0.0) (L+H* 0) (L* 0) (L*+H 0) NONE)) ((ssyl_in is 10) (((NONE 0.99726) (H* 0) (!H* 0.00273973) (L+H* 0) (L* 0) (L*+H 0) NONE)) ((R:SylStructure.parent.gpos is to) (((NONE 0.995984) (H* 0.00401606) (!H* 0) (L+H* 0) (L* 0) (L*+H 0) NONE)) ((R:SylStructure.parent.gpos is cc) (((NONE 0.987768) (H* 0.00611621) (!H* 0) (L+H* 0.00611621) (L* 0) (L*+H 0) NONE)) ((ssyl_out is 10) (((NONE 0.927273) (H* 0.0545455) (!H* 0) (L+H* 0.0181818) (L* 0) (L*+H 0) NONE)) ((R:SylStructure.parent.gpos is in) (((NONE 0.938322) (H* 0.0353618) (!H* 0.00493421) (L+H* 0.0197368) (L* 0.00164474) (L*+H 0) NONE)) ((R:SylStructure.parent.gpos is wp) (((NONE 0.895238) (H* 0.0857143) (!H* 0) (L+H* 0.0190476) (L* 0) (L*+H 0) NONE)) ((R:SylStructure.parent.gpos is aux) (((NONE 0.912281) (H* 0.0380117) (!H* 0.00584795) (L+H* 0.0350877) (L* 0.00584795) (L*+H 0.00292398) NONE)) ((R:SylStructure.parent.gpos is det) (((NONE 0.898004) (H* 0.0643016) (!H* 0.00332594) (L+H* 0.0332594) (L* 0) (L*+H 0.00110865) NONE)) ((stress is 0) (((NONE 0.978415) (H* 0.0144999) (!H* 0.00164772) (L+H* 0.00510793) (L* 0.000329544) (L*+H 0) NONE)) ((R:SylStructure.parent.R:Word.p.gpos is 0) (((NONE 0.209877) (H* 0.716049) (!H* 0) (L+H* 0.0617284) (L* 0.0123457) (L*+H 0) H*)) ((R:SylStructure.parent.gpos is md) (((NONE 0.693548) (H* 0.177419) (!H* 0.0322581) (L+H* 0.0967742) (L* 0) (L*+H 0) NONE)) ((p.syl_break is 3) ((syl_break is 1) (((NONE 0.4375) (H* 0.416667) (!H* 0) (L+H* 0.135417) (L* 0.0104167) (L*+H 0) NONE)) (((NONE 0.171171) (H* 0.666667) (!H* 0) (L+H* 0.144144) (L* 0.018018) (L*+H 0) H*))) ((pp.syl_break is 4) ((R:SylStructure.parent.R:Word.pp.gpos is in) (((NONE 0.0980392) (H* 0.803922) (!H* 0) (L+H* 0.0784314) (L* 0.0196078) (L*+H 0) H*)) ((syl_out is 0) (((NONE 0.0185185) (H* 0.796296) (!H* 0.037037) (L+H* 0.0925926) (L* 0.0555556) (L*+H 0) H*)) ((R:SylStructure.parent.R:Word.n.gpos is in) (((NONE 0.132353) (H* 0.676471) (!H* 0) (L+H* 0.161765) (L* 0.0294118) (L*+H 0) H*)) ((syl_break is 0) (((NONE 0.125) (H* 0.633929) (!H* 0.0133929) (L+H* 0.183036) (L* 0.0401786) (L*+H 0.00446429) H*)) ((n.stress is 0) (((NONE 0.364865) (H* 0.567568) (!H* 0) (L+H* 0.0540541) (L* 0.0135135) (L*+H 0) H*)) ((p.syl_break is 0) (((NONE 0.612903) (H* 0.290323) (!H* 0) (L+H* 0.0967742) (L* 0) (L*+H 0) NONE)) (((NONE 0.32) (H* 0.44) (!H* 0.02) (L+H* 0.22) (L* 0) (L*+H 0) H*)))))))) ((ssyl_in is 0) (((NONE 0.167769) (H* 0.628926) (!H* 0.0214876) (L+H* 0.142975) (L* 0.0363636) (L*+H 0.00247934) H*)) ((ssyl_out is 4) (((NONE 0.490385) (H* 0.240385) (!H* 0.0961538) (L+H* 0.163462) (L* 0.00961538) (L*+H 0) NONE)) ((pp.syl_break is 3) ((R:SylStructure.parent.R:Word.p.gpos is content) (((NONE 0.346154) (H* 0.346154) (!H* 0.0769231) (L+H* 0.192308) (L* 0.0384615) (L*+H 0) NONE)) (((NONE 0.160714) (H* 0.571429) (!H* 0.0178571) (L+H* 0.178571) (L* 0.0714286) (L*+H 0) H*))) ((syl_in is 2) ((n.stress is 0) ((R:SylStructure.parent.R:Word.p.gpos is in) (((NONE 0.218182) (H* 0.618182) (!H* 0.0363636) (L+H* 0.0909091) (L* 0.0181818) (L*+H 0.0181818) H*)) ((syl_out is 2) (((NONE 0.0961538) (H* 0.634615) (!H* 0.0961538) (L+H* 0.134615) (L* 0.0384615) (L*+H 0) H*)) ((R:SylStructure.parent.R:Word.p.gpos is content) ((syl_out is 4) (((NONE 0.56) (H* 0.12) (!H* 0.08) (L+H* 0.24) (L* 0) (L*+H 0) NONE)) (((NONE 0.262821) (H* 0.378205) (!H* 0.121795) (L+H* 0.192308) (L* 0.0448718) (L*+H 0) H*))) (((NONE 0.161905) (H* 0.590476) (!H* 0.0285714) (L+H* 0.171429) (L* 0.047619) (L*+H 0) H*))))) ((n.syl_break is 0) (((NONE 0.551724) (H* 0.293103) (!H* 0) (L+H* 0.155172) (L* 0) (L*+H 0) NONE)) (((NONE 0.408451) (H* 0.422535) (!H* 0.056338) (L+H* 0.112676) (L* 0) (L*+H 0) H*)))) ((R:SylStructure.parent.R:Word.n.gpos is 0) ((syl_break is 0) (((NONE 0.105263) (H* 0.315789) (!H* 0.157895) (L+H* 0.421053) (L* 0) (L*+H 0) L+H*)) (((NONE 0.641509) (H* 0.132075) (!H* 0.132075) (L+H* 0.0943396) (L* 0) (L*+H 0) NONE))) ((syl_break is 1) ((ssyl_in is 3) (((NONE 0.638889) (H* 0.152778) (!H* 0.125) (L+H* 0.0833333) (L* 0) (L*+H 0) NONE)) ((p.syl_break is 0) (((NONE 0.551402) (H* 0.186916) (!H* 0.158879) (L+H* 0.0841122) (L* 0.0186916) (L*+H 0) NONE)) ((n.stress is 0) ((pp.syl_break is 0) (((NONE 0.413043) (H* 0.184783) (!H* 0.152174) (L+H* 0.23913) (L* 0.0108696) (L*+H 0) NONE)) (((NONE 0.2125) (H* 0.3375) (!H* 0.1875) (L+H* 0.2125) (L* 0.05) (L*+H 0) H*))) (((NONE 0.449153) (H* 0.245763) (!H* 0.101695) (L+H* 0.20339) (L* 0) (L*+H 0) NONE))))) ((syl_out is 4) ((nn.syl_break is 0) ((pp.syl_break is 0) (((NONE 0.45614) (H* 0.210526) (!H* 0.192982) (L+H* 0.140351) (L* 0) (L*+H 0) NONE)) (((NONE 0.288462) (H* 0.25) (!H* 0.0961538) (L+H* 0.346154) (L* 0) (L*+H 0.0192308) L+H*))) (((NONE 0.163934) (H* 0.459016) (!H* 0.131148) (L+H* 0.245902) (L* 0) (L*+H 0) H*))) ((syl_out is 5) ((R:SylStructure.parent.R:Word.p.gpos is content) (((NONE 0.372881) (H* 0.20339) (!H* 0.169492) (L+H* 0.220339) (L* 0.0338983) (L*+H 0) NONE)) (((NONE 0.0961538) (H* 0.673077) (!H* 0.115385) (L+H* 0.0961538) (L* 0.0192308) (L*+H 0) H*))) ((R:SylStructure.parent.R:Word.pp.gpos is in) ((syl_in is 4) (((NONE 0.352113) (H* 0.422535) (!H* 0.15493) (L+H* 0.0704225) (L* 0) (L*+H 0) H*)) ((syl_in is 3) (((NONE 0.290323) (H* 0.467742) (!H* 0.0806452) (L+H* 0.145161) (L* 0.016129) (L*+H 0) H*)) ((pp.syl_break is 0) (((NONE 0.465517) (H* 0.293103) (!H* 0.172414) (L+H* 0.0689655) (L* 0) (L*+H 0) NONE)) ((R:SylStructure.parent.R:Word.p.gpos is content) (((NONE 0.18) (H* 0.36) (!H* 0.28) (L+H* 0.14) (L* 0.04) (L*+H 0) H*)) (((NONE 0.0877193) (H* 0.22807) (!H* 0.368421) (L+H* 0.298246) (L* 0.0175439) (L*+H 0) !H*)))))) ((ssyl_out is 2) ((p.syl_break is 0) (((NONE 0.634921) (H* 0.174603) (!H* 0.0793651) (L+H* 0.111111) (L* 0) (L*+H 0) NONE)) ((pp.syl_break is 0) (((NONE 0.388889) (H* 0.148148) (!H* 0.148148) (L+H* 0.259259) (L* 0.0185185) (L*+H 0.037037) NONE)) (((NONE 0.294118) (H* 0.137255) (!H* 0.215686) (L+H* 0.333333) (L* 0.0196078) (L*+H 0) L+H*)))) ((R:SylStructure.parent.R:Word.pp.gpos is to) (((NONE 0.0877193) (H* 0.350877) (!H* 0.210526) (L+H* 0.315789) (L* 0.0350877) (L*+H 0) H*)) ((syl_break is 3) ((pp.syl_break is 0) (((NONE 0.478261) (H* 0.141304) (!H* 0.195652) (L+H* 0.184783) (L* 0) (L*+H 0) NONE)) (((NONE 0.217822) (H* 0.366337) (!H* 0.257426) (L+H* 0.128713) (L* 0.029703) (L*+H 0) H*))) ((syl_in is 7) ((n.stress is 0) ((R:SylStructure.parent.R:Word.n.gpos is content) (((NONE 0.117647) (H* 0.220588) (!H* 0.441176) (L+H* 0.176471) (L* 0.0441176) (L*+H 0) !H*)) (((NONE 0.415385) (H* 0.0461538) (!H* 0.2) (L+H* 0.246154) (L* 0.0923077) (L*+H 0) NONE))) (((NONE 0.716981) (H* 0.113208) (!H* 0.0943396) (L+H* 0.0754717) (L* 0) (L*+H 0) NONE))) ((R:SylStructure.parent.R:Word.n.gpos is cc) (((NONE 0.292308) (H* 0.184615) (!H* 0.276923) (L+H* 0.246154) (L* 0) (L*+H 0) NONE)) ((nn.syl_break is 3) (((NONE 0.2) (H* 0.333333) (!H* 0.283333) (L+H* 0.15) (L* 0.0333333) (L*+H 0) H*)) ((ssyl_in is 4) (((NONE 0.383838) (H* 0.151515) (!H* 0.212121) (L+H* 0.20202) (L* 0.050505) (L*+H 0) NONE)) ((p.syl_break is 0) ((n.syl_break is 1) (((NONE 0.526316) (H* 0.210526) (!H* 0.0921053) (L+H* 0.171053) (L* 0) (L*+H 0) NONE)) ((ssyl_in is 3) (((NONE 0.509804) (H* 0.0980392) (!H* 0.215686) (L+H* 0.156863) (L* 0.0196078) (L*+H 0) NONE)) ((pp.syl_break is 0) (((NONE 0.506667) (H* 0.173333) (!H* 0.106667) (L+H* 0.2) (L* 0.0133333) (L*+H 0) NONE)) ((ssyl_in is 1) (((NONE 0.1) (H* 0.4) (!H* 0.266667) (L+H* 0.188889) (L* 0.0444444) (L*+H 0) H*)) (((NONE 0.326316) (H* 0.210526) (!H* 0.221053) (L+H* 0.189474) (L* 0.0526316) (L*+H 0) NONE)))))) ((R:SylStructure.parent.R:Word.p.gpos is in) (((NONE 0.0625) (H* 0.296875) (!H* 0.265625) (L+H* 0.328125) (L* 0.046875) (L*+H 0) L+H*)) ((syl_in is 6) (((NONE 0.271739) (H* 0.152174) (!H* 0.358696) (L+H* 0.184783) (L* 0.0326087) (L*+H 0) !H*)) ((syl_out is 2) (((NONE 0.111111) (H* 0.361111) (!H* 0.319444) (L+H* 0.138889) (L* 0.0555556) (L*+H 0.0138889) H*)) ((syl_in is 4) (((NONE 0.224) (H* 0.152) (!H* 0.328) (L+H* 0.24) (L* 0.056) (L*+H 0) !H*)) ((n.stress is 0) ((syl_in is 3) (((NONE 0.0833333) (H* 0.333333) (!H* 0.233333) (L+H* 0.216667) (L* 0.133333) (L*+H 0) H*)) (((NONE 0.283465) (H* 0.188976) (!H* 0.23622) (L+H* 0.204724) (L* 0.0708661) (L*+H 0.015748) NONE))) (((NONE 0.305263) (H* 0.284211) (!H* 0.210526) (L+H* 0.178947) (L* 0.0210526) (L*+H 0) NONE)))))))))))))))))))))))))))))))))))))))) ) ; NON L-L L-H H-L ; NONE13017 0 0 0 0 0 13017 [13017/13017] 100.000 ; H- 339 81 0 1 1 0 422 [81/422] 19.194 ; L- 223 52 0 5 0 0 280 [0/280] 0.000 ; L-L% 17 0 0 1057 96 0 1170 [1057/1170] 90.342 ; L-H% 16 0 0 457 139 0 612 [139/612] 22.712 ; H-L% 5 0 0 30 4 0 39 [0/39] 0.000 ; 13617 133 0 1550 240 0 ;total 15540 correct 14294.000 91.982% (set! f2b_int_tone_cart_tree '((lisp_syl_yn_question is 1) (((H-H% 1.0) H-H%)) ((R:SylStructure.parent.gpos is cc) (((NONE 0.996942) (H- 0.0030581) (L- 0) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((ssyl_in is 10) (((NONE 0.989041) (H- 0.00273973) (L- 0) (L-L% 0.00273973) (L-H% 0.00547945) (H-L% 0) NONE)) ((R:SylStructure.parent.gpos is md) (((NONE 0.986014) (H- 0) (L- 0) (L-L% 0.00699301) (L-H% 0.00699301) (H-L% 0) NONE)) ((p.old_syl_break is 4) (((NONE 0.99462) (H- 0.00239091) (L- 0.00119546) (L-L% 0) (L-H% 0.00119546) (H-L% 0.000597729) NONE)) ((R:SylStructure.parent.gpos is det) (((NONE 0.984635) (H- 0.00512164) (L- 0.00384123) (L-L% 0.00384123) (L-H% 0.00256082) (H-L% 0) NONE)) ((n.old_syl_break is 3) (((NONE 0.981848) (H- 0.00495049) (L- 0.00330033) (L-L% 0.00660066) (L-H% 0.00330033) (H-L% 0) NONE)) ((n.old_syl_break is 4) (((NONE 0.986982) (H- 0.000591716) (L- 0.0100592) (L-L% 0.00118343) (L-H% 0.00118343) (H-L% 0) NONE)) ((R:SylStructure.parent.gpos is in) (((NONE 0.977865) (H- 0.00390625) (L- 0.00390625) (L-L% 0.0078125) (L-H% 0.00651042) (H-L% 0) NONE)) ((old_syl_break is 4) ((R:SylStructure.parent.R:Word.n.gpos is 0) (((NONE 0) (H- 0.00892857) (L- 0) (L-L% 0.982143) (L-H% 0.00892857) (H-L% 0) L-L%)) ((R:SylStructure.parent.R:Word.p.gpos is aux) (((NONE 0) (H- 0) (L- 0) (L-L% 0.761905) (L-H% 0.238095) (H-L% 0) L-L%)) ((R:SylStructure.parent.R:Word.n.gpos is det) (((NONE 0) (H- 0) (L- 0) (L-L% 0.652542) (L-H% 0.347458) (H-L% 0) L-L%)) ((ssyl_in is 4) (((NONE 0) (H- 0) (L- 0) (L-L% 0.682243) (L-H% 0.313084) (H-L% 0.0046729) L-L%)) ((syl_in is 6) (((NONE 0) (H- 0) (L- 0.00649351) (L-L% 0.688312) (L-H% 0.298701) (H-L% 0.00649351) L-L%)) ((R:SylStructure.parent.R:Word.n.gpos is aux) (((NONE 0) (H- 0) (L- 0) (L-L% 0.464286) (L-H% 0.535714) (H-L% 0) L-H%)) ((syl_in is 5) (((NONE 0) (H- 0) (L- 0) (L-L% 0.666667) (L-H% 0.322034) (H-L% 0.0112994) L-L%)) ((sub_phrases is 2) (((NONE 0) (H- 0) (L- 0) (L-L% 0.696429) (L-H% 0.267857) (H-L% 0.0357143) L-L%)) ((R:SylStructure.parent.R:Word.p.gpos is det) (((NONE 0) (H- 0) (L- 0) (L-L% 0.628866) (L-H% 0.350515) (H-L% 0.0206186) L-L%)) ((sub_phrases is 0) ((R:SylStructure.parent.R:Word.n.gpos is in) ((n.old_syl_break is 0) (((NONE 0) (H- 0) (L- 0) (L-L% 0.68254) (L-H% 0.31746) (H-L% 0) L-L%)) (((NONE 0) (H- 0.0147059) (L- 0) (L-L% 0.338235) (L-H% 0.632353) (H-L% 0.0147059) L-H%))) ((n.stress is 0) (((NONE 0) (H- 0) (L- 0.0108303) (L-L% 0.599278) (L-H% 0.32491) (H-L% 0.064982) L-L%)) (((NONE 0) (H- 0) (L- 0) (L-L% 0.386364) (L-H% 0.579545) (H-L% 0.0340909) L-H%)))) (((NONE 0) (H- 0) (L- 0.00456621) (L-L% 0.652968) (L-H% 0.324201) (H-L% 0.0182648) L-L%)))))))))))) ((R:SylStructure.parent.gpos is pps) (((NONE 0.988764) (H- 0.011236) (L- 0) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((syl_in is 0) (((NONE 0.984848) (H- 0.0126263) (L- 0.00252525) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((R:SylStructure.parent.gpos is content) ((R:SylStructure.parent.R:Word.nn.gpos is 0) (((NONE 0.967914) (H- 0.0106952) (L- 0.0213904) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((pp.old_syl_break is 4) (((NONE 0.972315) (H- 0.0232558) (L- 0.00442968) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((syl_in is 1) (((NONE 0.951163) (H- 0.0372093) (L- 0.0116279) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((nn.old_syl_break is 4) (((NONE 0.956244) (H- 0.0127621) (L- 0.0291705) (L-L% 0) (L-H% 0) (H-L% 0.00182315) NONE)) ((R:SylStructure.parent.R:Word.nn.gpos is in) (((NONE 0.941919) (H- 0.0378788) (L- 0.020202) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((R:SylStructure.parent.R:Word.p.gpos is cc) (((NONE 0.919643) (H- 0.0714286) (L- 0.00892857) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((nn.old_syl_break is 3) (((NONE 0.927273) (H- 0.0472727) (L- 0.0254545) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((R:SylStructure.parent.R:Word.nn.gpos is cc) (((NONE 0.921569) (H- 0.0588235) (L- 0.0196078) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((ssyl_in is 0) (((NONE 0.911591) (H- 0.0825147) (L- 0.00589391) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((R:SylStructure.parent.R:Word.nn.gpos is to) (((NONE 0.912281) (H- 0.0350877) (L- 0.0526316) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((R:SylStructure.parent.R:Word.pp.gpos is to) (((NONE 0.894737) (H- 0.0526316) (L- 0.0526316) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((R:SylStructure.parent.R:Word.p.gpos is in) (((NONE 0.888554) (H- 0.0662651) (L- 0.0451807) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((R:SylStructure.parent.R:Word.pp.gpos is in) (((NONE 0.875817) (H- 0.0718954) (L- 0.0522876) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((syl_in is 2) (((NONE 0.869942) (H- 0.0867052) (L- 0.0433526) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((R:SylStructure.parent.R:Word.nn.gpos is aux) (((NONE 0.854839) (H- 0.0967742) (L- 0.0483871) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((sub_phrases is 1) (((NONE 0.836538) (H- 0.0721154) (L- 0.0913462) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((R:SylStructure.parent.R:Word.pp.gpos is det) (((NONE 0.832402) (H- 0.0949721) (L- 0.0726257) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((ssyl_in is 4) (((NONE 0.793103) (H- 0.103448) (L- 0.103448) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((n.old_syl_break is 0) (((NONE 0.850816) (H- 0.0839161) (L- 0.0652681) (L-L% 0) (L-H% 0) (H-L% 0) NONE)) ((R:SylStructure.parent.R:Word.n.gpos is content) (((NONE 0.889447) (H- 0.0753769) (L- 0.0251256) (L-L% 0) (L-H% 0) (H-L% 0.0100503) NONE)) ((old_syl_break is 3) (((NONE 0) (H- 0.609023) (L- 0.390977) (L-L% 0) (L-H% 0) (H-L% 0) H-)) (((NONE 1) (H- 0) (L- 0) (L-L% 0) (L-H% 0) (H-L% 0) NONE))))))))))))))))))))))) (((NONE 0.978947) (H- 0.0131579) (L- 0.00789474) (L-L% 0) (L-H% 0) (H-L% 0) NONE))))))))))))))) ) (defvar tobi_support_yn_questions t "tobi_support_yn_questions If set a crude final rise will be added at utterance that are judged to be yesy/no questions. Namely ending in a ? and not starting with a wh-for word.") (define (first_word syl) (let ((w (item.relation.parent syl 'SylStructure))) (item.relation.first w 'Word))) (define (syl_yn_question syl) "(syl_yn_question utt syl) Return 1 if this is the last syllable in a yes-no question. Basically if it ends in question mark and doesn't start with a wh-woerd. This isn't right but it depends on how much you want rising intonation." (if (and tobi_support_yn_questions (member_string (item.feat syl "syl_break") '("4" "3")) (not (member_string (downcase (item.name (first_word syl))) '("how" "why" "which" "who" "what" "where" "when"))) (string-matches (item.feat syl "R:SylStructure.parent.R:Token.parent.punc") ".*\\?.*")) "1" "0")) (provide 'tobi) festival/lib/f2bf0lr.scm0000644000175000017500000002702307262605650013522 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; First attempt at a linear regression model to predict F0 values. ;;; This is an attempt to reimplement the work in Black and ;;; Hunt ICSLP96, though this model probably isn't as good. ;;; ;;;start ;;; R2 = 0.251, F(74, 12711) = 57.5, Prob>F = 0.000 ;;; RMSE = 27.877 ;;;mid ;;; R2 = 0.332, F(74, 12711) = 85.6, Prob>F = 0.000 ;;; RMSE = 28.293 ;;;end ;;; R2 = 0.292, F(74, 12711) = 70.8, Prob>F = 0.000 ;;; RMSE = 27.139 (define (emph_syl syl) (if (string-equal (item.feat syl "tobi_accent") "NONE") 0.0 (if (string-equal (item.feat syl "R:SylStructure.parent.R:Token.parent.EMPH") "1") 2.0 0.0))) (set! f2b_f0_lr_start '( ( Intercept 160.584956 ) ( R:SylStructure.parent.R:Token.parent.EMPH 10.0 ) ( pp.tobi_accent 10.081770 (H*) ) ( pp.tobi_accent 3.358613 (!H*) ) ( pp.tobi_accent 4.144342 (*? X*? H*!H* * L+H* L+!H*) ) ( pp.tobi_accent -1.111794 (L*) ) ( pp.tobi_accent 19.646313 (L*+H L*+!H) ) ( p.tobi_accent 32.081029 (H*) ) ( p.tobi_accent 18.090033 (!H*) ) ( p.tobi_accent 23.255280 (*? X*? H*!H* * L+H* L+!H*) ) ( p.tobi_accent -9.623577 (L*) ) ( p.tobi_accent 26.517095 (L*+H L*+!H) ) ( tobi_accent 5.221081 (H*) ) ( tobi_accent 10.159194 (!H*) ) ( tobi_accent 3.645511 (*? X*? H*!H* * L+H* L+!H*) ) ( tobi_accent -5.720030 (L*) ) ( tobi_accent -6.355773 (L*+H L*+!H) ) ( n.tobi_accent -5.691933 (H*) ) ( n.tobi_accent 8.265606 (!H*) ) ( n.tobi_accent 0.861427 (*? X*? H*!H* * L+H* L+!H*) ) ( n.tobi_accent 1.270504 (L*) ) ( n.tobi_accent 3.499418 (L*+H L*+!H) ) ( nn.tobi_accent -3.785701 (H*) ) ( nn.tobi_accent 7.013446 (!H*) ) ( nn.tobi_accent 2.637494 (*? X*? H*!H* * L+H* L+!H*) ) ( nn.tobi_accent -0.392176 (L*) ) ( nn.tobi_accent -2.957502 (L*+H L*+!H) ) ( pp.tobi_endtone -3.531153 (L-L%) ) ( pp.tobi_endtone 0.131156 (L-) ) ( pp.tobi_endtone 2.729199 (H-L% !H-L% -X?) ) ( pp.tobi_endtone 8.258756 (L-H%) ) ( pp.tobi_endtone 5.836487 (H-) ) ( pp.tobi_endtone 11.213440 (!H- H-H%) ) ( R:Syllable.p.tobi_endtone -28.081359 (L-L%) ) ( R:Syllable.p.tobi_endtone -20.553145 (L-) ) ( R:Syllable.p.tobi_endtone -5.442577 (H-L% !H-L% -X?) ) ( R:Syllable.p.tobi_endtone -6.585836 (L-H%) ) ( R:Syllable.p.tobi_endtone 8.537044 (H-) ) ( R:Syllable.p.tobi_endtone 4.243342 (!H- H-H%) ) ( tobi_endtone -9.333926 (L-L%) ) ( tobi_endtone -0.346711 (L-) ) ( tobi_endtone -0.507352 (H-L% !H-L% -X?) ) ( tobi_endtone -0.937483 (L-H%) ) ( tobi_endtone 9.472265 (H-) ) ( tobi_endtone 14.256898 (!H- H-H%) ) ( n.tobi_endtone -13.084253 (L-L%) ) ( n.tobi_endtone -1.060688 (L-) ) ( n.tobi_endtone -7.947205 (H-L% !H-L% -X?) ) ( n.tobi_endtone -5.471592 (L-H%) ) ( n.tobi_endtone -0.095669 (H-) ) ( n.tobi_endtone 4.933708 (!H- H-H%) ) ( nn.tobi_endtone -14.993470 (L-L%) ) ( nn.tobi_endtone -3.784284 (L-) ) ( nn.tobi_endtone -15.505132 (H-L% !H-L% -X?) ) ( nn.tobi_endtone -11.352400 (L-H%) ) ( nn.tobi_endtone -5.551627 (H-) ) ( nn.tobi_endtone -0.661581 (!H- H-H%) ) ( pp.old_syl_break -3.367677 ) ( p.old_syl_break 0.641755 ) ( old_syl_break -0.659002 ) ( n.old_syl_break 1.217358 ) ( nn.old_syl_break 2.974502 ) ( pp.stress 1.588098 ) ( p.stress 3.693430 ) ( stress 2.009843 ) ( n.stress 1.645560 ) ( nn.stress 1.926870 ) ( syl_in 1.048362 ) ( syl_out 0.315553 ) ( ssyl_in -2.096079 ) ( ssyl_out 0.303531 ) ( asyl_in -4.257915 ) ( asyl_out -2.422424 ) ( last_accent -0.397647 ) ( next_accent -0.418613 ) ( sub_phrases -5.472055 ) )) (set! f2b_f0_lr_mid '( ( Intercept 169.183377 ) ( R:SylStructure.parent.R:Token.parent.EMPH 10.0 ) ( pp.tobi_accent 4.923247 (H*) ) ( pp.tobi_accent 0.955474 (!H*) ) ( pp.tobi_accent 1.193597 (*? X*? H*!H* * L+H* L+!H*) ) ( pp.tobi_accent 1.501383 (L*) ) ( pp.tobi_accent 7.992120 (L*+H L*+!H) ) ( p.tobi_accent 16.603350 (H*) ) ( p.tobi_accent 11.665814 (!H*) ) ( p.tobi_accent 13.063298 (*? X*? H*!H* * L+H* L+!H*) ) ( p.tobi_accent -2.288798 (L*) ) ( p.tobi_accent 29.168430 (L*+H L*+!H) ) ( tobi_accent 34.517868 (H*) ) ( tobi_accent 22.349656 (!H*) ) ( tobi_accent 23.551548 (*? X*? H*!H* * L+H* L+!H*) ) ( tobi_accent -14.117284 (L*) ) ( tobi_accent -5.978760 (L*+H L*+!H) ) ( n.tobi_accent -1.914945 (H*) ) ( n.tobi_accent 5.249441 (!H*) ) ( n.tobi_accent -1.929947 (*? X*? H*!H* * L+H* L+!H*) ) ( n.tobi_accent -3.287877 (L*) ) ( n.tobi_accent -4.980375 (L*+H L*+!H) ) ( nn.tobi_accent -6.147251 (H*) ) ( nn.tobi_accent 8.408949 (!H*) ) ( nn.tobi_accent 3.193500 (*? X*? H*!H* * L+H* L+!H*) ) ( nn.tobi_accent 1.323099 (L*) ) ( nn.tobi_accent 9.148058 (L*+H L*+!H) ) ( pp.tobi_endtone 4.255273 (L-L%) ) ( pp.tobi_endtone -1.033377 (L-) ) ( pp.tobi_endtone 11.992045 (H-L% !H-L% -X?) ) ( pp.tobi_endtone 6.989573 (L-H%) ) ( pp.tobi_endtone 2.598854 (H-) ) ( pp.tobi_endtone 12.178307 (!H- H-H%) ) ( R:Syllable.p.tobi_endtone -4.397973 (L-L%) ) ( R:Syllable.p.tobi_endtone -6.157077 (L-) ) ( R:Syllable.p.tobi_endtone 5.530608 (H-L% !H-L% -X?) ) ( R:Syllable.p.tobi_endtone 6.938086 (L-H%) ) ( R:Syllable.p.tobi_endtone 6.162763 (H-) ) ( R:Syllable.p.tobi_endtone 8.035727 (!H- H-H%) ) ( tobi_endtone -19.357902 (L-L%) ) ( tobi_endtone -13.877759 (L-) ) ( tobi_endtone -6.176061 (H-L% !H-L% -X?) ) ( tobi_endtone -7.328882 (L-H%) ) ( tobi_endtone 12.694193 (H-) ) ( tobi_endtone 30.923398 (!H- H-H%) ) ( n.tobi_endtone -17.727785 (L-L%) ) ( n.tobi_endtone -2.539592 (L-) ) ( n.tobi_endtone -8.126830 (H-L% !H-L% -X?) ) ( n.tobi_endtone -8.701685 (L-H%) ) ( n.tobi_endtone -1.006439 (H-) ) ( n.tobi_endtone 6.834498 (!H- H-H%) ) ( nn.tobi_endtone -15.407530 (L-L%) ) ( nn.tobi_endtone -2.974196 (L-) ) ( nn.tobi_endtone -12.287673 (H-L% !H-L% -X?) ) ( nn.tobi_endtone -7.621437 (L-H%) ) ( nn.tobi_endtone -0.458837 (H-) ) ( nn.tobi_endtone 3.170632 (!H- H-H%) ) ( pp.old_syl_break -4.196950 ) ( p.old_syl_break -5.176929 ) ( old_syl_break 0.047922 ) ( n.old_syl_break 2.153968 ) ( nn.old_syl_break 2.577074 ) ( pp.stress -2.368192 ) ( p.stress 1.080493 ) ( stress 1.135556 ) ( n.stress 2.447219 ) ( nn.stress 1.318122 ) ( syl_in 0.291663 ) ( syl_out -0.411814 ) ( ssyl_in -1.643456 ) ( ssyl_out 0.580589 ) ( asyl_in -5.649243 ) ( asyl_out 0.489823 ) ( last_accent 0.216634 ) ( next_accent 0.244134 ) ( sub_phrases -5.758156 ) )) (set! f2b_f0_lr_end '( ( Intercept 169.570381 ) ( R:SylStructure.parent.R:Token.parent.EMPH 10.0 ) ( pp.tobi_accent 3.594771 (H*) ) ( pp.tobi_accent 0.432519 (!H*) ) ( pp.tobi_accent 0.235664 (*? X*? H*!H* * L+H* L+!H*) ) ( pp.tobi_accent 1.513892 (L*) ) ( pp.tobi_accent 2.474823 (L*+H L*+!H) ) ( p.tobi_accent 11.214208 (H*) ) ( p.tobi_accent 9.619350 (!H*) ) ( p.tobi_accent 9.084690 (*? X*? H*!H* * L+H* L+!H*) ) ( p.tobi_accent 0.519202 (L*) ) ( p.tobi_accent 26.593112 (L*+H L*+!H) ) ( tobi_accent 25.217589 (H*) ) ( tobi_accent 13.759851 (!H*) ) ( tobi_accent 17.635192 (*? X*? H*!H* * L+H* L+!H*) ) ( tobi_accent -12.149974 (L*) ) ( tobi_accent 13.345913 (L*+H L*+!H) ) ( n.tobi_accent 4.944848 (H*) ) ( n.tobi_accent 7.398383 (!H*) ) ( n.tobi_accent 1.683011 (*? X*? H*!H* * L+H* L+!H*) ) ( n.tobi_accent -6.516900 (L*) ) ( n.tobi_accent -6.768201 (L*+H L*+!H) ) ( nn.tobi_accent -4.335797 (H*) ) ( nn.tobi_accent 5.656462 (!H*) ) ( nn.tobi_accent 0.263288 (*? X*? H*!H* * L+H* L+!H*) ) ( nn.tobi_accent 1.022002 (L*) ) ( nn.tobi_accent 6.702368 (L*+H L*+!H) ) ( pp.tobi_endtone 10.274958 (L-L%) ) ( pp.tobi_endtone 3.129947 (L-) ) ( pp.tobi_endtone 15.476240 (H-L% !H-L% -X?) ) ( pp.tobi_endtone 10.446935 (L-H%) ) ( pp.tobi_endtone 6.104384 (H-) ) ( pp.tobi_endtone 14.182688 (!H- H-H%) ) ( R:Syllable.p.tobi_endtone 1.767454 (L-L%) ) ( R:Syllable.p.tobi_endtone -1.040077 (L-) ) ( R:Syllable.p.tobi_endtone 18.438093 (H-L% !H-L% -X?) ) ( R:Syllable.p.tobi_endtone 8.750018 (L-H%) ) ( R:Syllable.p.tobi_endtone 5.000340 (H-) ) ( R:Syllable.p.tobi_endtone 10.913437 (!H- H-H%) ) ( tobi_endtone -12.637935 (L-L%) ) ( tobi_endtone -13.597961 (L-) ) ( tobi_endtone -6.501965 (H-L% !H-L% -X?) ) ( tobi_endtone 8.747483 (L-H%) ) ( tobi_endtone 15.165833 (H-) ) ( tobi_endtone 50.190326 (!H- H-H%) ) ( n.tobi_endtone -16.965781 (L-L%) ) ( n.tobi_endtone -5.222475 (L-) ) ( n.tobi_endtone -7.358555 (H-L% !H-L% -X?) ) ( n.tobi_endtone -7.833168 (L-H%) ) ( n.tobi_endtone 4.701087 (H-) ) ( n.tobi_endtone 10.349902 (!H- H-H%) ) ( nn.tobi_endtone -15.369483 (L-L%) ) ( nn.tobi_endtone -2.207161 (L-) ) ( nn.tobi_endtone -9.363835 (H-L% !H-L% -X?) ) ( nn.tobi_endtone -7.052374 (L-H%) ) ( nn.tobi_endtone 2.207854 (H-) ) ( nn.tobi_endtone 5.271546 (!H- H-H%) ) ( pp.old_syl_break -4.745862 ) ( p.old_syl_break -5.685178 ) ( old_syl_break -2.633291 ) ( n.old_syl_break 1.678340 ) ( nn.old_syl_break 2.274729 ) ( pp.stress -2.747198 ) ( p.stress 0.306724 ) ( stress -0.565613 ) ( n.stress 2.838327 ) ( nn.stress 1.285244 ) ( syl_in 0.169955 ) ( syl_out -1.045661 ) ( ssyl_in -1.487774 ) ( ssyl_out 0.752405 ) ( asyl_in -5.081677 ) ( asyl_out 3.016218 ) ( last_accent 0.312900 ) ( next_accent 0.837992 ) ( sub_phrases -5.397805 ) )) ;; groups ;; tobi_accent_1 25.217589 (H*) ) ;; tobi_accent_2 13.759851 (!H*) ) ;; tobi_accent_3 17.635192 (*? X*? H*!H* * L+H* L+!H*) ) ;; tobi_accent_4 -12.149974 (L*) ) ;; tobi_accent_5 13.345913 (L*+H L*+!H) ) ;; tobi_endtone_1 10.274958 (L-L%) ) ;; tobi_endtone_2 3.129947 (L-) ) ;; tobi_endtone_3 15.476240 (H-L% !H-L% -X?) ) ;; tobi_endtone_4 10.446935 (L-H%) ) ;; tobi_endtone_5 6.104384 (H-) ) ;; tobi_endtone_6 14.182688 (!H- H-H%) ) (provide 'f2bf0lr) festival/lib/tobi_rules.scm0000644000175000017500000010515707611241252014431 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Authors: Robert A. J. Clark and Alan W Black ;;; Modifications and Checking: ;;; Gregor Moehler (moehler@ims.uni-stuttgart.de) ;;; Matthew Stone (mdstone@cs.rutgers.edu) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Generate F0 points from tobi labels using rules given in: ;;; Jilka, Moehler & Dogil (forthcomming in Speech Communications) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; *** Converted to new Relation architecture -- but not checked yet -- awb ;;; -> crude (beta) checking: gm in Dec. 98 ;;; ;;; -> fixed TAKEOVER bug that used time value ;;; as pitch target (!) - MDS 1/02 ;;; -> hacked around bunches of target overlap problems - MDS 1/02 ;;; -> added primitive pitch range controls ;;; ;;; Known problems and bugs: ;;; Can't currently use voicing intervals which cross syllable boundaries, ;;; so pre/post-nuclear tones are currently places 0.2s before/after the ;;; nuclear tone even if no voicing occurs. Failing this they default a ;;; percentage of the voicing for that syllable. ;;; ;;; Don't know about target points ahead of the current syllable. ;;; (As you need to know what comes before them to calculate them) ;;; So: post accent tones are placed 0.2 ahead if following syllable exists ;;; ends before 0.2 from starred target and is not accented ;;; The H-target of the H+!H* is 0.2 sec instead of 0.15 sec before ;;; starred tone. ;;; ;;; Multi-utterance input has not been tested. ;;; ;;; !H- does not generate any targets ;;; ;;; Unfortunaltely some other modules may decide to put pauses in the ;;; middle of a phrase ;;; ;;; valleys are not tested yet ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; To use this in a voice ;;; (require 'tobi_rules) ;;; And in the voice call ;;; (setup_tobi_f0_method) ;;; Set the following for your speaker's F0 range ;;; (Parameter.set 'Default_Topline 146) ;;; (Parameter.set 'Default_Start_Baseline 61) ;;; (Parameter.set 'Valley_Dip 75) ;; level of debug printout (set! printdebug 0) (define (setup_tobi_f0_method) "(setup_tobi_f0_method) Set up parameters for current voice to use the implementaion of ToBI labels to F0 targets by rule." (Parameter.set 'Int_Method Intonation_Tree) (Parameter.set 'Int_Target_Method Int_Targets_General) (set! int_accent_cart_tree no_int_cart_tree) ; NONE always (set! int_tone_cart_tree no_int_cart_tree) ; NONE always (set! int_general_params (list (list 'targ_func tobi_f0_targets))) ; we will return a list of f0 targets here (Parameter.set 'Phrase_Method 'cart_tree) (set! phrase_cart_tree tobi_label_phrase_cart_tree) ; redefines the phrasebreak tree t) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;; ;;;;;; Define and set the new f0 rules ;;;;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Set global parameters ;;; You may want to reset these for different speakers (Parameter.set 'Default_Topline 146) ;146 (Parameter.set 'Default_Start_Baseline 61) ;61 (Parameter.set 'Current_Topline (Parameter.get 'Default_Topline)) (Parameter.set 'Current_Start_Baseline (Parameter.get 'Default_Start_Baseline)) (Parameter.set 'Current_End_Baseline (Parameter.get 'Current_Start_Baseline)) (Parameter.set 'Downstep_Factor 0.70) (Parameter.set 'Valley_Dip 75) ;;; function to add target points on a given syllable and fill in ;;; targets where necessary (define (tobi_f0_targets utt syl) "(tobi_f0_targets UTT ITEM) Returns a list of targets for the given syllable." (if (and (>= printdebug 1) (not(equal? 0 (item.feat syl "R:Intonation.daughter1.name")))) (format t "### %l (%.2f %.2f) %l ptarg: %l ###\n" (item.name syl) (item.feat syl "syllable_start")(item.feat syl "syllable_end") (item.feat syl "R:Intonation.daughter1.name") (ttt_last_target_time syl))) ;; only continue if there is a Word related to this syllable ;; I know there always should be, but there might be a bug elsewhere (cond ((not(equal? 0 (item.feat syl "R:SylStructure.parent.name"))) ; get current label. This assumes that there is only one accent and ; one endtone on a syllable. Although there can be one of each. (let ((voicing (ttt_get_voice_times syl)) ; voicing interval (pvoicing (ttt_get_voice_times ; previous voicing (item.relation.prev syl 'Syllable))) (nvoicing (ttt_get_voice_times ; next voicing (item.relation.next syl 'Syllable)))) ; if first syl of phrase set Phrase_Start and Phrase_End parameters ; and reset downstep (currently does so on big and little breaks.) ; only assignes Default values at this stage ; maybe trained from CART later - first steps now - MDS ; following Moehler and Mayer, SSW 2001 (if (eq 0 (item.feat syl 'syl_in)) ;; GM maybe something better needed here? (progn (Parameter.set 'Phrase_Start (item.feat syl 'R:SylStructure.parent.R:Phrase.last.word_start)) (Parameter.set 'Phrase_End (item.feat syl 'R:SylStructure.parent.R:Phrase.last.word_end)) (Parameter.set 'Current_Topline (/ (* (wagon syl ttt_topline_tree) (Parameter.get 'Default_Topline)) 100)) (Parameter.set 'Current_Start_Baseline (/ (* (wagon syl ttt_baseline_tree) (Parameter.get 'Default_Start_Baseline)) 100)) (Parameter.set 'Current_End_Baseline (Parameter.get 'Current_Start_Baseline)) (if (>= printdebug 3) (begin (print (format nil "new range: %f %f %f" (Parameter.get 'Current_Topline) (Parameter.get 'Current_Start_Baseline) (Parameter.get 'Current_End_Baseline) )))) )) ; do stuff (should go only if there is an accent/boundary?) (let ((new_targets (ttt_to_targets syl (wagon syl ttt_starttone_tree) voicing pvoicing nvoicing 'Starttones))) (set! new_targets (append new_targets (ttt_to_targets syl (wagon syl ttt_accent_tree) voicing pvoicing nvoicing 'Accents))) (set! new_targets (append new_targets (ttt_to_targets syl (wagon syl ttt_endtone_tree) voicing pvoicing nvoicing 'Endtones))) (if (and(not(equal? new_targets nil)) (>= printdebug 2)) (begin (format t ">> Targets: %l\n" new_targets) (format t ">> LastTarget: %l\n" (last new_targets)) )) new_targets))))) ;;; CART tree to specify no accents (set! no_int_cart_tree ' ((NONE))) ;;; ;;; Relate phrasing to boundary tones. ;;; Added downstepped tones - MDS (set! tobi_label_phrase_cart_tree ' ((tone in ("L-" "H-" "!H-")) ((B)) ((tone in ("H-H%" "H-L%" "!H-L%" "L-L%" "L-H%")) ((BB)) ((NB))))) ;;; ;;; The other functions ;;; ;;; process a list of relative targets and convert to actual targets (define (ttt_to_targets syl rlist voicing pvoicing nvoicing type) "Takes a list of target sets and returns a list of targets." (if (or (and (>= printdebug 2) rlist (atom (caar rlist)) (not (equal? 'NONE (caar rlist))) (not (equal? '(NONE) (caar rlist)))) (>= printdebug 3)) (begin (print "Entering ttt_to_targets with:") (print (format nil "rlist: %l vc: %l pvc: %l nvc: %l type: %s" rlist voicing pvoicing nvoicing type)))) (cond ;; nowt ((eq (length rlist) 0) ()) ;; a single target set ((atom (car (car rlist))) (cond ((eq type 'Accents) (ttt_accent_set_to_targets syl rlist voicing pvoicing nvoicing)) ((eq type 'Starttones) (ttt_bound_set_to_targets syl rlist voicing pvoicing)) ((eq type 'Endtones) (ttt_bound_set_to_targets syl rlist voicing pvoicing)) (t (error "unknown target set encountered in ttt_to_targets")))) ;; list of target sets ((atom (car (car (car rlist)))) (append (ttt_to_targets syl (cdr rlist) voicing pvoicing nvoicing type) (ttt_to_targets syl (car rlist) voicing pvoicing nvoicing type))) ;; error (t (error "something strange has happened in ttt_to_targets")))) ;; process a starttone/endtone target set. (define (ttt_bound_set_to_targets syl tset voicing pvoicing) "takes a start/endtone target set and returns a list of target points." (if (>= printdebug 3) (begin (print "Entering ttt_bound_set_to_targets with:") (pprintf (format nil "tset: %l vc: %l pvc: %l" tset voicing pvoicing)))) (cond ;; usually target given is NONE. (also ignore unknown!) ((or (eq (car (car tset)) 'NONE) (eq (car (car tset)) 'UNKNOWN)) nil) ;; a pair of target pairs ((eq (length tset) 2) (list (ttt_get_target (car tset) voicing) (ttt_get_target (car (cdr tset)) voicing))) ;; single target pair ((eq (length tset) 1) (cond ;; an actual target pair ((not (null (cdr (car tset)))) (list (ttt_get_target (car tset) voicing))) ;; a TAKEOVER marker ((eq (car (car tset)) 'TAKEOVER) (list (list (ttt_interval_percent voicing 0) (ttt_last_target_value syl)))) (t (error "unknown target pair in ttt_bound_set_to_targets")))) (t (error "unknown target set type in ttt_bound_set_to_targets")))) ;; process an accent target set. (define (ttt_accent_set_to_targets syl tset voicing pvoicing nvoicing) "takes a accent target set and returns a list of target points." (if (>= printdebug 3) (begin (print "Entering ttt_accent_set_to_targets with:") (pprintf (format nil "tset: %l vc: %l pvc: %l nvc: %l" tset voicing pvoicing nvoicing)))) (cond ;; single target in set ((null (cdr tset)) (cond ; target given is NONE. ((or (eq (car (car tset)) 'NONE) (eq (car (car tset)) 'UNKNOWN)) nil) ; V1 marker ((eq (car (car tset)) 'V1) (let ((target_time (+ (/ (- (next_accent_start syl) (ttt_last_target_time syl)) 2.0) (ttt_last_target_time syl)))) (list (list target_time (ttt_accent_pitch (Parameter.get 'Valley_Dip) target_time))))) ; V2 marker ((eq (car (car tset)) 'V2) (let ((target_time (+ (ttt_last_target_time syl) 0.25))) (list (list target_time (ttt_accent_pitch (Parameter.get 'Valley_Dip) target_time))))) ; V3 marker ((eq (car (car tset)) 'V3) (let ((target_time (- (next_accent_start syl) 0.25))) (list (list target_time (ttt_accent_pitch (Parameter.get 'Valley_Dip) target_time))))) ; single target pair (t (list (ttt_get_target (car tset) voicing))))) ;; a pair of targets ((length tset 2) (cond ;; a *ed tone with PRE type tone (as in L+H*) ((eq (car (car tset)) 'PRE) (let ((star_target (ttt_get_target (car (cdr tset)) voicing)) (last_target (parse-number(ttt_last_target_time syl)))) (cond ; normal 0.2s case (currently doesn't check for voicing) ((and (eqv? 0 (ip_initial syl)) (> (- (car star_target) 0.2) last_target)) (list (list (- (car star_target) 0.2) (ttt_accent_pitch (car (cdr (car tset))) (- (car star_target) 0.2))) ; the time star_target)) ; 90% prev voiced if not before last target - Added back in MDS, ; with parse-number added and new check for ip_initial ((and (eqv? 0 (ip_initial syl)) (> (parse-number (ttt_interval_percent pvoicing 90)) (parse-number (ttt_last_target_time syl)))) (list (list (ttt_interval_percent pvoicing 90) (ttt_accent_pitch (car (cdr (car tset))) (ttt_interval_percent pvoicing 90))) star_target)) ; otherwise (UNTESTED) [NOTE: Voicing for this syllable only] (t (list (list (ttt_interval_percent voicing 20) (ttt_accent_pitch (car (cdr (car tset))) (ttt_interval_percent voicing 20))) star_target))))) ; a *ed tone with POST type tone (as L*+H) ((eq (car(car(cdr tset))) 'POST) (let ((star_target (ttt_get_target (car tset) voicing)) (next_target nil ) ; interesting problem (next_syl (item.next syl))) (cond ; normal 0.2s case (UNTESTED) ((and (not (equal? next_syl nil)) (eq 0 (item.feat next_syl "accented"))) (cond ((< (+ (car star_target) 0.2) (item.feat next_syl "syllable_end")) (list star_target (list (+ (car star_target) 0.2) (ttt_accent_pitch (car (cdr (car (cdr tset)))) (+ (car star_target) 0.2) )))) (t (list star_target (list (ttt_interval_percent nvoicing 90) (ttt_accent_pitch (car (cdr (car (cdr tset)))) (ttt_interval_percent nvoicing 90) )))))) ; 20% next voiced (BUG: Can't do this as the next target hasn't been ; calculated yet!) (nil nil) ;otherwise (UNTESTED) (t (list star_target (list (ttt_interval_percent voicing 90) (ttt_accent_pitch (car (cdr (car (cdr tset)))) (ttt_interval_percent voicing 90) ))))))) (t ;; This case didn't use to happen, but now must ;; to avoid +H's clobbering endtones - MDS's hack. (list (ttt_get_target (car tset) voicing) (ttt_get_target (cadr tset) voicing))))) ;; something else... (t (error (format nil "unknown accent set in ttt_accent_set_to_targets: %l" tset))))) (define (ttt_get_target pair voicing) "Returns actual target pair, usually for a stared tone." (if (>= printdebug 4) (begin (print "Entering ttt_get_target with:") (pprintf pair) (pprintf voicing))) (list (ttt_interval_percent voicing (car pair)) (ttt_accent_pitch (car (cdr pair)) (ttt_interval_percent voicing (car pair))))) (define (ttt_accent_pitch value time) "Converts a accent pitch entry to a pitch value." (if (>= printdebug 4) (begin (print "Entering ttt_accent_pitch with:") (pprintf value))) (cond ;; a real value ((number? value) (ttt_interval_percent (list (ttt_get_current_baseline time) (Parameter.get 'Current_Topline)) value)) ;; Downstep then Topline ((eq value 'DHIGH) (progn (Parameter.set 'Current_Topline (+ (ttt_get_current_baseline time) (* (Parameter.get 'Downstep_Factor) (- (Parameter.get 'Current_Topline) (ttt_get_current_baseline time))))) (ttt_interval_percent (list (ttt_get_current_baseline time) (Parameter.get 'Current_Topline)) 100))) ;; Unknown (t (error "Unknown accent pitch value encountered")))) (define (ttt_get_current_baseline v) "Returns the current declined baseline at time v." (if (>= printdebug 4) (begin (print "Entering ttt_get_current_baseline with:") (pprintf v))) (let ((h (Parameter.get 'Current_Start_Baseline)) (l (Parameter.get 'Current_End_Baseline)) (e (Parameter.get 'Phrase_End)) (s (Parameter.get 'Phrase_Start))) (- h (* (/ (- h l) (- e s)) (- v s))))) ;;; find the time n% through an inteval (define (ttt_interval_percent pair percent) "Returns the time that is percent percent thought the pair." (if (>= printdebug 4) (begin (print "Entering ttt_interval_percent with:") (pprintf (format nil "%l, %l" pair percent)))) (cond ; no pair given: just return nil ((null pair) nil) ; otherwise do the calculation (t (let ((start (car pair)) (end (car(cdr pair)))) (+ start (* (- end start) (/ percent 100))))))) ;;; Getting start and end voicing times in a syllable (define (ttt_get_voice_times syl_item) "Returns a pair of start time of first voiced phone in syllable and end of last voiced phone in syllable, or nil if syllable is nil" (cond ((null syl_item) nil) (t (let ((segs (item.relation.daughters syl_item "SylStructure"))) (list (item.feat (ttt_first_voiced segs) "segment_start") (item.feat (ttt_first_voiced (reverse segs)) "end")))))) (define (ttt_first_voiced segs) "Returns first segment that is voiced (vowel or voiced consonant) returns last segment if all are unvoiced." (cond ((null (cdr segs)) (car segs)) ;; last possibility ((equal? "+" (item.feat (car segs) "ph_vc")) (car segs)) ((equal? "+" (item.feat (car segs) "ph_cvox")) (car segs)) (t (ttt_first_voiced (cdr segs))))) ;;; ttt_last_target has bifurcated into ;;; ttt_last_target_time and ;;; ttt_last_target_value ;;; to fix a place where f0 was set to last target time! ;;; - MDS (define (ttt_last_target_time syl) "Returns the end of the most recent previous target in the utterance or nil if there is not one present " (if (>= printdebug 3) (begin (print "Entering ttt_last_target_time") (print syl)) ) (let ((target (ttt_last_target syl))) (if (null? target) nil (item.feat target "R:Target.daughter1.pos")))) (define (ttt_last_target_value syl) "Returns the pitch of the most recent previous target in the utterance or nil if there is not one present " (if (>= printdebug 3) (begin (print "Entering ttt_last_target_time") (print syl)) ) (let ((target (ttt_last_target syl))) (if (null? target) nil (item.feat target "R:Target.daughter1.f0")))) ;; Changed to scan through segments in the segment relation, ;; to catch (notional) targets on pauses. - MDS ;; ;;; associated segments are: ;;; - the segments in the word ;;; - subsequent segments not in the syllable structure ;;; and on the first word, preceding segments ;;; not in the syllable structure (define (ttt_collect_following seg accum) (if (or (null? seg) (not (null? (item.relation seg 'SylStructure)))) accum (ttt_collect_following (item.next seg) (cons seg accum)))) (define (ttt_last_target syl) "Returns the most recent previous target in the utterance or nil if there is not one present " (if (>= printdebug 3) (begin (print "Entering ttt_last_target") (print syl)) ) (let ((prev_syl (item.relation.prev syl 'Syllable))) (cond ; ((symbol-bound? 'new_targets) (last (caar new_targets))) ((null prev_syl) nil) ((ttt_last_target_segs (ttt_collect_following (item.relation.next (item.relation.daughtern prev_syl "SylStructure") "Segment") (reverse (item.relation.daughters prev_syl "SylStructure"))))) ;list of segments of prev. syllable ;in reverse order, with pauses ;prepended. (t (ttt_last_target prev_syl))))) (define (ttt_last_target_segs segs) "Returns the first target no earlier than seg or nil if there is not one " (if (>= printdebug 4) (begin (print "Entering ttt_last_target_segs with:") (pprintf (format nil "%l" segs)) )) (cond ((null segs) nil) ((and (> (parse-number (item.feat (car segs) "R:Target.daughter1.f0")) 0) (eq 0 (item.feat (car segs) "R:SylStructure.parent.lisp_lh_condition")) (eq 0 (item.feat (car segs) "R:SylStructure.parent.lisp_hl_condition")) (eq 0 (item.feat (car segs) "R:SylStructure.parent.lisp_valley_condition"))) (car segs)) (t (ttt_last_target_segs (cdr segs))))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;; ;;;;;; CART TREES (ttt - tobi to target) ;;;;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Return a list of target lists. A target list comprises of a list ;;; of related targets (ie for the L and H in L+H*), just to confuse ;;; matters each target is also a list! (pos pitch) ;;; (set! ttt_endtone_tree ; BUG: does it check the current syl for last accent? ' ((tobi_endtone is NONE) ; ususally none ((((NONE)))) ((tobi_endtone is "H-H%") ; H-H% ((((100 120)))) ((tobi_endtone is "L-L%") ; L-L% ((((100 -20)))) ((tobi_endtone is "L-H%") ; L-H% ((lisp_last_accent > 2) ((lisp_last_accent_type is "L*") ((((0 25) (100 40)))) ; paper says 80 but AWB had 40 ((((0 0) (100 40))))) ((lisp_last_accent_type is "L*") ((((100 40)))) ((((50 0) (100 40)))))) ((tobi_endtone is "H-L%") ; H-L% ((lisp_last_accent_type is "L*") ((tobi_accent is"L*") ((((50 100) (100 100)))) ((((0 100) (100 100))))) ((((100 100))))) ((tobi_endtone is "!H-L%") ; !H-L% ((lisp_last_accent_type is "L*") ((tobi_accent is"L*") ((((50 DHIGH) (100 100)))) ((((0 DHIGH) (100 100))))) ((((100 DHIGH))))) ((tobi_endtone is "H-") ((((100 100)))) ((tobi_endtone is "!H-") ((((100 DHIGH)))) ((tobi_endtone is "L-") ((((100 0)))) ((((UNKNOWN)))))))))))))) (set! ttt_starttone_tree ' ((lisp_ip_initial = 1) ((tobi_endtone is "%H") ((((0 100)))) ((p.tobi_endtone in ("H-" "!H-" "L-")) ((((TAKEOVER)))) ; takeover case ((tobi_accent is NONE) ((lisp_next_accent > 2) ; default cases (dep. on whether next target is low) ((lisp_next_accent_type in ("L*" "L*+H" "L*+!H" "L+H*" "L+!H*" "L-" "L-H%" "L-L%")) ((((0 50)(100 25)))) ((((0 50)(100 75))))) ((lisp_next_accent_type in ("L*" "L*+H" "L*+!H" "L+H*" "L+!H*" "L-" "L-H%" "L-L%")) ((((0 30)))) ((((0 70)))))) ((tobi_accent in ("L*" "L*+H" "L*+!H" "L+H*" "L+!H*" "L-" "L-H%" "L-L%")) ((((0 30)))) ((((0 70)))))))) ((((NONE)))))) ; otherwise (and usually) nothing. ;; Redone after Jilka, Moehler and Dogil ;; - But treating one-syllable-ip's like ;; last-syllable-of-ip's in cases of ;; two tone switches per syllable (e.g. H* L-H%). ;; - And (hack) a 70% target for the initial ;; H*'s of phrases when the next accent is L+H* ;; - MDS (set! ttt_accent_tree ' ((tobi_accent is "H*" ) ; H* ((lisp_ip_final = 1) ((lisp_ip_one_syllable_case = 1) ((((50 100)))) ((((25 100))))) ((lisp_hstar_weak_target = 1) ((((60 70)))) ((lisp_ip_initial = 1) ((((85 100)))) ((((60 100))))))) ((tobi_accent is "!H*" ) ; !H* ((lisp_ip_final = 1) ((lisp_ip_one_syllable_case = 1) ((((50 DHIGH)))) ((((25 DHIGH))))) ((lisp_ip_initial = 1) ((((85 DHIGH)))) ((((60 DHIGH)))))) ((tobi_accent is "L*" ) ; L* ((lisp_ip_final = 1) ((lisp_ip_one_syllable_case = 1) ((((50 0)))) ((((25 0))))) ((lisp_ip_initial = 1) ((((85 0)))) ((((60 0)))))) ((tobi_accent is "L+H*") ; L+H* ((lisp_ip_final = 1) ((lisp_ip_one_syllable_case = 1) ((((PRE 20) (50 100)))) ; JMD estimated 70 ((((PRE 20) (25 100))))) ((lisp_ip_initial = 1) ((((PRE 20) (90 100)))) ((((PRE 20) (75 100)))))) ((tobi_accent is "L+!H*") ; L+!H* ((lisp_ip_final = 1) ((lisp_ip_one_syllable_case = 1) ((((PRE 20) (70 DHIGH)))) ((((PRE 20) (25 DHIGH))))) ((lisp_ip_initial = 1) ((((PRE 20) (90 DHIGH)))) ((((PRE 20) (75 DHIGH)))))) ((tobi_accent is "L*+H") ; L*+H ((lisp_ip_final = 1) ((lisp_ip_one_syllable_case = 1) ((((35 0) (80 100)))) ; POST would clobber endtones ((((15 0) (40 100))))) ; POST would clobber endtones - MDS ((lisp_ip_initial = 1) ((((55 0) (POST 100)))) ((((40 0) (POST 100)))))) ((tobi_accent is "L*+!H") ; L*+!H ((lisp_ip_final = 1) ((lisp_ip_one_syllable_case = 1) ((((35 0) (80 DHIGH)))) ; POST would clobber endtones - MDS ((((15 0) (40 DHIGH))))) ; POST would clobber endtones - MDS ((lisp_ip_initial = 1) ((((55 0) (POST DHIGH)))) ((((40 0) (POST DHIGH)))))) ((tobi_accent is "H+!H*") ; H+!H* ((lisp_ip_final = 1) ((lisp_ip_one_syllable_case = 1) ((((PRE 143) (60 DHIGH)))) ; the 143 is a hack to level out the downstep ((((PRE 143) (20 DHIGH))))) ((lisp_ip_initial = 1) ((((PRE 143) (90 DHIGH)))) ((((PRE 143) (60 DHIGH)))))) ((lisp_lh_condition = 1) ((((100 75)))) ((lisp_lh_condition = 2) ((((0 90)))) ((lisp_hl_condition = 1) ((((100 25)))) ((lisp_valley_condition = 1) ((((V1 85)))) ((lisp_valley_condition = 2) ((((V2 70)))) ((lisp_valley_condition = 3) ((((V3 70)))) ((tobi_accent is NONE) ; usually we find no accent ((((NONE)))) ((((UNKNOWN)))))))))))))))))))) ; UNKNOWN TARGET FOUND ;;; Cart tree to "predict" pitch range ;;; Right now just accesses a feature ;;; "register" following Moehler & Mayer 2001. ;;; Register must be one of ;;; H - primary high register (default): 133% lowest, 92% highest ;;; H-H - expanded high register: 134% lowest, 100% highest ;;; H-L - lowered high register: 128% lowest, 87% highest ;;; L - primary low register: 100% lowest, 73% highest ;;; L-L and HL-L - low compressed: 100% lowest, 66% highest ;;; HL - expanded register: 100% lowest, 84% highest ;;; HL-H - complete register: 100% lowest, 96% highest ;;; For their speaker, ,BASELINE was 42% of PEAK (set! ttt_topline_tree ' ((R:SylStructure.parent.register is "H") (92) ((R:SylStructure.parent.register is "H-H") (100) ((R:SylStructure.parent.register is "H-L") (87) ((R:SylStructure.parent.register is "L") (73) ((R:SylStructure.parent.register is "L-L") (66) ((R:SylStructure.parent.register is "HL") (84) ((R:SylStructure.parent.register is "HL-H") (96) (92))))))))) (set! ttt_baseline_tree ' ((R:SylStructure.parent.register is "H") (133) ((R:SylStructure.parent.register is "H-H") (134) ((R:SylStructure.parent.register is "H-L") (128) ((R:SylStructure.parent.register is "L") (100) ((R:SylStructure.parent.register is "L-L") (100) ((R:SylStructure.parent.register is "HL") (100) ((R:SylStructure.parent.register is "HL-H") (100) (133))))))))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;; ;;;;;; Lisp Feature functions. ;;;;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (valley_condition syl) "(valley_condition syl) Function to determine if a lowered target between two high target points is needed in this syllable. Returns: 0 - no target required 1 - the single target case 2 - the first of the two target case 3 - the second of the two target case " (if (>= printdebug 4) (begin (print "Entering valley_condition"))) (cond ((and (eq 0 (item.feat syl 'accented)) (string-matches (next_accent_type syl) "\\(H\\*\\|H\\-\\|H\\-L\\%\\|H\\-H\\%\\|\\!H\\*\\|\\!H\\-\\|\\!H\\-L\\%\\|\\!H\\-H\\%\\)") (string-matches (last_accent_type syl) "\\(H\\*\\|L\\+H\\*\\|L\\*\\+H\\\\|\\!H\\*\\|L\\+\\!H\\*\\|L\\*\\+\\!H\\)")) ;GM: excluded %H (returns nil for last target) (let ((nas (next_accent_start syl)) (syls (item.feat syl 'syllable_start)) (syle (item.feat syl 'syllable_end)) (las (ttt_last_target_time syl))) (if (>= printdebug 3) (begin (print (format nil "nas: %l syls: %l syle %l las %l" nas syls syle las)))) (cond ((and (< (- nas las) 0.5) (> (- nas las) 0.25) (< syls (+ (/ (- nas las) 2.0) (ttt_last_target_time syl))) (> syle (+ (/ (- nas las) 2.0) (ttt_last_target_time syl)))) 1) ((and (> (- nas las) 0.5) (< syls (+ (ttt_last_target_time syl) 0.25)) (> syle (+ (ttt_last_target_time syl) 0.25))) 2) ((and (> (- nas las) 0.5) (< syls (- nas 0.25)) (> syle (- nas 0.25))) 3) (t 0)))) (t 0))) (define (lh_condition syl) "(lh_condition syl) Function to determine the need for extra target points between an L and an H Returns: 1 - first extra target required 2 - second extra target required 0 - no target required. " (if (>= printdebug 4) (begin (print "Entering LH_condition"))) (cond ((and (eq 0 (item.feat syl 'accented)) (string-matches (last_accent_type syl) "\\(L\\*\\)") (string-matches (next_accent_type syl) "\\(H\\*\\|H\\-\\|H\\-L\\%\\|H\\-H\\%\\)")) (cond ((and (eq 1 (last_accent syl)) (< 2 (next_accent syl))) 1) ((and (< 2 (last_accent syl)) (eq 1 (next_accent syl))) 2) (t 0))) (t 0))) (define (hl_condition syl) "(lh_condition syl) Function to determine the need for extra target points between an H and an L Returns: 1 - extra target required 0 - no target required. " (if (>= printdebug 4) (begin (print "Entering HL_condition"))) (cond ((and (eq 0 (item.feat syl 'accented)) (string-matches (next_accent_type syl) "\\(L\\*\\|L\\+H\\*\\|L\\+\\!H\\*\\|L\\*\\+H\\|L\\*\\+!H\\|L\\-\\|L\\-L\\%\\|L-H\\%\\)") (string-matches (last_accent_type syl) "\\(H\\*\\|L\\+H\\*\\|L\\*\\+H\\\\|\\!H\\*\\|L\\+\\!H\\*\\|L\\*\\+\\!H\\|\\%H\\)") ;MDS: added !H's (eq 1 (last_accent syl)) ;; fall faster! -MDS (<= 2 (next_accent syl))) 1) (t 0))) (define (next_accent syl) "(next_accent syl) Wrapper for c++ func ff_next_accent. Returns the number of the syllables to the next accent in the following format. 0 - no next accent 1 - next syllable 2 - next next syllable etc..." (if (>= printdebug 4) (begin (print "Entering next_accent"))) (cond ((eq 0 (next_accent_type syl)) 0) (t (+ (item.feat syl 'next_accent) 1)))) ;; Fixed bug that crashed complex phrase tones. - MDS ;; Not sure how else to get a big number... (define infinity (/ 1 0)) ;; Modified to include current accent as well -MDS (define (last_accent syl) "(last_accent syl) Wrapper for c++ func ff_last_accent. Returns the number of the syllables to the previous accent in the following format. 0 - accent on current syllable 1 - prev syllable 2 - prev to prev syllable etc... infinity - no previous syllable" (if (>= printdebug 4) (begin (print "Entering last_accent"))) (cond ((not (equal? "NONE" (item.feat syl 'tobi_accent))) 0) ((equal? 0 (last_accent_type syl)) infinity) (t (+ (item.feat syl 'last_accent) 1)))) (define (next_accent_type syl) "(next_accent_type syl) Returns the type of the next accent." (cond ((not (eq 0 (item.feat syl "n.R:Intonation.daughter1.name"))) (item.feat syl "n.R:Intonation.daughter1.name")) ((eq 0 (item.feat syl 'syl_out)) 0) ;;GM real ip_final would be better (t (next_accent_type (item.relation.next syl 'Syllable))))) (define (last_accent_type syl) "(last_accent_type syl) Returns the type of the last (previous) accent." (if (>= printdebug 4) (begin (print "Entering last_accent_type"))) (cond ((not (equal? "NONE" (item.feat syl 'p.tobi_endtone))) (item.feat syl 'R:Syllable.p.tobi_endtone)) ((not (equal? "NONE" (item.feat syl 'p.tobi_accent))) (item.feat syl 'R:Syllable.p.tobi_accent)) ((eq 0 (item.feat syl 'syl_in)) 0) ;;GM real ip_initial would be better (t (last_accent_type (item.prev syl 'Syllable))))) (define (next_accent_start syl) "(next_accent_start syl) Returns the start time of the vowel of next accented syllable" (if (>= printdebug 4) (begin (print "Entering next_accent_start"))) (cond ((not (eq 0 (item.feat syl "n.R:Intonation.daughter1.name"))) (item.feat syl "R:Syllable.n.syllable_start")) ;;GM vowel start would be better ((eq 0 (item.feat syl 'syl_out)) 0) (t (next_accent_start (item.relation.next syl 'Syllable))))) ; new features (not used yet) (define (ip_final syl) "(ip_final SYL) returns 1 if the syllable is the final syllable of an ip (intermediate phrase)" (cond ((or (equal? 0 (item.feat syl "syl_out")) (equal? "L-" (item.feat syl "tobi_endtone")) (equal? "H-" (item.feat syl "tobi_endtone")) (equal? "!H-" (item.feat syl "tobi_endtone"))) 1) (t 0))) (define (ip_initial syl) "(ip_initial SYL) returns 1 if the syllable is the initial syllable of an ip (intermediate phrase)" (cond ((equal? 0 (item.feat syl "syl_in")) 1) ((equal? 1 (ip_final (item.relation.prev syl 'Syllable))) 1) (t 0))) ;; NEXT TWO FUNCTIONS ARE NEW - MDS (define (ip_one_syllable_case syl) "(ip_one_syllable_case SYL) returns true if the syllable is the initial syllable of an ip (intermediate phrase) and doesn't itself contain a complex tone that starts opposite this syllable's accent" (if (eqv? 0 (ip_initial syl)) 0 (let ((accent (item.feat syl "tobi_accent")) (tone (item.feat syl "tobi_endtone"))) (cond ((and (equal? tone "L-H%") (or (equal? accent "H*") (equal? accent "!H*") (equal? accent "L+H*") (equal? accent "L+!H*") (equal? accent "L*+H") (equal? accent "L*+!H*") (equal? accent "H+!H*"))) 0) ((and (or (equal? tone "H-L%") (equal? tone "!H-L%")) (equal? accent "L*")) 0) (t 1))))) (define (hstar_weak_target syl) (if (and (equal? 0 (item.feat syl 'asyl_in)) (member (next_accent_type syl) (list "L*" "L*+H" "L*+!H" "L+H*" "L+!H*"))) 1 0)) (provide 'tobi_rules) festival/lib/tilt.scm0000644000175000017500000007545607650552505013257 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Author: Alan W Black, Kurt Dusterhoff, Janet Hitzeman ;;; Date: April 1999 ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Tilt intonation modules, accent/boundary preditions and F0 generation ;;; The F0 generation is done using models as described in ;;; Dusterhoff, K. and Black, A. (1997). "Generating F0 contours for ;;; speech synthesis using the Tilt intonation theory" ;;; (http://www.cstr.ed.ac.uk/awb/papers/esca-int97.ps) ;;; Proceedings of ESCA Workshop of Intonation, pp 107-110, September, ;;; Athens, Greece. ;;; ;;; Intonation_Tilt assigns accents and boundaries by a CART tree ;;; the c and sil nodes are derived directly duration creation ;;; ;;; Int_Targets_Tilt generates the F0 using the CART trees as ;;; described in the paper referenced above. ;;; ;;; THIS CONTAINS *VERY* EXPERIMENTAL CODE ;;; it requires a thoroughly clean up and probably split into ;;; multiple files (defvar int_tilt_params nil "int_tilt_params Parameters for tilt intonation model.") (Parameter.def 'tilt_method 'cart) (define (Intonation_Tilt utt) "(Intonation_Tilt utt) Assign accent and boundary IntEvents to each syllable, and fill in spaces with silence and connections." (let (accent boundary) ;; Create basic intonation relations (utt.relation.create utt 'Intonation) (utt.relation.create utt 'IntonationSyllable) (mapcar (lambda (syl) ;; If first syllable in phrase add phrase_start (if (string-equal "pau" (item.feat syl "R:SylStructure.daughter1_to.Segment.p.name")) (tilt_add_intevent utt syl 'phrase_start)) (set! accent (wagon_predict syl tilt_a_cart_tree)) (set! boundary (wagon_predict syl tilt_b_cart_tree)) ; (format t "%s: accent %s boundary %s\n" ; (item.feat syl "R:WordStructure.root.name") ; accent ; boundary) (if (not (string-equal accent "0")) (tilt_add_intevent utt syl accent)) (if (not (string-equal boundary "0")) (if (and (string-equal boundary "afb") (not (string-equal accent "0"))) (tilt_add_intevent utt syl "fb") ;; can't have a/afb (tilt_add_intevent utt syl boundary))) ;; If last syllable in phrase add phrase_end (if (string-equal "pau" (item.feat syl "R:SylStructure.daughtern_to.Segment.n.name")) (tilt_add_intevent utt syl 'phrase_end))) (utt.relation.items utt 'Syllable)) ;; (utt.relation.print utt 'Intonation) utt)) (define (tilt_add_intevent utt syl name) "(tilt_add_intevent utt syl name) Add a new IntEvent related to syl with name." (let (ie) (set! ie (utt.relation.append utt 'Intonation (list name))) (if (not (item.relation syl 'IntonationSyllable)) (utt.relation.append utt 'IntonationSyllable syl)) (item.relation.append_daughter syl 'IntonationSyllable ie) (if (not (string-matches name "phrase_.*")) (item.set_feat ie "int_event" 1)) ie)) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Fo generate through tilt parameters and F0 rendering ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (Int_Targets_Tilt utt) "(Int_Targets_Tilt utt) Assign Tilt parameters to each IntEvent and then generate the F0 contour and assign targets." (utt.relation.set_feat utt "Intonation" "intonation_style" "tilt") (tilt_assign_parameters utt) ; (tilt_F0_and_targets utt) ;; this has to be C++, sorry ; (tilt_map_f0_range utt) (tilt_to_f0 utt "f0") (tilt_validate utt) utt ) (define (tilt_validate utt) "(tilt_validate utt) Checks that the predicted tilt parameter fall with reasonable limits and modify them where possible to be more reasonable." (mapcar (lambda (ie) (cond ((string-equal (item.name ie) "phrase_end") ;; check previous event does overflow segments ) (t t)) ) (utt.relation.items utt 'Intonation)) ) (define (tilt_map_f0_range utt) "(tilt_map_f0_range utt) In order fo better trained models to be used for voices which don't have the necessary data to train models from the targets may be mapped to a different pitch range. Note this is not optimal as pitch ranges don't map that easily, but the the results can sometimes be better than using a less sophisticated F0 generation model. The method used is to define the mean and standard deviation of the speaker the model was trained on and the mean and standard deciation of the desired speaker. Mapping is by converting the actual F0 value to zscores (distance from mean in number of stddev) and back into the other domain. The variable int_tilt_params is used to find the values." (let ((target_f0_mean (car (cdr (assoc 'target_f0_mean int_tilt_params)))) (target_f0_std (car (cdr (assoc 'target_f0_std int_tilt_params)))) (model_f0_std (car (cdr (assoc 'model_f0_std int_tilt_params)))) (model_f0_mean (car (cdr (assoc 'model_f0_mean int_tilt_params))))) (if target_f0_mean ;; only if one is specified (lambda (targ) (item.set_name targ (+ target_f0_mean (* target_f0_std (/ (- (parse-number (item.name targ)) model_f0_mean) model_f0_std))))) (utt.relation.leafs utt 'Target)))) (define (tilt_assign_parameters utt) "(tilt_assign_parameters utt) Assigned tilt parameters to IntEvents, depending on the value of the Parameter tilt_method uses wagon trees (cart) or linear regression models (lr)." (let ((method (Parameter.get 'tilt_method))) (cond ((equal? method 'cart) (tilt_assign_parameters_wagon utt)) ((equal? method 'lr) (tilt_assign_parameters_lr utt)) (t (error "Tilt: unknown tilt param prediction method: " tilt_method))))) (define (tilt_assign_parameters_wagon utt) "(tilt_assign_parameters_wagon utt) Assing parameters (start_f0, tilt, amplitude, peak_pos and duration) to each IntEvent. Uses Wagon trees to predict values" (mapcar (lambda (ie) (let ((param_trees (cdr (assoc_string (item.name ie) tilt_param_trees)))) (item.set_feat ie "time_path" "IntonationSyllable") (if (string-equal "1" (item.feat ie "int_event")) (item.set_function ie "time" "unisyn_tilt_event_position") (item.set_function ie "time" "unisyn_tilt_phrase_position")) (cond ((null param_trees) (format stderr "Tilt: unknown Intonation type %s, ignored\n" (item.name ie)) ;; *need* to assign default values (item.set_feat ie "ev.f0" 100) (item.set_feat ie "tilt.amp" 20.0) (item.set_feat ie "tilt.dur" 0.25) (item.set_feat ie "tilt.tilt" -0.2) (item.set_feat ie "rel_pos" 0.0) ) (t (tilt_assign_params_wagon ie param_trees))))) (utt.relation.items utt 'Intonation))) (define (tilt_assign_params_wagon ie trees) "(tilt_assign_params_wagon ie trees) Assign the names parameters to ie using the trees and names in trees." (mapcar (lambda (tree) (let ((val (wagon_predict ie (car (cdr tree))))) (item.set_feat ie (car tree) val))) trees)) (define (tilt_assign_parameters_lr utt) "(tilt_assign_parameters_lr utt) Assing parameters (start_f0, tilt, amplitude, peak_pos and duration) to each IntEvent. Prediction by linear regression models" (mapcar (lambda (ie) (let ((param_lrmodels (cdr (assoc_string (item.name ie) tilt_param_lrmodels)))) (cond ((null param_lrmodels) (format stderr "Tilt: unknown IntEvent type %s, ignored\n" (item.name ie)) ;; *need* to assign default values (item.set_feat ie "ev.f0" 100) (item.set_feat ie "tilt.amp" 20.0) (item.set_feat ie "tilt.dur" 0.25) (item.set_feat ie "tilt.tilt" -0.2) (item.set_feat ie "rel_pos" 0.0) ) (t (tilt_assign_params_lr ie param_lrmodels))))) (utt.relation.items utt 'IntEvent))) (define (tilt_assign_params_lr ie lrmodels) "(tilt_assign_params_lr ie lrmodels) Assign the names parameters to ie using the trees and names in trees." (mapcar (lambda (lrm) (let ((val (lr_predict ie (cdr lrm)))) (item.set_feat ie (car lrm) val))) lrmodels)) (define (utt.save.tilt_events utt filename) "(utt.save.til_events UTT FILENAME) Save tilt events in UTT to FILENAME in a format suitable for ev_synth." (let ((fd (fopen filename "w"))) (format fd "#\n") (mapcar (lambda (ie) (let ((name (item.name ie))) (cond ((or (string-equal name "sil") (string-equal name "c")) (format fd " %2.4f 100 %s; tilt: %2.6f\n" (item.feat ie 'end) name (item.feat ie "tilt_start_f0"))) (t ;; accent or boundary (format fd " %2.4f 100 %s; tilt: %2.6f %2.6f %2.6f %2.6f %2.6f\n" (item.feat ie 'end) name (item.feat ie "ev.f0") (item.feat ie "tilt.amp") (item.feat ie "tilt.dur") (item.feat ie "tilt.tilt") (item.feat ie "rel_pos")))))) (utt.relation.items utt 'IntEvent)) (fclose fd) utt)) ;;;;; ;;; Some features which should be pruned ;;;;; (def_feature_docstring 'Syllable.lisp_time_to_next_vowel "Syllable.lisp_time_to_next_vowel syl The time from vowel_start to next vowel_start") (define (time_to_next_vowel syl) "(time_to_next_vowel syl) The time from vowel_start to next vowel_start" (let (ttnv) (if (string-equal "0" (item.feat syl "n.vowel_start")) (set! ttnv 0.00) (set! ttnv (- (item.feat syl "n.vowel_start") (item.feat syl "vowel_start")))) ttnv)) (def_feature_docstring 'Syllable.lisp_next_stress "Syllable.lisp_next_stress Number of syllables to next stressed syllable. 0 if this syllable is stressed. It is effectively assumed the syllable after the last syllable is stressed.") (define (next_stress syl) (cond ((null syl) 0) ((string-equal (item.feat syl 'stress_num) "1") 0) (t (+ 1 (next_stress (item.relation.next syl 'Syllable)))))) (def_feature_docstring 'Syllable.lisp_last_stress "Syllable.lisp_last_stress Number of syllables from previous stressed syllable. 0 if this syllable is stressed. It is effectively assumed that the syllable before the first syllable is stressed.") (define (last_stress syl) (cond ((null syl) 0) ((string-equal (item.feat syl 'stress_num) "1") 0) (t (+ 1 (last_stress (item.relation.prev syl 'Syllable)))))) (def_feature_docstring 'SylStructure.lisp_length_to_last_seg "SylStructure.lisp_length_to_last_seg Length from start of the vowel to start of last segment of syllable.") (define (length_to_last_seg syl) (- (item.feat syl "daughtern_to.Segment.start") (item.feat syl "vowel_start"))) (def_feature_docstring 'SylStructure.lisp_get_rhyme_length "Syllable.lisp_get_rhyme_length Length from start of the vowel to end of syllable.") (define (get_rhyme_length syl) (- (item.feat syl 'end) (item.feat syl 'vowel_start syl))) (def_feature_docstring 'SylStructure.lisp_get_onset_length "Syllable.lisp_get_onset_length Length from start of syllable to start of vowel.") (define (get_onset_length syl) (cond ((< (- (item.feat syl 'vowel_start) (item.feat syl 'start)) 0.000) 0.000) ;; just in case (t (- (item.feat syl 'vowel_start) (item.feat syl 'start))))) (def_feature_docstring 'Syllable.lisp_tilt_accent "Syllable.lisp_tilt_accent Returns \"a\" if there is a tilt accent related to this syllable, 0 otherwise.") (define (tilt_accent syl) (let ((events (item.relation.daughters syl 'IntonationSyllable)) (r "0")) (mapcar (lambda (i) (if (member_string (item.name i) tilt_accent_list) (set! r "a"))) events) r)) (def_feature_docstring 'Syllable.lisp_tilt_boundary "Syllable.lisp_tilt_boundary Returns boundary label if there is a tilt boundary related to this syllable, 0 otherwise.") (define (tilt_boundary syl) (let ((events (item.relation.daughters syl 'IntonationSyllable)) (r "0")) (mapcar (lambda (i) (let ((name (item.name i))) (if (member_string name tilt_boundary_list) (cond ((string-matches name "a.*") (set! r (string-after name "a"))) ((string-matches name "m.*") (set! r (string-after name "m"))) (t (set! r name)))))) events) r)) (def_feature_docstring 'Syllable.lisp_tilt_accented "Syllable.lisp_tilt_accented Returns 1 if there is a tilt accent related to this syllable, 0 otherwise.") (define (tilt_accented syl) (let ((events (item.relation.daughters syl 'IntonationSyllable)) (r "0")) (mapcar (lambda (i) (if (member_string (item.name i) tilt_accent_list) (set! r "1"))) events) r)) (def_feature_docstring 'Syllable.lisp_tilt_boundaried "Syllable.lisp_tilt_boundaried Returns 1 if there is a tilt boundary related to this syllable, 0 otherwise.") (define (tilt_boundaried syl) (let ((events (item.relation.daughters syl 'IntonationSyllable)) (r "0")) (mapcar (lambda (i) (if (member_string (item.name i) tilt_boundary_list) (set! r "1"))) events) r)) (def_feature_docstring 'SylStructure.lisp_vowel_height "SylStructure.lisp_vowel_height syl Classifies vowels as high, low or mid") (define (vowel_height syl) (let ((vh (item.feat syl "daughtern.daughter1.daughter1.df.height"))) vh) ) (def_feature_docstring 'SylStructure.lisp_vowel_frontness "SylStructure.vowel_frontness syl Classifies vowels as front, back or mid") (define (vowel_frontness syl) (let ((vf (item.feat syl "daughtern.daughter1.daughter1.df.front"))) vf) ) (def_feature_docstring 'SylStructure.lisp_vowel_length "SylStructure.vowel_length syl Returns the df.length feature of a syllable's vowel") (define (vowel_length syl) (let ((vl (item.feat syl "daughtern.daughter1.daughter1.df.length"))) vl) ) (defvar sonority_vless_obst '("f" "h" "hh" "k" "p" "s" "sh" "t" "th" "ch") "sonority_vless_obst List of voiceless obstruents for use in sonority scaling (only good w/ radio_speech)" ) (defvar sonority_v_obst '("v" "b" "g" "z" "zh" "d" "dh" "jh") "sonority_v_obst List of voiced obstruents for use in sonority scaling (only good w/ radio_speech)" ) (defvar sonority_nas '("m" "n" "ng" "nx" "em" "en") "sonority_nas List of nasals (only good w/ radio_speech)" ) (defvar sonority_liq '("r" "l" "er" "el" "axr") "sonority_liq List of liquids (only good w/ radio_speech)" ) (defvar sonority_glides '("y" "w") "sonority_glides List of glides (only good w/ radio_speech)" ) (def_feature_docstring 'SylStructure.lisp_sonority_scale_coda "SylStructure.sonority_scale_coda syl Returns value on sonority scale (1 -6, where 6 is most sonorous) for the coda of a syllable, based on least sonorant portion.") (define (sonority_scale_coda syl) (let ((segs (item.daughters (item.daughtern (item.daughtern syl)))) (scale 6)) (mapcar (lambda (seg) (cond ((member_string (item.name seg) sonority_vless_obst) (if (> scale 1) (set! scale 1))) ((member_string (item.name seg) sonority_v_obst) (if (> scale 2) (set! scale 2))) ((member_string (item.name seg) sonority_nas) (if (> scale 3) (set! scale 3))) ((member_string (item.name seg) sonority_liq) (if (> scale 4) (set! scale 4))) ((member_string (item.name seg) sonority_glides) (if (> scale 5) (set! scale 5))) (t (if (> scale 6) (set! scale 6))) ) ) segs) scale)) (def_feature_docstring 'SylStructure.lisp_sonority_scale_onset "SylStructure.sonority_scale_onset syl Returns value on sonority scale (1 -6, where 6 is most sonorous) for the onset of a syllable, based on least sonorant portion.") (define (sonority_scale_onset syl) (if (string-equal "Onset" (item.feat (item.daughter1 syl) "sylval")) (let ((segs (item.daughters (item.daughter1 syl))) (scale 6)) (mapcar (lambda (seg) (cond ((member_string (item.name seg) sonority_vless_obst) (if (> scale 1) (set! scale 1))) ((member_string (item.name seg) sonority_v_obst) (if (> scale 2) (set! scale 2))) ((member_string (item.name seg) sonority_nas) (if (> scale 3) (set! scale 3))) ((member_string (item.name seg) sonority_liq) (if (> scale 4) (set! scale 4))) ((member_string (item.name seg) sonority_glides) (if (> scale 5) (set! scale 5))) (t (set! scale 6)) ) ) segs) scale) 0)) (def_feature_docstring 'SylStructure.lisp_num_postvocalic_c "SylStructure.lisp_num_postvocalic_c Finds the number of postvocalic consonants in a syllable.") (define (num_postvocalic_c syl) "(num_postvocalic_c syl) Finds the number of postvocalic consonants in a syllable." (let (segs (npc 0)) (set! segs (item.daughters (item.daughtern (item.daughtern syl)))) (mapcar (lambda (seg) (set! npc (+ npc 1)) ) segs) npc)) (def_feature_docstring 'SylStructure.lisp_syl_numphones "SylStructure.lisp_syl_numphones syl Finds the number segments in a syllable.") (define (syl_numphones syl) (length (mt_segs_from_syl syl)) ) (def_feature_docstring 'Segment.lisp_pos_in_syl "Segment.lisp_pos_in_syl seg Finds the position in a syllable of a segment - returns a number.") (define (pos_in_syl seg) (let ((segments (mt_segs_from_syl (item.relation (item.parent_to (item.relation seg 'SylStructure) 'Syllable) 'SylStructure))) (seg_count 1)) (mapcar (lambda (s) (if (not (eqv? s seg)) (set! seg_count (+ 1.0 seg_count)) nil)) segs) seg_count)) (def_feature_docstring 'Intonation.lisp_peak_anchor_segment_type "Intonation.peak_anchor_segment_type ie Determines whether the segment anchor for a peak is the first consonant of a syl - C0 -, the vowel of a syl - V0 -, or segments after that - C1->X,V1->X. If the segment is in a following syl, the return value will be preceded by a 1 - e.g. 1V1") (define (peak_anchor_segment_type ie) (let ( syl peak_anchor_num numsegs peak_anchor_type) (set! peak_anchor_num (peak_segment_anchor ie)) (if (> 9 peak_anchor_num) (set! syl (item.relation (item.parent (item.relation ie "IntonationSyllable")) "SylStructure"))) (if (> 9 peak_anchor_num) (set! numsegs (item.feat syl "syl_numphones"))) (cond ((< 9 peak_anchor_num) (set! peak_anchor_type "none")) ((> 0 peak_anchor_num) (set! peak_anchor_type (string-append "-1" (get_anchor_value (item.prev syl) (+ peak_anchor_num (item.feat syl "p.syl_numphones")))))) ((< peak_anchor_num numsegs) (set! peak_anchor_type (get_anchor_value syl numsegs))) ((> peak_anchor_num numsegs) (set! peak_anchor_type (string-append "1" (get_anchor_value (item.next syl) (- peak_anchor_num numsegs))))) (set! peak_anchor_type "none")) ; (format stderr "pat: %s\n" peak_anchor_type) peak_anchor_type)) (define (get_anchor_value sylSyl seg_num) "(get_anchor_value sylSyl seg_num) Gets the c/v value of the segment within a syllable." (let ((syl (item.relation sylSyl "SylStructure")) (seg_val "none") segs (ccnt -1) (vcnt -1) (vpis 0)) (set! segs (mt_segs_from_syl sylSyl)) (mapcar (lambda (seg) (cond ((string-equal "consonant" (item.feat seg "df.type")) (set! vcnt (+ 1 vcnt)) (set! vpis (item.feat seg "pos_in_syl"))) (t (set! ccnt (+ 1 ccnt)))) (cond ((and (eq (- seg_num 1.0) (item.feat seg "pos_in_syl")) ( string-equal "consonant" (item.feat seg "df.type"))) (set! seg_val (string-append "V" vcnt))) ((and (eq (- seg_num 1.0) (item.feat seg "pos_in_syl")) ( string-equal "vowel" (item.feat seg "df.type"))) (set! seg_val (string-append "C" (- (item.feat seg "pos_in_syl") vpis) "V" vcnt))) (t nil)) ) segs) seg_val)) (define (peak_segment_anchor ie) "peak_segment_anchor ie Determines what segment acts as the anchor for a peak. Returns number of segments from start of accented syllable to peak." ; (format stderr "accent: %s\n" ; (item.name ie)) (let ((pk_pos (item.feat ie "position")) (peak_seg_anchor 11)) (if (or (string-equal "phrase_start" (item.name ie)) (string-equal "phrase_end" (item.name ie)) (string-equal "pause" (item.name ie))) (set! peak_seg_anchor 10) (set! peak_seg_anchor (find_peak_seg_anchor ie pk_pos))) peak_seg_anchor)) (define (find_peak_seg_anchor ie pk_pos) "find_peak_seg_anchor ie pk_pos Part of the workings of peak_segment_anchor." (let (( syl (item.relation (item.parent (item.relation ie 'IntonationSyllable)) 'SylStructure)) (seg_anchor 11)) (cond ((not (eq 9.0 (segs_to_peak syl pk_pos))) (set! seg_anchor (segs_to_peak syl pk_pos))) ((and (item.prev syl) (not (eq 9.0 (segs_to_peak (item.prev syl) pk_pos)))) ; (format stderr "%s\n" (item.name (item.prev syl))) (set! seg_anchor (* -1 (- (+ 1 (item.feat syl "p.syl_numphones")) (segs_to_peak (item.prev syl) pk_pos))))) ((and (item.next syl) (> pk_pos (item.feat syl "n.start"))) ; (format stderr "%s\n" (item.name (item.next syl))) (set! seg_anchor (+ 1 (item.feat syl "syl_numphones") (segs_to_peak (item.next syl) pk_pos)))) (t (format stderr "No seg anchor could be found\n"))) ; (format stderr "seg_anchor: %f\n" seg_anchor) seg_anchor)) (define (segs_to_peak sylSyl pk_pos) "(segs_to_peak sylSyl pk_pos) Determines the number of segments from the start of a syllable to an intonation peak" (let ((syl (item.relation sylSyl "SylStructure")) (segs_2_peak 9) segs) (set! segs (mt_segs_from_syl syl)) (mapcar (lambda (seg) ; (format stderr "seg_end: %f pk: %f\n" (item.feat seg "end") ; pk_pos) (if (eq 1.0 (peak_wi_seg seg pk_pos)) (set! segs_2_peak (item.feat seg "pos_in_syl"))) ; (format stderr "segs_2_peak: %f\n" segs_2_peak) ) segs) segs_2_peak)) (define (peak_wi_seg segment pk_pos) "peak_wi_seg segment pk_pos Finds if a peak occurs w/i a segment" (let ((s_start (item.feat segment "start")) (s_end (item.feat segment "end")) (ret 0.0)) (if (and (< s_start pk_pos) (< pk_pos s_end)) (set! ret 1.0) nil) ret)) (defvar tilt_accent_list '("a" "arb" "afb" "m" "mfb" "mrb") "tilt_accent_list List of events containing accents in tilt model.") (defvar tilt_boundary_list '("rb" "arb" "afb" "fb" "mfb" "mrb") "tilt_boundary_list List of events containing boundaries in tilt model.") (def_feature_docstring 'Intonation.lisp_last_tilt_accent "Intonation.lisp_last_tilt_accent Returns the most recent tilt accent.") (define (last_tilt_accent intev) (let ((pie (item.relation.prev intev 'Intonation))) (cond ((not pie) "0") ((member_string (item.name pie) tilt_accent_list) (item.name pie)) (t (last_tilt_accent pie))))) (def_feature_docstring 'Intonation.lisp_next_tilt_accent "Intonation.lisp_next_tilt_accent Returns the next tilt accent.") (define (next_tilt_accent intev) (let ((nie (item.relation.next intev 'Intonation))) (cond ((not nie) "0") ((member_string (item.name nie) tilt_accent_list) (item.name nie)) (t (next_tilt_accent nie))))) (def_feature_docstring 'Intonation.lisp_last_tilt_boundary "Intonation.lisp_last_tilt_boundary Returns the most recent tilt boundary.") (define (last_tilt_boundary intev) (let ((pie (item.relation.prev intev 'Intonation))) (cond ((not pie) "0") ((member_string (item.name pie) tilt_boundary_list) (item.name pie)) (t (last_tilt_boundary pie))))) (def_feature_docstring 'Intonation.lisp_next_tilt_boundary "Intonation.lisp_next_tilt_boundary Returns the next tilt boundary.") (define (next_tilt_boundary intev) (let ((nie (item.relation.next intev 'Intonation))) (cond ((not nie) "0") ((member_string (item.name nie) tilt_boundary_list) (item.name nie)) (t (next_tilt_boundary nie))))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Some basic function to metrical tree structure ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (mt_syl_from_seg seg) (if seg (item.root (item.relation seg 'SylStructure)) nil)) (define (mt_word_from_syl syl) (if syl (item.root (item.relation syl 'WordStructure)) nil)) (define (mt_word_from_seg seg) (mt_word_from_syl (mt_syl_from_seg seg))) (define (mt_segs_from_syl s) (cond ((null s) nil) ((member_string 'Segment (item.relations s)) (list s)) (t (apply append (mapcar mt_segs_from_syl (item.relation.daughters s 'SylStructure)))))) (define (sylmtval s) (let ((syl (mt_syl_from_seg s))) (if syl (item.feat syl "MetricalValue") "0"))) (define (sylpmtval s) (let ((syl (mt_syl_from_seg s))) (if syl (item.feat syl "R:MetricalTree.parent.MetricalValue") "0"))) (define (mt_numsyls w) (let ((s1 (item.daughter1_to (item.relation w 'WordStructure) 'Syllable)) (sn (item.daughtern_to (item.relation w 'WordStructure) 'Syllable)) (count 1)) (while (and s1 (not (equal? s1 sn))) (set! count (+ 1 count)) (set! s1 (item.next s1))) (if s1 count 0))) (define (mt_seg_numsyls s) (let ((w (mt_word_from_seg s))) (if w (mt_num_syls w) 0))) ;;; These functions should be sort out some time ;;; Difference between this syl and the next ;;; number of closing brackets, number of opening brackets ;;; difference between them (define (mt_close n) "(mt_close n) The number of consituents this is the end of, Effectively the number of closing brackets after this word." (if (or (not n) (item.next n)) 0 (+ 1 (mt_close (item.parent n))))) (define (mt_open n) "(mt_open n) The number of consituents this is the start of, Effectively the number of opening brackets before this word." (if (or (not n) (item.prev n)) 0 (+ 1 (mt_open (item.parent n))))) (define (mt_postype syl) "(mt_postype syl) Returns single, initial, final or middle." (let ((w (mt_word_from_syl syl)) (psw (mt_word_from_syl (item.relation.prev syl 'Syllable))) (nsw (mt_word_from_syl (item.relation.next syl 'Syllable)))) (cond ((and (equal? w psw) (equal? w nsw)) 'middle) ((and (not (equal? w psw)) (not (equal? w nsw))) 'single) ((equal? w psw) 'final) (t 'initial)))) (define (mt_accent syl) "(mt_accent syl) Accent or 0 if none." (let ((a 0)) (mapcar (lambda (i) (if (string-matches (item.name i) "^a.*") (set! a "a"))) (item.relation.daughters syl 'IntonationSyllable)) a)) (define (mt_break syl) "(mt_break syl) Break or 0 if none." (let ((a 0)) (mapcar (lambda (i) (if (string-matches (item.name i) ".*b$") (set! a (item.name i)))) (item.relation.daughters syl 'IntonationSyllable)) a)) (define (mt_ssyl_out s) (cond ((null s) 0) ((not (string-equal "0" (item.feat s "R:WordStructure.root.lisp_word_mt_break"))) 0) ((string-equal "s" (item.feat s "MetricalValue")) (+ 1 (mt_ssyl_out (item.relation.next s 'Syllable)))) (t (mt_ssyl_out (item.relation.next s 'Syllable))))) (define (mt_num_s s) "(mt_num_s s) The number of s MetricalValues from here to a w or top." (cond ((null s) 0) ((string-equal "w" (item.feat s "MetricalValue")) 0) (t (+ 1 (mt_num_s (item.parent s)))))) (define (mt_num_w s) "(mt_num_w s) The number of w MetricalValues from here to a s or top." (cond ((null s) 0) ((string-equal "s" (item.feat s "MetricalValue")) 0) (t (+ 1 (mt_num_w (item.parent s)))))) (define (mt_strong s) "(mt_strong s) 1 if all MetricalValues a s to a word, 0 otherwise." (cond ((string-equal "w" (item.feat s "MetricalValue")) "0") ((member_string 'Word (item.relations s)) "1") (t (mt_strong (item.relation.parent s 'MetricalTree))))) (define (mt_lssp s) "(mt_lssp s) 1 if last stressed syllable in phrase, 0 otherwise." (if (and (string-equal "s" (item.feat s "MetricalValue")) (equal? 0 (mt_ssyl_out s))) "1" "0")) (define (mt_fssw s) "(mt_fssw s) 1 if first stressed syllable in word, 0 otherwise." (if (and (string-equal "s" (item.feat s "MetricalValue")) (mt_no_stress_before (item.relation.prev s 'Syllable))) "1" "0")) (define (mt_nfssw s) "(nfssw s) 1 if second or later stressed syllable in word, 0 otherwise." (if (and (string-equal "s" (item.feat s "MetricalValue")) (null (mt_no_stress_before (item.relation.prev s 'Syllable)))) "1" "0")) (define (mt_no_stress_before ss) (cond ((null ss) t) ((not (string-equal (item.feat ss "R:WordStructure.root.addr") (item.feat (item.next ss) "R:WordStructure.root.addr"))) t) ((string-equal "s" (item.feat ss "MetricalValue")) nil) (t (mt_no_stress_before (item.prev ss))))) (define (word_mt_break w) (cond ((string-equal "1" (item.feat w "sentence_end")) "BB") ((string-equal "1" (item.feat w "phrase_end")) "B") (t "0"))) (provide 'tilt) festival/lib/apml.scm0000644000175000017500000004217012442333465013214 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 2002 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Rob Clark ;;; Date: July 2002 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; Sets up the current voice to synthesise from APML. ;; ;; (require 'apml_f2bf0lr) (require 'apml_kaldurtreeZ) ;; Default pitch settings (if unspecified in current voice.) (defvar apml_default_pitch_mean 170 ) (defvar apml_default_pitch_standard_deviation 34 ) ;; apml sythesis wrappers. (define (apml_client_synth apml) "(apml_client_synth apml) Synthesise apml and return waveform(s) to client." (utt.send.wave.client (apml_synth apml))) (define (apml_synth apml) "(apml_synth xml) Synthesis an apml string." (let ((tmpfile (make_tmp_filename)) utt) (string_to_file tmpfile apml) (set! utt (apml_file_synth tmpfile)) (delete-file tmpfile) utt)) (define (apml_file_synth filename) "(apml_file_synth filename) Synthesis an apml file." (let ((utt (Utterance Tokens nil))) (utt.load utt filename) (utt.synth utt))) (define (string_to_file file s) "(string_to_file file string) Write string to file." (let ((fd)) (set! fd (fopen file "wb")) (format fd "%s" s) (fclose fd))) ;;; ;;; Phrasing. ;;; ;; phrasing CART. ; ; It has been decided that by default, only punctuation should affect ; phrasing (and subsequently pauses) ; (set! apml_phrase_tree ' ((lisp_apml_punc in ("?" "." ":")) ; big punctuation ((BB)) ((lisp_apml_punc in ("'" "\"" "," ";")) ; else little punctuation ((B)) ((lisp_apml_last_word is 1) ((BB)) ; need a BB at the end! ((NB)))))) ; else nothing ;; feature functions for phrasing (define (apml_punc word) (item.feat (item.relation.parent word 'Token) 'punc)) (define (apml_last_word word) (if (item.next word) "0" "1")) ;;; ;;; Pauses ;;; ;; feature functions for pauses (define (apml_is_pause word) (if (item.relation (item.relation.parent word 'Token) 'Pause) t nil)) (define (apml_pause word) (if (item.relation word 'Pause) (item.feat (item.relation.parent (item.relation.parent word 'Token) 'Pause) "sec") 0)) (define (Apml_Pauses utt) "(Pauses UTT) Predict pause insertion for apml." (let ((words (utt.relation.items utt 'Word)) lastword tpname) (if words (begin (insert_initial_pause utt) ;; always have a start pause (set! lastword (car (last words))) (mapcar (lambda (w) (let ((pbreak (item.feat w "pbreak")) (emph (item.feat w "R:Token.parent.EMPH"))) (cond ((apml_is_pause w) (insert_pause utt w)) ((or (string-equal "B" pbreak) (string-equal "BB" pbreak)) (insert_pause utt w)) ((equal? w lastword) (insert_pause utt w))))) words) ;; The embarassing bit. Remove any words labelled as punc or fpunc (mapcar (lambda (w) (let ((pos (item.feat w "pos"))) (if (or (string-equal "punc" pos) (string-equal "fpunc" pos)) (let ((pbreak (item.feat w "pbreak")) (wp (item.relation w 'Phrase))) (if (and (string-matches pbreak "BB?") (item.relation.prev w 'Word)) (item.set_feat (item.relation.prev w 'Word) "pbreak" pbreak)) (item.relation.remove w 'Word) ;; can't refer to w as we've just deleted it (item.relation.remove wp 'Phrase))))) words))) utt)) ;;; ;;; Intonation. ;;; ;; Accent prediction (well transfer really). ;; ;; We treat L+H* L-H% on a single syllable as a special case. (set! apml_accent_cart ' ((lisp_apml_accent is "Hstar") ((H*)) ((lisp_apml_accent is "Lstar") ((L*)) ((lisp_apml_LHLH is "LHLH") ((L+H*L-H%)) ((lisp_apml_accent is "LplusHstar") ((L+H*)) ((lisp_apml_accent is "LstarplusH") ((L*+H)) ((NONE)))))))) (set! apml_boundary_cart ' ((lisp_apml_boundary is "LL") ((L-L%)) ((lisp_apml_LHLH is "LHLH") ((NONE)) ; this is dealt with by the accent feature ((lisp_apml_boundary is "LH") ((L-H%)) ((lisp_apml_boundary is "HH") ((H-H%)) ((lisp_apml_boundary is "HL") ((H-L%)) ((NONE)))))))) ;; feature functions. (define (apml_accent syl) (let ((token (item.relation.parent (item.relation.parent syl 'SylStructure) 'Token))) (if (and (eq (item.feat syl 'stress) 1) (item.relation.parent token 'Emphasis)) (item.feat (item.relation.parent token 'Emphasis) 'x-pitchaccent) 0))) (define (apml_boundary syl) (let ((token (item.relation.parent (item.relation.parent syl 'SylStructure) 'Token))) (if (and (> (item.feat syl 'syl_break) 0) (item.relation.parent token 'Boundary)) (item.feat (item.relation.parent token 'Boundary) 'type) 0))) (define (apml_LHLH syl) (let ((accent (apml_accent syl)) (boundary (apml_boundary syl))) (if (and (string-equal accent "LplusHstar") (string-equal boundary "LH")) "LHLH" 0))) (define (apml_seg_is_LHLH_vowel seg) (if (and (string-equal (apml_LHLH (item.relation.parent seg 'SylStructure)) "LHLH") (string-equal (item.feat seg 'ph_vc) "+")) "LHLH" 0)) ;;;; feature functions: (define (apml_tgtype syl) (let ((l (apml_boundl (item.relation.parent syl 'SylStructure))) (r (apml_boundr (item.relation.parent syl 'SylStructure)))) (if (eq (item.feat syl 'accented) 0) 0 ; this is a quirk related to the way the models were trained (cond ((eq l 0) 1) ((eq r 1) 3) (t 2))))) (define (apml_iecount syl) (if (eq (item.feat syl 'accented) 0) 0 ; this is a quirk related to the way the models were trained (+ (item.feat syl 'asyl_in) 1))) ;; suport functions. (define (apml_boundl word) "(apml_boundl word) Number of boundaries in this performative to the left of this word." (let ((w (item.prev word)) (c 0)) (while (and w (apml_same_p w word)) (if (item.relation.parent (item.relation.parent w 'Token) 'Boundary) (set! c (+ c 1))) (set! w (item.prev w))) c)) (define (apml_boundr word) "(apml_boundr word) Number of boundaries in this performative to the right of this word." (let ((w word) (c 0)) (while (and w (apml_same_p w word)) (if (item.relation.parent (item.relation.parent w 'Token) 'Boundary) (set! c (+ c 1))) (set! w (item.next w))) c)) (define (apml_same_p w1 w2) "(apml_same_p w1 w2) Are these two words in the same performative?" (let ((p1 (item.relation.parent (item.relation.parent w1 'Token) 'SemStructure)) (p2 (item.relation.parent (item.relation.parent w1 'Token) 'SemStructure))) (if (and (item.parent p1) (item.parent p2)) ; not true if theme/rheme omitted. (equal? (item.parent p1) (item.parent p2)) (equal? p1 p2)))) ;;; ;;; segment timings ;;; (define (apml_seg_times utt) "(apml_seg_times utt) Output the segment timings for an apml utterance." (let ((segs (utt.relation.items utt 'Segment))) (mapcar (lambda (x) (format t "%s %s\n" (item.name x) (item.feat x 'end))) segs) t)) ;;; ;;; Additional functions for f0model. ;;; (define (find_hstar_left syl) "(find_hstar_left syl) If the closest accent or boundary to the left is H* return how many syllables away it is. Returns 0 if nearest accent is not H*" (let ((count 0)) ;; if this syllable has a pitch event (if (or (not (string-equal (item.feat syl 'tobi_accent) "NONE")) (not (string-equal (item.feat syl 'tobi_endtone) "NONE"))) 0) (while (and syl (string-equal (item.feat syl 'tobi_accent) "NONE") (string-equal (item.feat syl 'tobi_endtone) "NONE")) (set! count (+ count 1)) (set! syl (item.prev syl))) (cond ;; run out of syllables before finding accent ((null syl) 0) ((string-equal (item.feat syl 'tobi_accent) "H*") count) (t 0)))) (define (find_ll_right syl) "(find_ll_right syl) If the closest accent or boundary to the right is L-L% return how many syllables away it is. Returns 0 if nearest is not L-L%." (let ((count 0)) ;; if this syllable has a pitch event (if (or (not (string-equal (item.feat syl 'tobi_accent) "NONE")) (not (string-equal (item.feat syl 'tobi_endtone) "NONE"))) 0) (while (and syl (string-equal (item.feat syl 'tobi_accent) "NONE") (string-equal (item.feat syl 'tobi_endtone) "NONE")) (set! count (+ count 1)) (set! syl (item.next syl))) (cond ;; run out of syllables before finding boundary ((null syl) 0) ((string-equal (item.feat syl 'tobi_endtone) "L-L%") count) (t 0)))) (define (l_spread syl) "(l_spread syl) Proportion of pitch lowering required due to L- spreading backwards." (let ((l (find_hstar_left syl)) (r (find_ll_right syl))) (cond ((or (eq l 0) (eq r 0)) 0) (t (/ r (- (+ l r) 1)))))) ;;; ;;; Debuging and other useful stuff. ;;; (define (apml_print_semstruct utt) "(apml_print_semstruct utt) Pretty print APML semantic structure." (let ((i (utt.relation.first utt 'SemStructure))) (while (not (null i)) (apml_pss_item 0 i) (apml_pss_daughters 1 (item.daughters i)) (set! i (item.next i))))) (define (apml_pss_daughters depth list) (mapcar (lambda (x) (apml_pss_item depth x) (apml_pss_daughters (+ depth 1) (item.daughters x)) ) list)) (define (apml_pss_item depth item) (let ((c 0)) (while (< c depth) (format t " ") (set! c (+ c 1))) (format t "%s\n" (item.name item)))) (define (apml_print_words utt) "(apml_print_words utt) Pretty print APML words with associated accents." (mapcar (lambda (x) (format t "%s (" (item.name x)) (apml_pww_accent x) (apml_pww_boundary x) (apml_pww_pause x) (format t ")\n")) (utt.relation.items utt 'Word)) t) (define (apml_pww_accent item) (let ((p (item.relation.parent (item.relation.parent item 'Token) 'Emphasis))) (if p (apml_ppw_list (item.features p))))) (define (apml_pww_boundary item) (let ((p (item.relation.parent (item.relation.parent item 'Token) 'Boundary))) (if p (apml_ppw_list (item.features p))))) (define (apml_pww_pause item) (let ((p (item.relation.parent (item.relation.parent item 'Token) 'Pause))) (if p (apml_ppw_list (item.features p))))) (define (apml_ppw_list l) (mapcar (lambda (x) (format t " %s" x)) (flatten l))) (define (apml_print_sylstructure utt filename) "(apml_print_sylstructure utt filename) Pretty print APML syllable structure. Filename t for stdout" (let (fd) (if (not (eq? filename t)) (set! fd (fopen filename "wb")) (set! fd t)) (mapcar (lambda (x) (format fd "%s\n" (item.name x)) (apml_psyl fd x)) (utt.relation.items utt 'Word)) t)) (define (apml_psyl fd word) (mapcar (lambda (x) (apml_psegs fd x) (if (eq (item.feat x 'stress) 1) (format fd " (1)")) (if (item.relation.daughter1 x 'Intonation) (begin (let ((ie (item.relation.daughter1 x 'Intonation))) (format fd " [") (while ie (format fd "%s" (item.name ie)) (set! ie (item.next ie)) (if ie (format t " "))) (format fd "]")))) (format fd "\n")) (item.daughters (item.relation word 'SylStructure)))) (define (apml_psegs fd syl) (let ((segs (item.daughters syl))) (format fd " ") (while segs (format fd "%s" (item.name (car segs))) (if (cdr segs) (format fd ".")) (set! segs (cdr segs))))) (define (apml_get_lr_params) (let ((m 0) (s 0)) (if (or (equal? (Parameter.get 'Int_Target_Method) Int_Targets_LR) (equal? (Parameter.get 'Int_Target_Method) Int_Targets_5_LR)) (begin (set! m (car (cdr (car int_lr_params)))) (set! s (car (cdr (car (cdr int_lr_params)))))) (begin (set! m apml_default_pitch_mean) (set! s apml_default_pitch_standard_deviation))) (list m s))) (define (apml_initialise) "(apml_initialise) Set up the current voice for apml use." (if (not (string-matches current-voice ".*multisyn.*")) ; nothing if multisyn (cond ((or (string-equal (Parameter.get 'Language) "americanenglish") (string-equal (Parameter.get 'Language) "britishenglish")) (begin (format t "Initialising APML for English.\n") ;; Phrasing. (Parameter.set 'Phrase_Method 'cart_tree) (set! phrase_cart_tree apml_phrase_tree) ;; Pauses. ;;(set! duration_cart_tree apml_kal_duration_cart_tree) ;;(set! duration_ph_info apml_kal_durs) ;;(Parameter.set 'Pause_Method Apml_Pauses) ;; Lexicon. ;;;; We now assume the lexicon you have already set is suitable, ;;;; You probably want to ensure this is "apmlcmu" or "unilex" ;;(if (not (member_string "apmlcmu" (lex.list))) ;; (load (path-append lexdir "apmlcmu/apmlcmulex.scm"))) ;;(lex.select "apmlcmu") ;; Add other lex entries here: ;;(lex.add.entry '("minerals" nil (((m ih n) 1) ((er) 0) ((ax l z) 0)))) ;;(lex.add.entry '("fibre" nil (((f ay b) 1) ((er) 0)))) ;;(lex.add.entry '("dont" v (((d ow n t) 1)))) ;;(lex.add.entry '("pectoris" nil (((p eh k) 2) ((t ao r) 1) ((ih s) 0)))) ;;(lex.add.entry '("sideeffects" nil (((s ay d) 1) ((ax f) 0) ((eh k t s) 2)))) ;; Intonation events. (set! int_accent_cart_tree apml_accent_cart) (set! int_tone_cart_tree apml_boundary_cart) (Parameter.set 'Int_Method Intonation_Tree) ;; Intonation f0 contour. (set! f0_lr_start apml_f2b_f0_lr_start) (set! f0_lr_left apml_f2b_f0_lr_left) (set! f0_lr_mid apml_f2b_f0_lr_mid) (set! f0_lr_right apml_f2b_f0_lr_right) (set! f0_lr_end apml_f2b_f0_lr_end) (set! int_lr_params (list (list 'target_f0_mean (car (apml_get_lr_params))) (list 'target_f0_std (car (cdr (apml_get_lr_params)))) (list 'model_f0_mean 170) (list 'model_f0_std 40))) (Parameter.set 'Int_Target_Method Int_Targets_5_LR) nil)) ((string-equal (Parameter.get 'Language) "italian") (begin (format t "Initialising APML for Italian.\n") ;; Phrasing. (Parameter.set 'Phrase_Method 'cart_tree) (set! phrase_cart_tree apml_phrase_tree) ;; Intonation events. (set! int_accent_cart_tree apml_accent_cart) (set! int_tone_cart_tree apml_boundary_cart) (Parameter.set 'Int_Method Intonation_Tree) ;; Intonation f0 contour. (set! f0_lr_start apml_f2b_f0_lr_start) (set! f0_lr_mid apml_f2b_f0_lr_mid) (set! f0_lr_end apml_f2b_f0_lr_end) (set! int_lr_params (list (list 'target_f0_mean (car (apml_get_lr_params))) (list 'target_f0_std (car (cdr (apml_get_lr_params)))) (list 'model_f0_mean 170) (list 'model_f0_std 34))) (Parameter.set 'Int_Target_Method Int_Targets_LR) nil)) (t nil)))) (provide 'apml) festival/lib/apml_f2bf0lr.scm0000644000175000017500000006266510065567162014546 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 2002 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Rob Clark ;;; Date: July 2002 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; APML.f0 trees. ;; ;; (set! apml_f2b_f0_lr_start '( ( Intercept 163.9871 ) ( pp.lisp_apml_tgtype -3.1750 (1) ) ( p.lisp_apml_tgtype 5.0332 (1) ) ( lisp_apml_tgtype 0.0000 (1) ) ( n.lisp_apml_tgtype 17.7799 (1) ) ( nn.lisp_apml_tgtype 13.6845 (1) ) ( pp.lisp_apml_tgtype 0.0000 (2) ) ( p.lisp_apml_tgtype 0.0000 (2) ) ( lisp_apml_tgtype 0.0000 (2) ) ( n.lisp_apml_tgtype 0.0000 (2) ) ( nn.lisp_apml_tgtype 0.0000 (2) ) ( pp.lisp_apml_tgtype 0.0000 (3) ) ( p.lisp_apml_tgtype 0.0000 (3) ) ( lisp_apml_tgtype -9.7245 (3) ) ( n.lisp_apml_tgtype 0.0000 (3) ) ( nn.lisp_apml_tgtype -2.4009 (3) ) ( pp.lisp_apml_iecount 0.0000 ) ( p.lisp_apml_iecount -0.4484 ) ( lisp_apml_iecount 0.0000 ) ( n.lisp_apml_iecount -2.0165 ) ( nn.lisp_apml_iecount 0.0000 ) ( pp.tobi_accent 0.0000 (H*) ) ( p.tobi_accent 11.1239 (H*) ) ( tobi_accent 21.5164 (H*) ) ( n.tobi_accent -2.5990 (H*) ) ( nn.tobi_accent -6.5307 (H*) ) ( pp.tobi_accent 0.0000 (L*) ) ( p.tobi_accent -10.0000 (L*) ) ( tobi_accent -5.0000 (L*) ) ( n.tobi_accent -10.6798 (L*) ) ( nn.tobi_accent -5.6561 (L*) ) ( pp.tobi_accent 5.3577 (L*+H) ) ( p.tobi_accent 60.0000 (L*+H) ) ( tobi_accent -5.0000 (L*+H) ) ( n.tobi_accent 0.0000 (L*+H) ) ( nn.tobi_accent 0.0000 (L*+H) ) ( pp.tobi_accent 0.0000 (L+H*) ) ( p.tobi_accent 11.1200 (L+H*) ) ( tobi_accent 21.5200 (L+H*) ) ( n.tobi_accent -2.6000 (L+H*) ) ( nn.tobi_accent -6.5300 (L+H*) ) ( pp.tobi_endtone 0.0000 (L-L%) ) ( p.tobi_endtone -0.6164 (L-L%) ) ( tobi_endtone -50 (L-L%) ) ( n.tobi_endtone -10.8729 (L-L%) ) ( nn.tobi_endtone -7.6522 (L-L%) ) ( pp.tobi_endtone 0.7583 (L-H%) ) ( p.tobi_endtone 0.0000 (L-H%) ) ( tobi_endtone -20.0000 (L-H%) ) ( n.tobi_endtone -11.8935 (L-H%) ) ( nn.tobi_endtone -7.2012 (L-H%) ) ( pp.tobi_endtone 0.0000 (H-L%) ) ( p.tobi_endtone 0.0000 (H-L%) ) ( tobi_endtone 4.0790 (H-L%) ) ( n.tobi_endtone -19.3463 (H-L%) ) ( nn.tobi_endtone -29.3615 (H-L%) ) ( pp.tobi_endtone 0.0000 (H-H%) ) ( p.tobi_endtone 0.0000 (H-H%) ) ( tobi_endtone 0.0000 (H-H%) ) ( n.tobi_endtone 0.0000 (H-H%) ) ( nn.tobi_endtone 0.0000 (H-H%) ) ( pp.tobi_endtone 0.0000 (L-) ) ( p.tobi_endtone -15.1702 (L-) ) ( tobi_endtone 0.0000 (L-) ) ( n.tobi_endtone -14.5562 (L-) ) ( nn.tobi_endtone 0.0000 (L-) ) ( pp.tobi_endtone -13.5046 (H-) ) ( p.tobi_endtone 0.0000 (H-) ) ( tobi_endtone 6.3377 (H-) ) ( n.tobi_endtone -6.8631 (H-) ) ( nn.tobi_endtone 0.0000 (H-) ) ( p.tobi_accent 60.0000 (L+H*L-H%) ) ( tobi_accent -60.0000 (L+H*L-H%) ) ( n.tobi_accent 0.0000 (L+H*L-H%) ) ( pp.syl_break 0.0000 ) ( p.syl_break 0.0000 ) ( syl_break 0.6417 ) ( n.syl_break 1.3532 ) ( nn.syl_break 1.0724 ) ( pp.stress 0.0000 ) ( p.stress -0.6193 ) ( stress 2.4121 ) ( n.stress 0.0000 ) ( nn.stress 2.5478 ) ( syl_in -1.4373 ) ( syl_out 0.4181 ) ( ssyl_in 0.0000 ) ( ssyl_out 0.6125 ) ( asyl_in 0.0000 ) ( asyl_out 0.9906 ) ( last_accent 0.0000 ) ( next_accent -0.3700 ) ( sub_phrases 0.0000 ) ( lisp_l_spread -60.0000 ) )) (set! apml_f2b_f0_lr_left '( ( Intercept 162.1173 ) ( pp.lisp_apml_tgtype -1.5875 (1) ) ( p.lisp_apml_tgtype 4.8101 (1) ) ( lisp_apml_tgtype 12.8265 (1) ) ( n.lisp_apml_tgtype 16.3027 (1) ) ( nn.lisp_apml_tgtype 13.3225 (1) ) ( pp.lisp_apml_tgtype 0.0000 (2) ) ( p.lisp_apml_tgtype 1.7434 (2) ) ( lisp_apml_tgtype 6.7783 (2) ) ( n.lisp_apml_tgtype 0.6679 (2) ) ( nn.lisp_apml_tgtype 0.0000 (2) ) ( pp.lisp_apml_tgtype 1.6494 (3) ) ( p.lisp_apml_tgtype 1.2861 (3) ) ( lisp_apml_tgtype -2.0724 (3) ) ( n.lisp_apml_tgtype 0.0000 (3) ) ( nn.lisp_apml_tgtype -1.2004 (3) ) ( pp.lisp_apml_iecount 0.0000 ) ( p.lisp_apml_iecount -0.5857 ) ( lisp_apml_iecount 0.0000 ) ( n.lisp_apml_iecount -2.3543 ) ( nn.lisp_apml_iecount 0.0000 ) ( pp.tobi_accent 0.0000 (H*) ) ( p.tobi_accent 8.5867 (H*) ) ( tobi_accent 21.2169 (H*) ) ( n.tobi_accent -1.2995 (H*) ) ( nn.tobi_accent -6.5056 (H*) ) ( pp.tobi_accent 0.0000 (L*) ) ( p.tobi_accent -7.5000 (L*) ) ( tobi_accent -25.0000 (L*) ) ( n.tobi_accent -8.3939 (L*) ) ( nn.tobi_accent -4.5688 (L*) ) ( pp.tobi_accent 2.6789 (L*+H) ) ( p.tobi_accent 45.0000 (L*+H) ) ( tobi_accent -17.5000 (L*+H) ) ( n.tobi_accent -1.3600 (L*+H) ) ( nn.tobi_accent 0.0000 (L*+H) ) ( pp.tobi_accent 0.0000 (L+H*) ) ( p.tobi_accent 8.5850 (L+H*) ) ( tobi_accent 21.2200 (L+H*) ) ( n.tobi_accent -1.3000 (L+H*) ) ( nn.tobi_accent -6.5050 (L+H*) ) ( pp.tobi_endtone 1.8117 (L-L%) ) ( p.tobi_endtone -0.1681 (L-L%) ) ( tobi_endtone -70 (L-L%) ) ( n.tobi_endtone -8.9334 (L-L%) ) ( nn.tobi_endtone -8.4034 (L-L%) ) ( pp.tobi_endtone 1.2099 (L-H%) ) ( p.tobi_endtone 1.1220 (L-H%) ) ( tobi_endtone -10.0000 (L-H%) ) ( n.tobi_endtone -5.9467 (L-H%) ) ( nn.tobi_endtone -6.9072 (L-H%) ) ( pp.tobi_endtone 0.0000 (H-L%) ) ( p.tobi_endtone 0.0000 (H-L%) ) ( tobi_endtone 2.0395 (H-L%) ) ( n.tobi_endtone -12.3940 (H-L%) ) ( nn.tobi_endtone -24.2593 (H-L%) ) ( pp.tobi_endtone 0.0000 (H-H%) ) ( p.tobi_endtone 0.0000 (H-H%) ) ( tobi_endtone 0.0000 (H-H%) ) ( n.tobi_endtone 0.0000 (H-H%) ) ( nn.tobi_endtone 16.1076 (H-H%) ) ( pp.tobi_endtone -1.8913 (L-) ) ( p.tobi_endtone -15.5650 (L-) ) ( tobi_endtone -18.3620 (L-) ) ( n.tobi_endtone -9.8322 (L-) ) ( nn.tobi_endtone -1.8182 (L-) ) ( pp.tobi_endtone -13.4429 (H-) ) ( p.tobi_endtone 0.0000 (H-) ) ( tobi_endtone 1.9053 (H-) ) ( n.tobi_endtone -3.4315 (H-) ) ( nn.tobi_endtone 0.0000 (H-) ) ( p.tobi_accent 0.0000 (L+H*L-H%) ) ( tobi_accent 10.0000 (L+H*L-H%) ) ( n.tobi_accent 0.0000 (L+H*L-H%) ) ( pp.syl_break 0.3501 ) ( p.syl_break -0.8121 ) ( syl_break 0.3209 ) ( n.syl_break 0.7486 ) ( nn.syl_break 0.8182 ) ( pp.stress -0.9778 ) ( p.stress -0.3096 ) ( stress 2.7752 ) ( n.stress 0.9976 ) ( nn.stress 2.7343 ) ( syl_in -1.9845 ) ( syl_out 0.7142 ) ( ssyl_in 1.0376 ) ( ssyl_out 0.3062 ) ( asyl_in 0.0000 ) ( asyl_out 0.4953 ) ( last_accent 0.0000 ) ( next_accent 0.1084 ) ( sub_phrases 0.0000 ) ( lisp_l_spread -60.0000 ) )) (set! apml_f2b_f0_lr_mid '( ( Intercept 160.2474 ) ( pp.lisp_apml_tgtype 0.0000 (1) ) ( p.lisp_apml_tgtype 4.5869 (1) ) ( lisp_apml_tgtype 25.6530 (1) ) ( n.lisp_apml_tgtype 14.8255 (1) ) ( nn.lisp_apml_tgtype 12.9605 (1) ) ( pp.lisp_apml_tgtype 0.0000 (2) ) ( p.lisp_apml_tgtype 3.4867 (2) ) ( lisp_apml_tgtype 13.5566 (2) ) ( n.lisp_apml_tgtype 1.3359 (2) ) ( nn.lisp_apml_tgtype 0.0000 (2) ) ( pp.lisp_apml_tgtype 3.2989 (3) ) ( p.lisp_apml_tgtype 2.5723 (3) ) ( lisp_apml_tgtype 5.5798 (3) ) ( n.lisp_apml_tgtype 0.0000 (3) ) ( nn.lisp_apml_tgtype 0.0000 (3) ) ( pp.lisp_apml_iecount 0.0000 ) ( p.lisp_apml_iecount -0.7231 ) ( lisp_apml_iecount 0.0000 ) ( n.lisp_apml_iecount -2.6922 ) ( nn.lisp_apml_iecount 0.0000 ) ( pp.tobi_accent 0.0000 (H*) ) ( p.tobi_accent 6.0496 (H*) ) ( tobi_accent 20.9174 (H*) ) ( n.tobi_accent 0.0000 (H*) ) ( nn.tobi_accent -6.4804 (H*) ) ( pp.tobi_accent 0.0000 (L*) ) ( p.tobi_accent -5.0000 (L*) ) ( tobi_accent -45.0000 (L*) ) ( n.tobi_accent -6.1079 (L*) ) ( nn.tobi_accent -3.4815 (L*) ) ( pp.tobi_accent 0.0000 (L*+H) ) ( p.tobi_accent 30.0000 (L*+H) ) ( tobi_accent -30.0000 (L*+H) ) ( n.tobi_accent -2.7200 (L*+H) ) ( nn.tobi_accent 0.0000 (L*+H) ) ( pp.tobi_accent 0.0000 (L+H*) ) ( p.tobi_accent 6.0500 (L+H*) ) ( tobi_accent 20.9200 (L+H*) ) ( n.tobi_accent 0.0000 (L+H*) ) ( nn.tobi_accent -6.4800 (L+H*) ) ( pp.tobi_endtone 3.6235 (L-L%) ) ( p.tobi_endtone 0.2801 (L-L%) ) ( tobi_endtone -80 (L-L%) ) ( n.tobi_endtone -6.9938 (L-L%) ) ( nn.tobi_endtone -9.1546 (L-L%) ) ( pp.tobi_endtone 1.6616 (L-H%) ) ( p.tobi_endtone 2.2441 (L-H%) ) ( tobi_endtone 0.0000 (L-H%) ) ( n.tobi_endtone 0.0000 (L-H%) ) ( nn.tobi_endtone -6.6132 (L-H%) ) ( pp.tobi_endtone 0.0000 (H-L%) ) ( p.tobi_endtone 0.0000 (H-L%) ) ( tobi_endtone 0.0000 (H-L%) ) ( n.tobi_endtone -5.4416 (H-L%) ) ( nn.tobi_endtone -19.1570 (H-L%) ) ( pp.tobi_endtone 0.0000 (H-H%) ) ( p.tobi_endtone 0.0000 (H-H%) ) ( tobi_endtone 0.0000 (H-H%) ) ( n.tobi_endtone 0.0000 (H-H%) ) ( nn.tobi_endtone 32.2151 (H-H%) ) ( pp.tobi_endtone -3.7825 (L-) ) ( p.tobi_endtone -15.9598 (L-) ) ( tobi_endtone -36.7241 (L-) ) ( n.tobi_endtone -5.1082 (L-) ) ( nn.tobi_endtone -3.6363 (L-) ) ( pp.tobi_endtone -13.3813 (H-) ) ( p.tobi_endtone 0.0000 (H-) ) ( tobi_endtone -2.5270 (H-) ) ( n.tobi_endtone 0.0000 (H-) ) ( nn.tobi_endtone 0.0000 (H-) ) ( p.tobi_accent 0.0000 (L+H*L-H%) ) ( tobi_accent 40.0000 (L+H*L-H%) ) ( n.tobi_accent 0.0000 (L+H*L-H%) ) ( pp.syl_break 0.7003 ) ( p.syl_break -1.6241 ) ( syl_break 0.0000 ) ( n.syl_break 0.1439 ) ( nn.syl_break 0.5640 ) ( pp.stress -1.9556 ) ( p.stress 0.0000 ) ( stress 3.1383 ) ( n.stress 1.9952 ) ( nn.stress 2.9208 ) ( syl_in -2.5317 ) ( syl_out 1.0103 ) ( ssyl_in 2.0751 ) ( ssyl_out 0.0000 ) ( asyl_in 0.0000 ) ( asyl_out 0.0000 ) ( last_accent 0.0000 ) ( next_accent 0.5869 ) ( sub_phrases 0.0000 ) ( lisp_l_spread -60.0000 ) )) (set! apml_f2b_f0_lr_right '( ( Intercept 162.6687 ) ( pp.lisp_apml_tgtype -4.0459 (1) ) ( p.lisp_apml_tgtype 3.0601 (1) ) ( lisp_apml_tgtype 27.8166 (1) ) ( n.lisp_apml_tgtype 7.4127 (1) ) ( nn.lisp_apml_tgtype 11.3458 (1) ) ( pp.lisp_apml_tgtype -3.8091 (2) ) ( p.lisp_apml_tgtype 1.7434 (2) ) ( lisp_apml_tgtype 17.1672 (2) ) ( n.lisp_apml_tgtype 0.6679 (2) ) ( nn.lisp_apml_tgtype 0.0000 (2) ) ( pp.lisp_apml_tgtype 1.6494 (3) ) ( p.lisp_apml_tgtype 1.2861 (3) ) ( lisp_apml_tgtype 9.5674 (3) ) ( n.lisp_apml_tgtype -3.1085 (3) ) ( nn.lisp_apml_tgtype 0.0000 (3) ) ( pp.lisp_apml_iecount 0.0000 ) ( p.lisp_apml_iecount -0.7829 ) ( lisp_apml_iecount -0.5447 ) ( n.lisp_apml_iecount -1.3461 ) ( nn.lisp_apml_iecount -0.7178 ) ( pp.tobi_accent 0.7904 (H*) ) ( p.tobi_accent 3.0248 (H*) ) ( tobi_accent 14.1116 (H*) ) ( n.tobi_accent 0.0000 (H*) ) ( nn.tobi_accent -3.2402 (H*) ) ( pp.tobi_accent 0.0000 (L*) ) ( p.tobi_accent -2.5000 (L*) ) ( tobi_accent -32.5000 (L*) ) ( n.tobi_accent -3.0539 (L*) ) ( nn.tobi_accent -1.7408 (L*) ) ( pp.tobi_accent 0.0000 (L*+H) ) ( p.tobi_accent 17.5000 (L*+H) ) ( tobi_accent -9.0000 (L*+H) ) ( n.tobi_accent -2.8025 (L*+H) ) ( nn.tobi_accent -0.5455 (L*+H) ) ( pp.tobi_accent 0.7900 (L+H*) ) ( p.tobi_accent 3.0250 (L+H*) ) ( tobi_accent 14.1150 (L+H*) ) ( n.tobi_accent 0.0000 (L+H*) ) ( nn.tobi_accent -3.2400 (L+H*) ) ( pp.tobi_endtone 5.7534 (L-L%) ) ( p.tobi_endtone 0.1401 (L-L%) ) ( tobi_endtone -65 (L-L%) ) ( n.tobi_endtone -11.1795 (L-L%) ) ( nn.tobi_endtone -7.8158 (L-L%) ) ( pp.tobi_endtone 4.4276 (L-H%) ) ( p.tobi_endtone 1.1220 (L-H%) ) ( tobi_endtone 20.0000 (L-H%) ) ( n.tobi_endtone -6.8995 (L-H%) ) ( nn.tobi_endtone -6.1219 (L-H%) ) ( pp.tobi_endtone 2.4327 (H-L%) ) ( p.tobi_endtone 0.0000 (H-L%) ) ( tobi_endtone -7.5781 (H-L%) ) ( n.tobi_endtone -2.7208 (H-L%) ) ( nn.tobi_endtone -14.4838 (H-L%) ) ( pp.tobi_endtone 0.0000 (H-H%) ) ( p.tobi_endtone 0.0000 (H-H%) ) ( tobi_endtone 0.0000 (H-H%) ) ( n.tobi_endtone 0.0000 (H-H%) ) ( nn.tobi_endtone 16.1076 (H-H%) ) ( pp.tobi_endtone -1.8913 (L-) ) ( p.tobi_endtone -15.5651 (L-) ) ( tobi_endtone -40.2021 (L-) ) ( n.tobi_endtone -2.5541 (L-) ) ( nn.tobi_endtone -2.2224 (L-) ) ( pp.tobi_endtone -6.6906 (H-) ) ( p.tobi_endtone -3.5483 (H-) ) ( tobi_endtone -1.2635 (H-) ) ( n.tobi_endtone 0.0000 (H-) ) ( nn.tobi_endtone 0.0000 (H-) ) ( p.tobi_accent 0.0000 (L+H*L-H%) ) ( tobi_accent -40.0000 (L+H*L-H%) ) ( n.tobi_accent 0.0000 (L+H*L-H%) ) ( pp.syl_break 0.3501 ) ( p.syl_break -1.0003 ) ( syl_break -1.5536 ) ( n.syl_break 0.0720 ) ( nn.syl_break 0.5989 ) ( pp.stress -0.9778 ) ( p.stress -0.8046 ) ( stress 1.2124 ) ( n.stress 3.9715 ) ( nn.stress 2.3914 ) ( syl_in -2.3468 ) ( syl_out 0.9792 ) ( ssyl_in 2.0463 ) ( ssyl_out 0.0000 ) ( asyl_in -0.1460 ) ( asyl_out 0.0000 ) ( last_accent -1.0992 ) ( next_accent 0.2935 ) ( sub_phrases 0.0000 ) ( lisp_l_spread -60.0000 ) )) (set! apml_f2b_f0_lr_end '( ( Intercept 165.0901 ) ( pp.lisp_apml_tgtype -8.0918 (1) ) ( p.lisp_apml_tgtype 1.5332 (1) ) ( lisp_apml_tgtype 29.9802 (1) ) ( n.lisp_apml_tgtype 0.0000 (1) ) ( nn.lisp_apml_tgtype 9.7312 (1) ) ( pp.lisp_apml_tgtype -7.6181 (2) ) ( p.lisp_apml_tgtype 0.0000 (2) ) ( lisp_apml_tgtype 20.7778 (2) ) ( n.lisp_apml_tgtype 0.0000 (2) ) ( nn.lisp_apml_tgtype 0.0000 (2) ) ( pp.lisp_apml_tgtype 0.0000 (3) ) ( p.lisp_apml_tgtype 0.0000 (3) ) ( lisp_apml_tgtype 13.5550 (3) ) ( n.lisp_apml_tgtype -6.2170 (3) ) ( nn.lisp_apml_tgtype 0.0000 (3) ) ( pp.lisp_apml_iecount 0.0000 ) ( p.lisp_apml_iecount -0.8428 ) ( lisp_apml_iecount -1.0894 ) ( n.lisp_apml_iecount 0.0000 ) ( nn.lisp_apml_iecount -1.4355 ) ( pp.tobi_accent 1.5807 (H*) ) ( p.tobi_accent 0.0000 (H*) ) ( tobi_accent 7.3057 (H*) ) ( n.tobi_accent 0.0000 (H*) ) ( nn.tobi_accent 0.0000 (H*) ) ( pp.tobi_accent 0.0000 (L*) ) ( p.tobi_accent 0.0000 (L*) ) ( tobi_accent -20.0000 (L*) ) ( n.tobi_accent 0.0000 (L*) ) ( nn.tobi_accent 0.0000 (L*) ) ( pp.tobi_accent 0.0000 (L*+H) ) ( p.tobi_accent 5.0000 (L*+H) ) ( tobi_accent 12.0000 (L*+H) ) ( n.tobi_accent -2.8850 (L*+H) ) ( nn.tobi_accent -1.0910 (L*+H) ) ( pp.tobi_accent 1.5800 (L+H*) ) ( p.tobi_accent 0.0000 (L+H*) ) ( tobi_accent 7.3100 (L+H*) ) ( n.tobi_accent 0.0000 (L+H*) ) ( nn.tobi_accent 0.0000 (L+H*) ) ( pp.tobi_endtone 7.8833 (L-L%) ) ( p.tobi_endtone 0.0000 (L-L%) ) ( tobi_endtone -80 (L-L%) ) ( n.tobi_endtone -35 (L-L%) ) ( nn.tobi_endtone -6.4769 (L-L%) ) ( pp.tobi_endtone 7.1936 (L-H%) ) ( p.tobi_endtone 0.0000 (L-H%) ) ( tobi_endtone 40.0000 (L-H%) ) ( n.tobi_endtone -13.7990 (L-H%) ) ( nn.tobi_endtone -5.6305 (L-H%) ) ( pp.tobi_endtone 4.8654 (H-L%) ) ( p.tobi_endtone 0.0000 (H-L%) ) ( tobi_endtone -15.1561 (H-L%) ) ( n.tobi_endtone 0.0000 (H-L%) ) ( nn.tobi_endtone -9.8107 (H-L%) ) ( pp.tobi_endtone 0.0000 (H-H%) ) ( p.tobi_endtone 0.0000 (H-H%) ) ( tobi_endtone 0.0000 (H-H%) ) ( n.tobi_endtone 0.0000 (H-H%) ) ( nn.tobi_endtone 0.0000 (H-H%) ) ( pp.tobi_endtone 0.0000 (L-) ) ( p.tobi_endtone -15.1705 (L-) ) ( tobi_endtone -43.6801 (L-) ) ( n.tobi_endtone 0.0000 (L-) ) ( nn.tobi_endtone -0.8085 (L-) ) ( pp.tobi_endtone 0.0000 (H-) ) ( p.tobi_endtone -7.0967 (H-) ) ( tobi_endtone 0.0000 (H-) ) ( n.tobi_endtone 0.0000 (H-) ) ( nn.tobi_endtone 0.0000 (H-) ) ( p.tobi_accent 0.0000 (L+H*L-H%) ) ( tobi_accent 60.0000 (L+H*L-H%) ) ( n.tobi_accent -60.0000 (L+H*L-H%) ) ( pp.syl_break 0.0000 ) ( p.syl_break -0.3765 ) ( syl_break -3.1072 ) ( n.syl_break 0.0000 ) ( nn.syl_break 0.6338 ) ( pp.stress 0.0000 ) ( p.stress -1.6093 ) ( stress -0.7136 ) ( n.stress 5.9479 ) ( nn.stress 1.8619 ) ( syl_in -2.1619 ) ( syl_out 0.9481 ) ( ssyl_in 2.0175 ) ( ssyl_out 0.0000 ) ( asyl_in -0.2919 ) ( asyl_out 0.0000 ) ( last_accent -2.1984 ) ( next_accent 0.0000 ) ( sub_phrases 0.0000 ) ( lisp_l_spread -60.0000 ) )) festival/lib/apml_kaldurtreeZ.scm0000644000175000017500000010616407707512373015601 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; A tree to predict zcore durations build from f2b ;;; doesn't use actual phonemes so it can have better generalizations ;;; ;;; Basically copied from ked ;;; (set! apml_kal_durs '( (uh 0.067 0.025) (hh 0.061 0.028) (ao 0.138 0.046) (hv 0.053 0.020) (v 0.051 0.019) (ih 0.058 0.023) (el 0.111 0.043) (ey 0.132 0.042) (em 0.080 0.033) (jh 0.094 0.024) (w 0.054 0.023) (uw 0.107 0.044) (ae 0.120 0.036) (en 0.117 0.056) (k 0.089 0.034) (y 0.048 0.025) (axr 0.147 0.035) ; (l 0.056 0.026) (l 0.066 0.026) (ng 0.064 0.024) (zh 0.071 0.030) (z 0.079 0.034) (brth 0.246 0.046) (m 0.069 0.028) (iy 0.097 0.041) (n 0.059 0.025) (ah 0.087 0.031) (er 0.086 0.010) (b 0.069 0.024) (pau 0.200 0.1) (aw 0.166 0.053) (p 0.088 0.030) (ch 0.115 0.025) (ow 0.134 0.039) (dh 0.031 0.016) (nx 0.049 0.100) (d 0.048 0.021) (ax 0.046 0.024) (h# 0.060 0.083) (r 0.053 0.031) (eh 0.095 0.036) (ay 0.137 0.047) (oy 0.183 0.050) (f 0.095 0.033) (sh 0.108 0.031) (s 0.102 0.037) (g 0.064 0.021) (dx 0.031 0.016) (th 0.093 0.050) (aa 0.094 0.037) (t 0.070 0.020) ) ) (set! apml_kal_duration_cart_tree ' ((name is pau) ((emph_sil is +) ((0.0 -0.5)) ((p.R:SylStructure.parent.parent.lisp_apml_pause = 0.2) ((0.0 0.0)) ((p.R:SylStructure.parent.parent.lisp_apml_pause = 0.4) ((0.0 2.0)) ((p.R:SylStructure.parent.parent.lisp_apml_pause = 0.6) ((0.0 4.0)) ((p.R:SylStructure.parent.parent.lisp_apml_pause = 0.8) ((0.0 6.0)) ((p.R:SylStructure.parent.parent.lisp_apml_pause = 1.0) ((0.0 8.0)) ((p.R:SylStructure.parent.parent.lisp_apml_pause = 1.5) ((0.0 13.0)) ((p.R:SylStructure.parent.parent.lisp_apml_pause = 2.0) ((0.0 18.0)) ((p.R:SylStructure.parent.parent.lisp_apml_pause = 2.5) ((0.0 23.0)) ((p.R:SylStructure.parent.parent.lisp_apml_pause = 3.0) ((0.0 28.0)) ((p.R:SylStructure.parent.parent.pbreak is BB) ((0.0 2.0)) ((0.0 0.0))))))))))))) ((R:SylStructure.parent.accented is 0) ((n.ph_ctype is 0) ((p.ph_vlng is 0) ((R:SylStructure.parent.syl_codasize < 1.5) ((p.ph_ctype is n) ((ph_ctype is f) ((0.559208 -0.783163)) ((1.05215 -0.222704))) ((ph_ctype is s) ((R:SylStructure.parent.syl_break is 2) ((0.589948 0.764459)) ((R:SylStructure.parent.asyl_in < 0.7) ((1.06385 0.567944)) ((0.691943 0.0530272)))) ((ph_vlng is l) ((pp.ph_vfront is 1) ((1.06991 0.766486)) ((R:SylStructure.parent.syl_break is 1) ((0.69665 0.279248)) ((0.670353 0.0567774)))) ((p.ph_ctype is s) ((seg_onsetcoda is coda) ((0.828638 -0.038356)) ((ph_ctype is f) ((0.7631 -0.545853)) ((0.49329 -0.765994)))) ((R:SylStructure.parent.parent.gpos is det) ((R:SylStructure.parent.last_accent < 0.3) ((R:SylStructure.parent.sub_phrases < 1) ((0.811686 0.160195)) ((0.799015 0.713958))) ((0.731599 -0.215472))) ((ph_ctype is r) ((0.673487 0.092772)) ((R:SylStructure.parent.asyl_in < 1) ((0.745273 0.00132813)) ((0.75457 -0.334898))))))))) ((pos_in_syl < 0.5) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((0.902446 -0.041618)) ((R:SylStructure.parent.sub_phrases < 2.3) ((0.900629 0.262952)) ((1.18474 0.594794)))) ((seg_onset_stop is 0) ((R:SylStructure.parent.position_type is mid) ((0.512323 -0.760444)) ((R:SylStructure.parent.syl_out < 6.8) ((pp.ph_vlng is a) ((0.640575 -0.450449)) ((ph_ctype is f) ((R:SylStructure.parent.sub_phrases < 1.3) ((0.862876 -0.296956)) ((R:SylStructure.parent.syl_out < 2.4) ((0.803215 0.0422868)) ((0.877856 -0.154465)))) ((R:SylStructure.parent.syl_out < 3.6) ((R:SylStructure.parent.syl_out < 1.2) ((0.567081 -0.264199)) ((0.598043 -0.541738))) ((0.676843 -0.166623))))) ((0.691678 -0.57173)))) ((R:SylStructure.parent.parent.gpos is cc) ((1.15995 0.313289)) ((pp.ph_vfront is 1) ((0.555993 0.0695819)) ((R:SylStructure.parent.asyl_in < 1.2) ((R:SylStructure.parent.sub_phrases < 2.7) ((0.721635 -0.367088)) ((0.71919 -0.194887))) ((0.547052 -0.0637491))))))) ((ph_ctype is s) ((R:SylStructure.parent.syl_break is 0) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((0.650007 -0.333421)) ((0.846301 -0.165383))) ((0.527756 -0.516332))) ((R:SylStructure.parent.syl_break is 0) ((p.ph_ctype is s) ((0.504414 -0.779112)) ((0.812498 -0.337611))) ((pos_in_syl < 1.4) ((0.513041 -0.745807)) ((p.ph_ctype is s) ((0.350582 -1.04907)) ((0.362 -0.914974)))))))) ((R:SylStructure.parent.syl_break is 0) ((ph_ctype is n) ((R:SylStructure.parent.position_type is initial) ((pos_in_syl < 1.2) ((0.580485 0.172658)) ((0.630973 -0.101423))) ((0.577937 -0.360092))) ((R:SylStructure.parent.syl_out < 2.9) ((R:SylStructure.parent.syl_out < 1.1) ((R:SylStructure.parent.position_type is initial) ((0.896092 0.764189)) ((R:SylStructure.parent.sub_phrases < 3.6) ((ph_ctype is s) ((0.877362 0.555132)) ((0.604511 0.369882))) ((0.799982 0.666966)))) ((seg_onsetcoda is coda) ((p.ph_vlng is a) ((R:SylStructure.parent.last_accent < 0.4) ((0.800736 0.240634)) ((0.720606 0.486176))) ((1.18173 0.573811))) ((0.607147 0.194468)))) ((ph_ctype is r) ((0.88377 0.499383)) ((R:SylStructure.parent.last_accent < 0.5) ((R:SylStructure.parent.position_type is initial) ((R:SylStructure.parent.parent.word_numsyls < 2.4) ((0.62798 0.0737318)) ((0.787334 0.331014))) ((ph_ctype is s) ((0.808368 0.0929299)) ((0.527948 -0.0443271)))) ((seg_coda_fric is 0) ((p.ph_vlng is a) ((0.679745 0.517681)) ((R:SylStructure.parent.sub_phrases < 1.1) ((0.759979 0.128316)) ((0.775233 0.361383)))) ((R:SylStructure.parent.last_accent < 1.3) ((0.696255 0.054136)) ((0.632425 0.246742)))))))) ((pos_in_syl < 0.3) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((0.847602 0.621547)) ((ph_ctype is s) ((0.880645 0.501679)) ((R:SylStructure.parent.sub_phrases < 3.3) ((R:SylStructure.parent.sub_phrases < 0.3) ((0.901014 -0.042049)) ((0.657493 0.183226))) ((0.680126 0.284799))))) ((ph_ctype is s) ((p.ph_vlng is s) ((0.670033 -0.820934)) ((0.863306 -0.348735))) ((ph_ctype is n) ((R:SylStructure.parent.asyl_in < 1.2) ((0.656966 -0.40092)) ((0.530966 -0.639366))) ((seg_coda_fric is 0) ((1.04153 0.364857)) ((pos_in_syl < 1.2) ((R:SylStructure.parent.syl_out < 3.4) ((0.81503 -0.00768613)) ((0.602665 -0.197753))) ((0.601844 -0.394632))))))))) ((n.ph_ctype is f) ((pos_in_syl < 1.5) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((pos_in_syl < 0.1) ((1.63863 0.938841)) ((R:SylStructure.parent.position_type is initial) ((0.897722 -0.0796637)) ((nn.ph_vheight is 0) ((0.781081 0.480026)) ((0.779711 0.127175))))) ((ph_ctype is r) ((p.ph_ctype is s) ((0.581329 -0.708767)) ((0.564366 -0.236212))) ((ph_vlng is a) ((p.ph_ctype is r) ((0.70992 -0.273389)) ((R:SylStructure.parent.parent.gpos is in) ((0.764696 0.0581338)) ((nn.ph_vheight is 0) ((0.977737 0.721904)) ((R:SylStructure.parent.sub_phrases < 2.2) ((pp.ph_vfront is 0) ((0.586708 0.0161206)) ((0.619949 0.227372))) ((0.707285 0.445569)))))) ((ph_ctype is n) ((R:SylStructure.parent.syl_break is 1) ((nn.ph_vfront is 2) ((0.430295 -0.120097)) ((0.741371 0.219042))) ((0.587492 0.321245))) ((p.ph_ctype is n) ((0.871586 0.134075)) ((p.ph_ctype is r) ((0.490751 -0.466418)) ((R:SylStructure.parent.syl_codasize < 1.3) ((R:SylStructure.parent.sub_phrases < 2.2) ((p.ph_ctype is s) ((0.407452 -0.425925)) ((0.644771 -0.542809))) ((0.688772 -0.201899))) ((ph_vheight is 1) ((nn.ph_vheight is 0) ((0.692018 0.209018)) ((0.751345 -0.178136))) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.3) ((R:SylStructure.parent.asyl_in < 1.5) ((0.599633 -0.235593)) ((0.60042 0.126118))) ((p.ph_vlng is a) ((0.7148 -0.174812)) ((R:SylStructure.parent.parent.gpos is content) ((0.761296 -0.231509)) ((0.813081 -0.536405))))))))))))) ((ph_ctype is n) ((0.898844 0.163343)) ((p.ph_vlng is s) ((seg_coda_fric is 0) ((0.752921 -0.45528)) ((0.890079 -0.0998025))) ((ph_ctype is f) ((0.729376 -0.930547)) ((ph_ctype is s) ((R:SylStructure.parent.R:Syllable.p.syl_break is 0) ((0.745052 -0.634119)) ((0.521502 -0.760176))) ((R:SylStructure.parent.syl_break is 1) ((0.766575 -0.121355)) ((0.795616 -0.557509)))))))) ((p.ph_vlng is 0) ((p.ph_ctype is r) ((ph_vlng is 0) ((0.733659 -0.402734)) ((R:SylStructure.parent.sub_phrases < 1.5) ((ph_vlng is s) ((0.326176 -0.988478)) ((n.ph_ctype is s) ((0.276471 -0.802536)) ((0.438283 -0.900628)))) ((nn.ph_vheight is 0) ((ph_vheight is 2) ((0.521 -0.768992)) ((0.615436 -0.574918))) ((ph_vheight is 1) ((0.387376 -0.756359)) ((pos_in_syl < 0.3) ((0.417235 -0.808937)) ((0.384043 -0.93315))))))) ((ph_vlng is a) ((ph_ctype is 0) ((n.ph_ctype is s) ((p.ph_ctype is f) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((0.415908 -0.428493)) ((pos_in_syl < 0.1) ((0.790441 0.0211071)) ((0.452465 -0.254485)))) ((p.ph_ctype is s) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((0.582447 -0.389966)) ((0.757648 0.185781))) ((R:SylStructure.parent.sub_phrases < 1.4) ((0.628965 0.422551)) ((0.713613 0.145576))))) ((seg_onset_stop is 0) ((R:SylStructure.parent.R:Syllable.p.syl_break is 0) ((pp.ph_vfront is 1) ((0.412363 -0.62319)) ((R:SylStructure.parent.syl_out < 3.6) ((0.729259 -0.317324)) ((0.441633 -0.591051)))) ((R:SylStructure.parent.syl_break is 1) ((R:SylStructure.parent.sub_phrases < 2.7) ((0.457728 -0.405607)) ((0.532411 -0.313148))) ((R:SylStructure.parent.last_accent < 0.3) ((1.14175 0.159416)) ((0.616396 -0.254651))))) ((R:SylStructure.parent.position_type is initial) ((0.264181 -0.799896)) ((0.439801 -0.551309))))) ((R:SylStructure.parent.position_type is final) ((0.552027 -0.707084)) ((0.585661 -0.901874)))) ((ph_ctype is s) ((pos_in_syl < 1.2) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((pp.ph_vfront is 1) ((0.607449 0.196466)) ((0.599662 0.00382414))) ((0.64109 -0.12859))) ((pp.ph_vfront is 1) ((0.720484 -0.219339)) ((0.688707 -0.516734)))) ((ph_vlng is s) ((n.ph_ctype is s) ((R:SylStructure.parent.parent.gpos is content) ((R:SylStructure.parent.position_type is single) ((0.659206 0.159445)) ((R:SylStructure.parent.parent.word_numsyls < 3.5) ((R:SylStructure.parent.sub_phrases < 2) ((0.447186 -0.419103)) ((0.631822 -0.0928561))) ((0.451623 -0.576116)))) ((ph_vheight is 3) ((0.578626 -0.64583)) ((0.56636 -0.4665)))) ((R:SylStructure.parent.parent.gpos is in) ((0.771516 -0.217292)) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((0.688571 -0.304382)) ((R:SylStructure.parent.parent.gpos is content) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((n.ph_ctype is n) ((0.556085 -0.572203)) ((0.820173 -0.240338))) ((R:SylStructure.parent.parent.word_numsyls < 2.2) ((0.595398 -0.588171)) ((0.524737 -0.95797)))) ((R:SylStructure.parent.sub_phrases < 3.9) ((0.371492 -0.959427)) ((0.440479 -0.845747))))))) ((R:SylStructure.parent.R:Syllable.p.syl_break is 0) ((p.ph_ctype is f) ((0.524088 -0.482247)) ((nn.ph_vheight is 1) ((0.587666 -0.632362)) ((ph_vlng is l) ((R:SylStructure.parent.position_type is final) ((0.513286 -0.713117)) ((0.604613 -0.924308))) ((R:SylStructure.parent.syl_codasize < 2.2) ((0.577997 -0.891342)) ((0.659804 -1.15252)))))) ((pp.ph_vlng is s) ((ph_ctype is f) ((0.813383 -0.599624)) ((0.984027 -0.0771909))) ((p.ph_ctype is f) ((R:SylStructure.parent.parent.gpos is in) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((0.313572 -1.03242)) ((0.525854 -0.542799))) ((R:SylStructure.parent.syl_out < 2.8) ((0.613007 -0.423979)) ((0.570258 -0.766379)))) ((R:SylStructure.parent.syl_break is 1) ((R:SylStructure.parent.parent.gpos is to) ((0.364585 -0.792895)) ((ph_vlng is l) ((0.69143 -0.276816)) ((0.65673 -0.523721)))) ((R:SylStructure.parent.syl_out < 3.6) ((R:SylStructure.parent.position_type is initial) ((0.682096 -0.488102)) ((0.406364 -0.731758))) ((0.584694 -0.822229))))))))))) ((n.ph_ctype is r) ((R:SylStructure.parent.position_type is initial) ((p.ph_vlng is a) ((0.797058 1.02334)) ((ph_ctype is s) ((1.0548 0.536277)) ((0.817253 0.138201)))) ((R:SylStructure.parent.sub_phrases < 1.1) ((R:SylStructure.parent.syl_out < 3.3) ((0.884574 -0.23471)) ((0.772063 -0.525292))) ((nn.ph_vfront is 1) ((1.25254 0.417485)) ((0.955557 -0.0781996))))) ((pp.ph_vfront is 0) ((ph_ctype is f) ((n.ph_ctype is s) ((R:SylStructure.parent.parent.gpos is content) ((R:SylStructure.parent.R:Syllable.p.syl_break is 0) ((0.583506 -0.56941)) ((0.525949 -0.289362))) ((0.749316 -0.0921038))) ((p.ph_vlng is s) ((0.734234 0.139463)) ((0.680119 -0.0708717)))) ((ph_vlng is s) ((ph_vheight is 1) ((0.908712 -0.618971)) ((0.55344 -0.840495))) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 1.2) ((pos_in_syl < 1.2) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((0.838715 0.00913392)) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((ph_vheight is 2) ((0.555513 -0.512523)) ((R:SylStructure.parent.position_type is initial) ((0.758711 0.121704)) ((0.737555 -0.25637)))) ((R:SylStructure.parent.syl_out < 3.1) ((n.ph_ctype is s) ((0.611756 -0.474522)) ((1.05437 -0.247206))) ((R:SylStructure.parent.syl_codasize < 2.2) ((R:SylStructure.parent.position_type is final) ((0.567761 -0.597866)) ((0.785599 -0.407765))) ((0.575598 -0.741256)))))) ((ph_ctype is s) ((n.ph_ctype is s) ((0.661069 -1.08426)) ((0.783184 -0.39789))) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((R:SylStructure.parent.sub_phrases < 2.6) ((0.511323 -0.666011)) ((0.691878 -0.499492))) ((ph_ctype is r) ((0.482131 -0.253186)) ((0.852955 -0.372832)))))) ((0.854447 -0.0936489))))) ((R:SylStructure.parent.position_type is final) ((0.685939 -0.249982)) ((R:SylStructure.parent.syl_out < 3.2) ((0.989843 0.18086)) ((0.686805 -0.0402908))))))))) ((R:SylStructure.parent.syl_out < 2.4) ((R:SylStructure.parent.syl_out < 0.2) ((seg_onsetcoda is coda) ((ph_ctype is s) ((R:SylStructure.parent.syl_break is 4) ((pp.ph_vlng is 0) ((0.959737 1.63203)) ((1.20714 0.994933))) ((n.ph_ctype is 0) ((R:SylStructure.parent.syl_break is 2) ((0.864809 0.214457)) ((0.874278 0.730381))) ((pp.ph_vfront is 0) ((seg_coda_fric is 0) ((1.20844 -0.336221)) ((1.01357 0.468302))) ((0.658106 -0.799121))))) ((n.ph_ctype is f) ((ph_ctype is f) ((1.26332 0.0300613)) ((ph_vlng is d) ((1.02719 1.1649)) ((ph_ctype is 0) ((R:SylStructure.parent.asyl_in < 1.2) ((1.14048 2.2668)) ((ph_vheight is 1) ((1.15528 1.50375)) ((1.42406 2.07927)))) ((R:SylStructure.parent.sub_phrases < 1.1) ((0.955892 1.10243)) ((R:SylStructure.parent.syl_break is 2) ((1.32682 1.8432)) ((1.27582 1.59853))))))) ((n.ph_ctype is 0) ((ph_ctype is n) ((R:SylStructure.parent.syl_break is 2) ((1.45399 1.12927)) ((1.05543 0.442376))) ((R:SylStructure.parent.syl_break is 4) ((R:SylStructure.parent.position_type is final) ((ph_ctype is f) ((1.46434 1.76508)) ((0.978055 0.7486))) ((1.2395 2.30826))) ((ph_ctype is 0) ((0.935325 1.69917)) ((nn.ph_vfront is 1) ((1.20456 1.31128)) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((nn.ph_vheight is 0) ((1.16907 0.212421)) ((0.952091 0.653094))) ((p.ph_ctype is 0) ((1.05502 1.25802)) ((0.818731 0.777568)))))))) ((ph_ctype is f) ((p.ph_ctype is 0) ((1.03918 0.163941)) ((0.737545 -0.167063))) ((R:SylStructure.parent.position_type is final) ((n.ph_ctype is n) ((R:SylStructure.parent.last_accent < 0.5) ((R:SylStructure.parent.sub_phrases < 2.8) ((0.826207 -0.000859005)) ((0.871119 0.273433))) ((R:SylStructure.parent.parent.word_numsyls < 2.4) ((1.17405 1.05694)) ((0.858394 0.244916)))) ((R:SylStructure.parent.syl_codasize < 2.2) ((p.ph_ctype is 0) ((1.14092 1.21187)) ((R:SylStructure.parent.syl_break is 2) ((1.02653 0.59865)) ((0.94248 1.1634)))) ((seg_coda_fric is 0) ((1.07441 0.292935)) ((1.15736 0.92574))))) ((ph_vlng is s) ((R:SylStructure.parent.syl_break is 2) ((1.34638 1.23484)) ((0.951514 2.02008))) ((ph_ctype is 0) ((p.ph_ctype is r) ((0.806106 0.697089)) ((R:SylStructure.parent.syl_break is 2) ((1.10891 0.992197)) ((1.04657 1.51093)))) ((1.18165 0.520952))))))))) ((p.ph_vlng is 0) ((pos_in_syl < 0.7) ((R:SylStructure.parent.position_type is final) ((ph_ctype is r) ((0.966357 0.185827)) ((ph_ctype is s) ((0.647163 0.0332298)) ((0.692972 -0.534917)))) ((ph_ctype is s) ((0.881521 0.575107)) ((p.ph_ctype is f) ((0.8223 -0.111275)) ((R:SylStructure.parent.last_accent < 0.3) ((0.969188 0.09447)) ((0.894438 0.381947)))))) ((p.ph_ctype is f) ((0.479748 -0.490108)) ((0.813125 -0.201268)))) ((ph_ctype is s) ((0.908566 1.20397)) ((R:SylStructure.parent.last_accent < 1.2) ((0.88078 0.636568)) ((0.978087 1.07763)))))) ((pos_in_syl < 1.3) ((R:SylStructure.parent.syl_break is 0) ((pos_in_syl < 0.1) ((R:SylStructure.parent.position_type is initial) ((p.ph_ctype is n) ((0.801651 -0.0163359)) ((ph_ctype is s) ((n.ph_ctype is r) ((0.893307 1.07253)) ((p.ph_vlng is 0) ((0.92651 0.525806)) ((0.652444 0.952792)))) ((p.ph_vlng is 0) ((seg_onsetcoda is coda) ((0.820151 0.469117)) ((p.ph_ctype is f) ((0.747972 -0.0716448)) ((ph_ctype is f) ((0.770882 0.457137)) ((0.840905 0.102492))))) ((R:SylStructure.parent.syl_out < 1.1) ((0.667824 0.697337)) ((0.737967 0.375114)))))) ((ph_vheight is 1) ((0.624353 0.410671)) ((R:SylStructure.parent.asyl_in < 0.8) ((0.647905 -0.331055)) ((p.ph_ctype is s) ((0.629039 -0.240616)) ((0.749277 -0.0191273)))))) ((ph_vheight is 3) ((p.ph_ctype is s) ((0.626922 0.556537)) ((0.789357 0.153892))) ((seg_onsetcoda is coda) ((n.ph_ctype is 0) ((R:SylStructure.parent.parent.word_numsyls < 3.4) ((0.744714 0.123242)) ((0.742039 0.295753))) ((seg_coda_fric is 0) ((R:SylStructure.parent.parent.word_numsyls < 2.4) ((ph_vheight is 1) ((0.549715 -0.341018)) ((0.573641 -0.00893114))) ((nn.ph_vfront is 2) ((0.67099 -0.744625)) ((0.664438 -0.302803)))) ((p.ph_vlng is 0) ((0.630028 0.113815)) ((0.632794 -0.128733))))) ((ph_ctype is r) ((0.367169 -0.854509)) ((0.94334 -0.216179)))))) ((n.ph_ctype is f) ((ph_vlng is 0) ((1.3089 0.46195)) ((R:SylStructure.parent.syl_codasize < 1.3) ((1.07673 0.657169)) ((pp.ph_vlng is 0) ((0.972319 1.08222)) ((1.00038 1.46257))))) ((p.ph_vlng is l) ((1.03617 0.785204)) ((p.ph_vlng is a) ((R:SylStructure.parent.position_type is final) ((1.00681 0.321168)) ((0.928115 0.950834))) ((ph_vlng is 0) ((pos_in_syl < 0.1) ((R:SylStructure.parent.position_type is final) ((0.863682 -0.167374)) ((nn.ph_vheight is 0) ((p.ph_ctype is f) ((0.773591 -0.00374425)) ((R:SylStructure.parent.syl_out < 1.1) ((0.951802 0.228448)) ((1.02282 0.504252)))) ((1.09721 0.736476)))) ((R:SylStructure.parent.position_type is final) ((1.04302 0.0590974)) ((0.589208 -0.431535)))) ((n.ph_ctype is 0) ((1.27879 1.00642)) ((ph_vlng is s) ((R:SylStructure.parent.asyl_in < 1.4) ((0.935787 0.481652)) ((0.9887 0.749861))) ((R:SylStructure.parent.syl_out < 1.1) ((R:SylStructure.parent.position_type is final) ((0.921307 0.0696307)) ((0.83675 0.552212))) ((0.810076 -0.0479225)))))))))) ((ph_ctype is s) ((n.ph_ctype is s) ((0.706959 -1.0609)) ((p.ph_ctype is n) ((0.850614 -0.59933)) ((n.ph_ctype is r) ((0.665947 0.00698725)) ((n.ph_ctype is 0) ((R:SylStructure.parent.position_type is initial) ((0.762889 -0.0649044)) ((0.723956 -0.248899))) ((R:SylStructure.parent.sub_phrases < 1.4) ((0.632957 -0.601987)) ((0.889114 -0.302401))))))) ((ph_ctype is f) ((R:SylStructure.parent.syl_codasize < 2.2) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((R:SylStructure.parent.syl_out < 1.1) ((0.865267 0.164636)) ((0.581827 -0.0989051))) ((nn.ph_vfront is 2) ((0.684459 -0.316836)) ((0.778854 -0.0961191)))) ((R:SylStructure.parent.syl_out < 1.1) ((p.ph_ctype is s) ((0.837964 -0.429437)) ((0.875304 -0.0652743))) ((0.611071 -0.635089)))) ((p.ph_ctype is r) ((R:SylStructure.parent.syl_out < 1.1) ((0.762012 0.0139361)) ((0.567983 -0.454845))) ((R:SylStructure.parent.syl_codasize < 2.2) ((ph_ctype is l) ((1.18845 0.809091)) ((R:SylStructure.parent.position_type is initial) ((ph_ctype is n) ((0.773548 -0.277092)) ((1.01586 0.281001))) ((p.ph_ctype is 0) ((1.06831 0.699145)) ((0.924189 0.241873))))) ((R:SylStructure.parent.syl_break is 0) ((ph_ctype is n) ((0.592321 -0.470784)) ((0.778688 -0.072112))) ((n.ph_ctype is s) ((1.08848 0.0733489)) ((1.25674 0.608371)))))))))) ((pos_in_syl < 0.7) ((p.ph_vlng is 0) ((R:SylStructure.parent.position_type is mid) ((ph_ctype is 0) ((ph_vheight is 2) ((0.456225 -0.293282)) ((0.561529 -0.0816115))) ((0.6537 -0.504024))) ((ph_ctype is s) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((1.31586 0.98395)) ((R:SylStructure.parent.position_type is single) ((0.816869 0.634789)) ((R:SylStructure.parent.syl_out < 4.4) ((1.05578 0.479029)) ((R:SylStructure.parent.asyl_in < 0.4) ((1.11813 0.143214)) ((0.87178 0.406834)))))) ((n.ph_ctype is n) ((R:SylStructure.parent.last_accent < 0.6) ((0.838154 -0.415599)) ((0.924024 0.110288))) ((seg_onsetcoda is coda) ((nn.ph_vfront is 2) ((0.670096 0.0314187)) ((n.ph_ctype is f) ((1.00363 0.693893)) ((R:SylStructure.parent.syl_out < 6) ((0.772363 0.215675)) ((0.920313 0.574068))))) ((R:SylStructure.parent.position_type is final) ((0.673837 -0.458142)) ((R:SylStructure.parent.sub_phrases < 2.8) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((0.894817 0.304628)) ((ph_ctype is n) ((0.787302 -0.23094)) ((R:SylStructure.parent.asyl_in < 1.2) ((ph_ctype is f) ((R:SylStructure.parent.last_accent < 0.5) ((1.12278 0.326954)) ((0.802236 -0.100616))) ((0.791255 -0.0919132))) ((0.95233 0.219053))))) ((R:SylStructure.parent.position_type is initial) ((ph_ctype is f) ((1.0616 0.216118)) ((0.703216 -0.00834086))) ((ph_ctype is f) ((1.22277 0.761763)) ((0.904811 0.332721)))))))))) ((ph_vheight is 0) ((p.ph_vlng is s) ((0.873379 0.217178)) ((n.ph_ctype is r) ((0.723915 1.29451)) ((n.ph_ctype is 0) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((R:SylStructure.parent.sub_phrases < 4) ((seg_coda_fric is 0) ((p.ph_vlng is l) ((0.849154 0.945261)) ((0.633261 0.687498))) ((0.728546 0.403076))) ((0.850962 1.00255))) ((0.957999 1.09113))) ((0.85771 0.209045))))) ((ph_vheight is 2) ((0.803401 -0.0544067)) ((0.681353 0.256045))))) ((n.ph_ctype is f) ((ph_ctype is s) ((p.ph_vlng is 0) ((0.479307 -0.9673)) ((0.700477 -0.351397))) ((ph_ctype is f) ((0.73467 -0.6233)) ((R:SylStructure.parent.syl_break is 0) ((p.ph_ctype is s) ((0.56282 0.266234)) ((p.ph_ctype is r) ((0.446203 -0.302281)) ((R:SylStructure.parent.sub_phrases < 2.7) ((ph_ctype is 0) ((0.572016 -0.0102436)) ((0.497358 -0.274514))) ((0.545477 0.0482177))))) ((ph_vlng is s) ((0.805269 0.888495)) ((ph_ctype is n) ((0.869854 0.653018)) ((R:SylStructure.parent.sub_phrases < 2.2) ((0.735031 0.0612886)) ((0.771859 0.346637)))))))) ((R:SylStructure.parent.syl_codasize < 1.4) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.3) ((R:SylStructure.parent.position_type is initial) ((0.743458 0.0411808)) ((1.13068 0.613305))) ((pos_in_syl < 1.2) ((R:SylStructure.parent.R:Syllable.p.syl_break is 1) ((1.11481 0.175467)) ((0.937893 -0.276407))) ((0.74264 -0.550878)))) ((pos_in_syl < 3.4) ((seg_onsetcoda is coda) ((ph_ctype is r) ((n.ph_ctype is s) ((0.714319 -0.240328)) ((p.ph_ctype is 0) ((0.976987 0.330352)) ((1.1781 -0.0816682)))) ((ph_ctype is l) ((n.ph_ctype is 0) ((1.39137 0.383533)) ((0.725585 -0.324515))) ((ph_vheight is 3) ((ph_vlng is d) ((0.802626 -0.62487)) ((n.ph_ctype is r) ((0.661091 -0.513869)) ((R:SylStructure.parent.position_type is initial) ((R:SylStructure.parent.parent.word_numsyls < 2.4) ((0.482285 0.207874)) ((0.401601 -0.0204711))) ((0.733755 0.397372))))) ((n.ph_ctype is r) ((p.ph_ctype is 0) ((pos_in_syl < 1.2) ((0.666325 0.271734)) ((nn.ph_vheight is 0) ((0.642401 -0.261466)) ((0.783684 -0.00956571)))) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((0.692225 -0.381895)) ((0.741921 -0.0898767)))) ((nn.ph_vfront is 2) ((ph_ctype is s) ((0.697527 -1.12626)) ((n.ph_ctype is s) ((ph_vlng is 0) ((R:SylStructure.parent.sub_phrases < 2.4) ((0.498719 -0.906926)) ((0.635342 -0.625651))) ((0.45886 -0.385089))) ((0.848596 -0.359702)))) ((p.ph_vlng is a) ((p.ph_ctype is 0) ((0.947278 0.216904)) ((0.637933 -0.394349))) ((p.ph_ctype is r) ((R:SylStructure.parent.syl_break is 0) ((0.529903 -0.860573)) ((0.581378 -0.510488))) ((ph_vlng is 0) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((seg_onset_stop is 0) ((R:SylStructure.parent.syl_break is 0) ((p.ph_vlng is d) ((0.768363 0.0108428)) ((ph_ctype is s) ((0.835756 -0.035054)) ((ph_ctype is f) ((p.ph_vlng is s) ((0.602016 -0.179727)) ((0.640126 -0.297341))) ((0.674628 -0.542602))))) ((ph_ctype is s) ((0.662261 -0.60496)) ((0.662088 -0.432058)))) ((R:SylStructure.parent.syl_out < 4.4) ((0.582448 -0.389079)) ((ph_ctype is s) ((0.60413 -0.73564)) ((0.567153 -0.605444))))) ((R:SylStructure.parent.R:Syllable.p.syl_break is 2) ((0.761115 -0.827377)) ((ph_ctype is n) ((0.855183 -0.275338)) ((R:SylStructure.parent.syl_break is 0) ((0.788288 -0.802801)) ((R:SylStructure.parent.syl_codasize < 2.2) ((0.686134 -0.371234)) ((0.840184 -0.772883))))))) ((pos_in_syl < 1.2) ((R:SylStructure.parent.syl_break is 0) ((n.ph_ctype is n) ((0.423592 -0.655006)) ((R:SylStructure.parent.syl_out < 4.4) ((0.595269 -0.303751)) ((0.478433 -0.456882)))) ((0.688133 -0.133182))) ((seg_onset_stop is 0) ((1.27464 0.114442)) ((0.406837 -0.167545)))))))))))) ((ph_ctype is r) ((0.462874 -0.87695)) ((R:SylStructure.parent.R:Syllable.n.syl_onsetsize < 0.2) ((0.645442 -0.640572)) ((0.673717 -0.321322))))) ((0.61008 -0.925472)))))))) ;; RMSE 0.8085 Correlation is 0.5899 Mean (abs) Error 0.6024 (0.5393) )) (provide 'apml_kaldurtreeZ) festival/lib/sec.ts20.quad.ngrambin0000644000175000017500000010311007262605650015562 0ustar awbawbNgramBin_2 4 cc cd dt ex fw in j md n of pdt pos prp punc r to uh v wdt wp wrb 1 B NB ?¢ûc*R-ÀF?ü¬€”Ò&Ö?¢ûc*R-À?ü¬€”Ò&Ö?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-À@,?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@?ꆗ ñß—À?¢ûc*R-À?ü¬€”Ò&Ö?ꆗ ñß—?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-@@€?¢ûc*R-À?ꆗ ñß—@Y@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À"@?¢ûc*R-À?ꆗ ñß—@"?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À;?ü¬€”Ò&Ö?¢ûc*R-À&@"?¢ûc*R-ÀJ€?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-@7?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-@?¢ûc*R-À@*?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@ ?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À @?ü¬€”Ò&Ö?¢ûc*R-À @@?ꆗ ñß—@?¢ûc*R-À@@H€?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-@5?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À,?ꆗ ñß—?¢ûc*R-À3?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À ?ꆗ ñß—À?¢ûc*R-À?ꆗ ñß—@@?¢ûc*R-À@?ü¬€”Ò&Ö@@@@@ ?¢ûc*R-À?ꆗ ñß—À?¢ûc*R-À@U€?ü¬€”Ò&Ö@?ü¬€”Ò&ÖÀ?¢ûc*R-À@3@6@?¢ûc*R-À8?ꆗ ñß—?¢ûc*R-À1?ꆗ ñß—?¢ûc*R-À*?ü¬€”Ò&Ö?¢ûc*R-À3?ꆗ ñß—?¢ûc*R-ÀRÀ?ü¬€”Ò&Ö?¢ûc*R-@?¢ûc*R-@;?¢ûc*R-À"?ꆗ ñß—À@?¢ûc*R-À@T?¢ûc*R-À*?ꆗ ñß—?¢ûc*R-À@.?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-?ꆗ ñß—?¢ûc*R-À"@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À ?ꆗ ñß—?¢ûc*R-?ꆗ ñß—@@@(?¢ûc*R-À@@,?ꆗ ñß—@,?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-@(?¢ûc*R-@$?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—@.?¢ûc*R-ÀE€@?¢ûc*R-ÀJ?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—@@&?¢ûc*R-À@@:?ꆗ ñß—@?¢ûc*R-À@ ?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@A€?¢ûc*R-@0?ꆗ ñß—@A?ü¬€”Ò&Ö@"?¢ûc*R-À@&?¢ûc*R-À?ꆗ ñß—À?¢ûc*R-À2?ꆗ ñß—?¢ûc*R-À3@?¢ûc*R-ÀE€@?¢ûc*R-À@€@?¢ûc*R-À"?ü¬€”Ò&Ö?¢ûc*R-ÀK€@6?¢ûc*R-@?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À@?¢ûc*R-À ?ü¬€”Ò&Ö@$?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-À?ü¬€”Ò&Ö@&?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À@?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À:@?¢ûc*R-À]@@?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@0?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-À5?ꆗ ñß—?¢ûc*R-À@4?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-À*?ꆗ ñß—?¢ûc*R-ÀG€?ü¬€”Ò&Ö@@ ?ü¬€”Ò&Ö@?¢ûc*R-À@*@&?¢ûc*R-@?ꆗ ñß—?¢ûc*R-?ꆗ ñß—@9?ü¬€”Ò&Ö@?¢ûc*R-À@?ꆗ ñß—?¢ûc*R-À@L?¢ûc*R-@1@@?¢ûc*R-À@,@ ?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-@4?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-Àa @?¢ûc*R-À @ ?ꆗ ñß—@?ꆗ ñß—@3?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@ ?ü¬€”Ò&Ö?¢ûc*R-À@?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À*?ꆗ ñß—À?¢ûc*R-?ꆗ ñß—?¢ûc*R-À*@?¢ûc*R-?ꆗ ñß—?¢ûc*R-À1?ꆗ ñß—?¢ûc*R-ÀU?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-ÀUÀ?ü¬€”Ò&Ö?¢ûc*R-Àa @"?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@B€?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-ÀF?ꆗ ñß—?¢ûc*R-ÀRÀ?ꆗ ñß—?¢ûc*R-À9?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À@?¢ûc*R-À*?ꆗ ñß—?¢ûc*R-À@@$?ü¬€”Ò&Ö@0?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@@@*@M?¢ûc*R-À@@ƒˆ@@"?¢ûc*R-À@8?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@@?¢ûc*R-ÀJ@K@:?ꆗ ñß—À@(@$?ü¬€”Ò&Ö?¢ûc*R-À@a @``@@"@,@2@(@mà@O@q€?¢ûc*R-À@G@"@8?¢ûc*R-@€@2@A@A€@6?¢ûc*R-À@]€@g @7@.@@@"@?¢ûc*R-À@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À&?ü¬€”Ò&Ö?¢ûc*R-ÀPÀ?ꆗ ñß—?¢ûc*R-ÀD€?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À1?ꆗ ñß—?¢ûc*R-À ?ü¬€”Ò&Ö?¢ûc*R-?ꆗ ñß—À?¢ûc*R-À@?ꆗ ñß—?ü¬€”Ò&Ö@?¢ûc*R-À?ꆗ ñß—@"?ꆗ ñß—?¢ûc*R-À@?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—À?¢ûc*R-À@?¢ûc*R-À1?ü¬€”Ò&Ö@C?¢ûc*R-À*?ü¬€”Ò&Ö?¢ûc*R-@?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@ ?¢ûc*R-ÀE€?ꆗ ñß—?¢ûc*R-ÀL?ꆗ ñß—@0?¢ûc*R-À?ü¬€”Ò&Ö@?¢ûc*R-@?¢ûc*R-À@7?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-@?¢ûc*R-@ ?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-ÀE€?ꆗ ñß—?¢ûc*R-ÀB€?ꆗ ñß—?¢ûc*R-À@,?¢ûc*R-À{p?ꆗ ñß—?¢ûc*R-Àc ?ü¬€”Ò&Ö?¢ûc*R-ÀR€?ü¬€”Ò&Ö?¢ûc*R-À@?¢ûc*R-ÀR€?ü¬€”Ò&Ö?¢ûc*R-ÀXÀ@?¢ûc*R-@A?¢ûc*R-À@?¢ûc*R-À@?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-@$?¢ûc*R-À@?¢ûc*R-À€x?ꆗ ñß—?¢ûc*R-ÀA€?ꆗ ñß—?¢ûc*R-À^@?ꆗ ñß—?¢ûc*R-ÀQ€?ꆗ ñß—?¢ûc*R-ÀyÀ?ꆗ ñß—?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-@(?ꆗ ñß—À?¢ûc*R-À@@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@G€?¢ûc*R-@"?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@.?¢ûc*R-@?¢ûc*R-À0?ꆗ ñß—@*?ꆗ ñß—?¢ûc*R-À@?ü¬€”Ò&Ö@qÀ?¢ûc*R-À?ü¬€”Ò&Ö@†?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@8?¢ûc*R-@(?¢ûc*R-À@@3?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-ÀC€@?¢ûc*R-ÀK€?ꆗ ñß—À@1?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-@?¢ûc*R-À@$?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À ?ꆗ ñß—@?¢ûc*R-?ꆗ ñß—À?¢ûc*R-À@À@&?¢ûc*R-À@@[À?¢ûc*R-@?¢ûc*R-À@*?¢ûc*R-À?ꆗ ñß—À?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À9?ꆗ ñß—?¢ûc*R-À:@:@=?ꆗ ñß—@&@À?¢ûc*R-À@J€@B€?ü¬€”Ò&Ö@?ü¬€”Ò&ÖÀ@`@@@7?¢ûc*R-À@<?ü¬€”Ò&Ö@?¢ûc*R-@o€?ꆗ ñß—@@&@?¢ûc*R-À@A@C@?ꆗ ñß—À?¢ûc*R-@?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À&@?¢ûc*R-À@?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À*@?¢ûc*R-À3?ü¬€”Ò&Ö?¢ûc*R-ÀO@?ꆗ ñß—À?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@"?ü¬€”Ò&Ö?¢ûc*R-@F€?ꆗ ñß—@;?¢ûc*R-@\À?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À@I€?ꆗ ñß—@"?¢ûc*R-@?¢ûc*R-À@@]€?¢ûc*R-À @?¢ûc*R-@?ꆗ ñß—@?ü¬€”Ò&Ö?¢ûc*R-À@@?ꆗ ñß—À?¢ûc*R-À@@?¢ûc*R-À@?ü¬€”Ò&Ö?¢ûc*R-À@?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-@?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-?ꆗ ñß—@ ?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@ ?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À"@0?ꆗ ñß—@?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-?ꆗ ñß—?¢ûc*R-À"?ü¬€”Ò&Ö?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À&?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À"@?¢ûc*R-ÀJ?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@@1?¢ûc*R-À?ü¬€”Ò&Ö@?ꆗ ñß—@ ?¢ûc*R-À@(?¢ûc*R-À@@?¢ûc*R-@?¢ûc*R-@ ?¢ûc*R-À@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À"@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À3?ü¬€”Ò&Ö?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-À3@?¢ûc*R-À(?ꆗ ñß—?¢ûc*R-À4?ꆗ ñß—?¢ûc*R-ÀF€?ꆗ ñß—?¢ûc*R-À5@?¢ûc*R-À@?¢ûc*R-@J?¢ûc*R-À@?¢ûc*R-À@?¢ûc*R-À@ ?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-À,?ꆗ ñß—@(?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À5?ꆗ ñß—?¢ûc*R-À@"?¢ûc*R-À]@@$?¢ûc*R-@N€?¢ûc*R-À@?¢ûc*R-@"?¢ûc*R-À?ꆗ ñß—@>?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@6?¢ûc*R-?ꆗ ñß—?¢ûc*R-@?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À ?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À@@_?¢ûc*R-À"?ü¬€”Ò&Ö?¢ûc*R-ÀI€?ꆗ ñß—?¢ûc*R-À @E@.?¢ûc*R-@@?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-À@bà@S@@"@0@ @$@[@@E@[@?¢ûc*R-À@&@$@?¢ûc*R-@~Ð@@0@*À?¢ûc*R-À@ZÀ@K€@0?ü¬€”Ò&Ö@?ꆗ ñß—@?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-@3?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À@?¢ûc*R-À@*?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-ÀZÀ?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@?¢ûc*R-À @D€@@?¢ûc*R-@??ꆗ ñß—À?¢ûc*R-À@9?ü¬€”Ò&Ö@&@?ꆗ ñß—?¢ûc*R-@3@?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@.?ü¬€”Ò&Ö?¢ûc*R-À@9?ü¬€”Ò&ÖÀ?ꆗ ñß—?¢ûc*R-@0?ꆗ ñß—À?¢ûc*R-@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À*?ü¬€”Ò&Ö?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À"@?¢ûc*R-?ꆗ ñß—?ü¬€”Ò&Ö?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À*@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À"@=?¢ûc*R-ÀJ€?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À@?¢ûc*R-À@@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-?ꆗ ñß—À?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À\@?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-Àd`?ꆗ ñß—?¢ûc*R-Àf`?ꆗ ñß—?¢ûc*R-À1?ꆗ ñß—?¢ûc*R-ÀS@?ꆗ ñß—?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-Àc ?ꆗ ñß—?¢ûc*R-À&?ü¬€”Ò&Ö?¢ûc*R-À@?¢ûc*R-À(?ꆗ ñß—?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À,?ꆗ ñß—?¢ûc*R-À5?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À?ü¬€”Ò&Ö@A€?¢ûc*R-ÀE€?ü¬€”Ò&Ö?¢ûc*R-ÀJ?ꆗ ñß—?¢ûc*R-À@?ü¬€”Ò&Ö@8?¢ûc*R-À@@4?ü¬€”Ò&Ö@*?¢ûc*R-À@*?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@1?¢ûc*R-@1?ꆗ ñß—@8?ꆗ ñß—@,?¢ûc*R-À?ꆗ ñß—@E?¢ûc*R-À?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-ÀfÀ?ü¬€”Ò&Ö?¢ûc*R-@G€?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-@6?¢ûc*R-?ꆗ ñß—À@bà?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@<?¢ûc*R-@?¢ûc*R-@1?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@I?¢ûc*R-À ?ü¬€”Ò&Ö@?¢ûc*R-@?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-À"@5?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À(?ꆗ ñß—?¢ûc*R-À"@*?¢ûc*R-À@B?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-ÀA€?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À6?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-À2?ꆗ ñß—@B@@xP?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-@G€?¢ûc*R-?ꆗ ñß—@@r ?¢ûc*R-@?¢ûc*R-@?¢ûc*R-À@Y@?¢ûc*R-@$?¢ûc*R-@.?¢ûc*R-À?ꆗ ñß—@<?¢ûc*R-@ ?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?ü¬€”Ò&ÖÀ?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@@&?¢ûc*R-@?¢ûc*R-À?ü¬€”Ò&Ö@4?ꆗ ñß—À?¢ûc*R-À@*?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-ÀB€?ꆗ ñß—?¢ûc*R-?ꆗ ñß—À@0?¢ûc*R-À@T€?¢ûc*R-À @A@"@À?ü¬€”Ò&ÖÀ?¢ûc*R-À?ü¬€”Ò&Ö@P€@E@@@*@@,@b@@(@,?¢ûc*R-À@>?ꆗ ñß—@?¢ûc*R-@y@@,@@$@ ?¢ûc*R-À@U@@S@?ꆗ ñß—@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-@k ?¢ûc*R-À@?¢ûc*R-@L€?¢ûc*R-À@@qà?¢ûc*R-À@?¢ûc*R-À@>?¢ûc*R-@?¢ûc*R-@?¢ûc*R-À@6?¢ûc*R-À@?¢ûc*R-ÀI?ꆗ ñß—@?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-@@??¢ûc*R-À@*@X@?ꆗ ñß—?¢ûc*R-À @<?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-À ?ꆗ ñß—?¢ûc*R-À"?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?ü¬€”Ò&Ö@ ?¢ûc*R-@?¢ûc*R-À"@(?¢ûc*R-@?¢ûc*R-À@@C€?¢ûc*R-À @tp@@>?¢ûc*R-@tp@"@9?¢ûc*R-À@p @1@N@@?¢ûc*R-@q@4@"?ü¬€”Ò&Ö?ꆗ ñß—?¢ûc*R-À@s@5?ꆗ ñß—?¢ûc*R-@d @*@(?¢ûc*R-@(?ꆗ ñß—@pÀ@7@G?¢ûc*R-@??ꆗ ñß—@A€?¢ûc*R-À?ü¬€”Ò&Ö?ꆗ ñß—À?¢ûc*R-?ü¬€”Ò&Ö@?¢ûc*R-À@@@?ꆗ ñß—@?¢ûc*R-À?ꆗ ñß—À?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@M€@@(?¢ûc*R-@?¢ûc*R-À@ @@€?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À @>?¢ûc*R-À@?¢ûc*R-À@4?¢ûc*R-À@"?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@[À?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-?ꆗ ñß—?¢ûc*R-ÀH?ꆗ ñß—?ü¬€”Ò&Ö@@@1@]?ꆗ ñß—@?¢ûc*R-À@A€@[@@@P?¢ûc*R-?ꆗ ñß—?ü¬€”Ò&Ö@D?¢ûc*R-@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@&@M€?¢ûc*R-@R@&@a€@@M?¢ûc*R-À@"@p`?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@"?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-@?¢ûc*R-À@N?¢ûc*R-À5?ü¬€”Ò&Ö?¢ûc*R-À@?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@0?¢ûc*R-À(?ꆗ ñß—@?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-À@(?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-ÀX@@?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—À?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À3@ ?¢ûc*R-À ?ꆗ ñß—@R?¢ûc*R-À@gà?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-@?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-ÀWÀ?ꆗ ñß—?¢ûc*R-À(?ꆗ ñß—?¢ûc*R-À,?ꆗ ñß—?¢ûc*R-À*?ꆗ ñß—?¢ûc*R-À ?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-@?¢ûc*R-À@@I€?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-ÀN@(@3?¢ûc*R-?ꆗ ñß—?ü¬€”Ò&Ö?ꆗ ñß—?¢ûc*R-À@>@"?ü¬€”Ò&Ö@?ü¬€”Ò&Ö?¢ûc*R-@@H€?¢ûc*R-@?¢ûc*R-À@3?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-@` @@?ü¬€”Ò&Ö?ꆗ ñß—?¢ûc*R-À@0@@?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-ÀI€?ꆗ ñß—@?¢ûc*R-ÀW@?ꆗ ñß—?¢ûc*R-@?ꆗ ñß—?¢ûc*R-?ꆗ ñß—@@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@0?ꆗ ñß—À?¢ûc*R-À@@?¢ûc*R-À$?ꆗ ñß—@?¢ûc*R-À*?ꆗ ñß—?¢ûc*R-À0?ꆗ ñß—À?¢ûc*R-À&?ü¬€”Ò&Ö?¢ûc*R-À"?ü¬€”Ò&Ö?¢ûc*R-À0?ꆗ ñß—?¢ûc*R-ÀZ@@@$?¢ûc*R-À@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@?¢ûc*R-À&?ꆗ ñß—À?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-ÀA€?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À:?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À"?ü¬€”Ò&Ö?¢ûc*R-Àh @&?¢ûc*R-À@*?¢ûc*R-À"?ꆗ ñß—À?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—À?¢ûc*R-À?ü¬€”Ò&Ö?ꆗ ñß—?¢ûc*R-Àz@?¢ûc*R-À}ð?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-Àeà?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À8?ꆗ ñß—?¢ûc*R-À(?ü¬€”Ò&Ö?¢ûc*R-À@B?¢ûc*R-À1?ꆗ ñß—?¢ûc*R-ÀJ@?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@"@$?¢ûc*R-À?ü¬€”Ò&Ö@"@@$?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@D€?ꆗ ñß—À@?¢ûc*R-À@$@?¢ûc*R-À9?ꆗ ñß—?¢ûc*R-ÀcÀ@?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@?ü¬€”Ò&Ö?¢ûc*R-À@?ꆗ ñß—?¢ûc*R-À@?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À2?ü¬€”Ò&Ö?¢ûc*R-À5?ꆗ ñß—?¢ûc*R-ÀY@?ꆗ ñß—?¢ûc*R-À&?ü¬€”Ò&Ö?¢ûc*R-Àià?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-ÀC€@?¢ûc*R-À@0?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-ÀYÀ?ꆗ ñß—?¢ûc*R-@0?¢ûc*R-À?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@"?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@7?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À @?ü¬€”Ò&Ö?¢ûc*R-À"@?ü¬€”Ò&Ö@?¢ûc*R-À?ꆗ ñß—@TÀ?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À?ꆗ ñß—À?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À:?ꆗ ñß—?¢ûc*R-À"@?¢ûc*R-@3?ꆗ ñß—À?¢ûc*R-À?ꆗ ñß—@Y@?¢ûc*R-À @3@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@<@A?ꆗ ñß—?ü¬€”Ò&ÖÀ@?ꆗ ñß—@6@@ ?¢ûc*R-À@?ꆗ ñß—À?¢ûc*R-@b`@ @À@(?¢ûc*R-À@>@F€?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-ÀB€?ꆗ ñß—?¢ûc*R-À``@?¢ûc*R-À @,?¢ûc*R-?ꆗ ñß—?¢ûc*R-@(?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@(?ü¬€”Ò&Ö@?¢ûc*R-@?¢ûc*R-@1@?¢ûc*R-À@D@?¢ûc*R-À@,?ü¬€”Ò&Ö?¢ûc*R-À@?¢ûc*R-@4@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À"@@ ?¢ûc*R-@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-?ꆗ ñß—?¢ûc*R-À"@(?ꆗ ñß—@?ü¬€”Ò&Ö@?¢ûc*R-À@@I?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À@?¢ûc*R-À&@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@4?¢ûc*R-ÀJ€?ꆗ ñß—?ü¬€”Ò&Ö@@@_À?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@0@SÀ@@L?¢ûc*R-?ꆗ ñß—À@H€?ꆗ ñß—@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@@S€?¢ûc*R-@VÀ?ü¬€”Ò&Ö@``@ @H?¢ûc*R-À@@\À?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-À3?ꆗ ñß—?¢ûc*R-ÀRÀ?ꆗ ñß—?¢ûc*R-ÀU@@"?¢ûc*R-@VÀ?ꆗ ñß—@"?¢ûc*R-À@@I€?¢ûc*R-@4?¢ûc*R-@?¢ûc*R-@H€?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@TÀ?¢ûc*R-@4?ü¬€”Ò&Ö@PÀ?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—@Q?¢ûc*R-@?¢ûc*R-@?¢ûc*R-@?¢ûc*R-À@?¢ûc*R-@?¢ûc*R-À@?ü¬€”Ò&Ö@?¢ûc*R-À@8?¢ûc*R-@?¢ûc*R-À@&?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À@(?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-@\À?¢ûc*R-À@t°?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-@$?¢ûc*R-À@@6?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-À@€?ü¬€”Ò&Ö?¢ûc*R-À@B?¢ûc*R-ÀK€@*?ꆗ ñß—@c ?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-@@€?¢ûc*R-À@Z@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—@S@?¢ûc*R-?ꆗ ñß—?¢ûc*R-@0?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@1?¢ûc*R-@?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö@?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö@?¢ûc*R-@"?¢ûc*R-À@P@?¢ûc*R-@?¢ûc*R-À@ ?¢ûc*R-À?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-À@€@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À @@&?ꆗ ñß—@?ꆗ ñß—?¢ûc*R-À@@(?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-@@@`à?¢ûc*R-@.?¢ûc*R-À@5?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-@[À?ꆗ ñß—@?ꆗ ñß—À?¢ûc*R-À@>@V@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À ?ꆗ ñß—?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@(?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À"?ü¬€”Ò&Ö?¢ûc*R-À*?ü¬€”Ò&Ö?¢ûc*R-ÀTÀ?ꆗ ñß—?¢ûc*R-À&@?¢ûc*R-@I€?¢ûc*R-@K€?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-@A€?¢ûc*R-À?ü¬€”Ò&Ö@z@?¢ûc*R-À@€?ꆗ ñß—?¢ûc*R-À4?ü¬€”Ò&Ö@@@.?ü¬€”Ò&Ö?¢ûc*R-À@@F?ü¬€”Ò&Ö@6?ꆗ ñß—À?¢ûc*R-@?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@ @@€?¢ûc*R-@UÀ?ü¬€”Ò&Ö@A€?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—@<?¢ûc*R-À?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À$?ꆗ ñß—À?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À"@6?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?@;?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À ?ü¬€”Ò&Ö?ꆗ ñß—À@"?ꆗ ñß—@T?¢ûc*R-À@$@H?¢ûc*R-@1?¢ûc*R-À?ꆗ ñß—@@€?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?ü¬€”Ò&Ö@T€?¢ûc*R-@,@@L?¢ûc*R-@5?¢ûc*R-?ꆗ ñß—À@J€?¢ûc*R-À?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-@?¢ûc*R-À@?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@E€?¢ûc*R-À*?ü¬€”Ò&Ö?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-@?¢ûc*R-À@?¢ûc*R-À@?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@C€?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-@"?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@ ?¢ûc*R-À@8?¢ûc*R-@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@"?¢ûc*R-ÀB€?ꆗ ñß—?¢ûc*R-À1?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À*@?¢ûc*R-À@ ?¢ûc*R-À@"?¢ûc*R-À&@?¢ûc*R-À?ü¬€”Ò&Ö?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@0?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?ü¬€”Ò&Ö@9?¢ûc*R-À@Q€?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À"@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À3?ꆗ ñß—?¢ûc*R-À3?ü¬€”Ò&Ö?¢ûc*R-ÀK€@ ?ü¬€”Ò&Ö@`?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-@ ?¢ûc*R-À?ꆗ ñß—@M?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@K?¢ûc*R-@?¢ûc*R-@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@?¢ûc*R-À @(@?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-À@=@;?ꆗ ñß—@?¢ûc*R-?ꆗ ñß—@@J?ꆗ ñß—@"?¢ûc*R-À?ꆗ ñß—À?¢ûc*R-@J?¢ûc*R-?ꆗ ñß—?ü¬€”Ò&Ö@$?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À0?ꆗ ñß—?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-?ꆗ ñß—À?¢ûc*R-À.?ü¬€”Ò&Ö?¢ûc*R-À"?ü¬€”Ò&Ö?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@?ꆗ ñß—@?¢ûc*R-@?¢ûc*R-À@?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À"?ü¬€”Ò&Ö?¢ûc*R-À*@?¢ûc*R-ÀWÀ@?¢ûc*R-@(?¢ûc*R-@(?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@A€?¢ûc*R-À @D?ü¬€”Ò&ÖÀ?¢ûc*R-@B€?ü¬€”Ò&Ö@?¢ûc*R-À@<@@?ꆗ ñß—?¢ûc*R-À@:@?¢ûc*R-À@G€@.?¢ûc*R-À@5@&@?ꆗ ñß—@?¢ûc*R-@@€@$?ü¬€”Ò&Ö?ꆗ ñß—@?¢ûc*R-@?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-À@@>?ü¬€”Ò&Ö@8?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-@C€?ꆗ ñß—@(?¢ûc*R-@?¢ûc*R-À@@<?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-@6?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@(?¢ûc*R-À@?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@6?¢ûc*R-ÀJ@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-@$@G?¢ûc*R-À@<@A€@@.?ü¬€”Ò&Ö?¢ûc*R-@@4?¢ûc*R-À@@;?¢ûc*R-@@@@@€@@=?¢ûc*R-À@@D€?¢ûc*R-À?ꆗ ñß—À?¢ûc*R-ÀB€?ꆗ ñß—?¢ûc*R-À*?ꆗ ñß—?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À ?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À&@?¢ûc*R-À[@?ꆗ ñß—?¢ûc*R-À*@?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À1?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-@>?¢ûc*R-À@T€?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-ÀY@?ꆗ ñß—?¢ûc*R-ÀI€?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-ÀQ€?ꆗ ñß—À?¢ûc*R-@?¢ûc*R-À@À?ꆗ ñß—À?¢ûc*R-À@(?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-@1?ꆗ ñß—?ü¬€”Ò&Ö?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?ꆗ ñß—?¢ûc*R-ÀL€?ꆗ ñß—?¢ûc*R-ÀT€?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@2?¢ûc*R-À"@&?ꆗ ñß—?¢ûc*R-À@?ü¬€”Ò&Ö?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À@?ꆗ ñß—?¢ûc*R-À5?ꆗ ñß—?¢ûc*R-ÀX?ü¬€”Ò&Ö@?ꆗ ñß—?ü¬€”Ò&Ö@ @Q?¢ûc*R-À@9@J€@@2?¢ûc*R-À@@D€?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö@D?¢ûc*R-@F?ü¬€”Ò&Ö@B@@&?¢ûc*R-À?ü¬€”Ò&Ö@A?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À2?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-ÀG€?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À"?ü¬€”Ò&Ö?¢ûc*R-Àt?ꆗ ñß—?¢ûc*R-Àu ?ꆗ ñß—À?¢ûc*R-À?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-À@À?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À[@@?¢ûc*R-ÀA€?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-Àh ?ꆗ ñß—?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@5?¢ûc*R-À"@?¢ûc*R-@?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—@1?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À&@1?¢ûc*R-À?ꆗ ñß—@?ꆗ ñß—@@h ?¢ûc*R-À?ꆗ ñß—@}Ð?¢ûc*R-@?¢ûc*R-À@?ꆗ ñß—@2?¢ûc*R-À@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@@?¢ûc*R-À?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-ÀJ@?ü¬€”Ò&Ö?ꆗ ñß—@3@@qÀ?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@@3?ü¬€”Ò&Ö@H?¢ûc*R-À@@f?ꆗ ñß—@?¢ûc*R-À?ꆗ ñß—@X@?¢ûc*R-@A€@@ ?¢ûc*R-@?¢ûc*R-À@@.?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?ꆗ ñß—?¢ûc*R-À@.@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@:@0?ü¬€”Ò&Ö@$?¢ûc*R-À?ꆗ ñß—@T?¢ûc*R-@$?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-@T@?ü¬€”Ò&Ö@@"À?¢ûc*R-À?ꆗ ñß—?¢ûc*R-ÀC?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À @,@?¢ûc*R-?ꆗ ñß—À?¢ûc*R-À@C€@=?¢ûc*R-@?ꆗ ñß—À?ü¬€”Ò&Ö@M@@4?¢ûc*R-À@1?ü¬€”Ò&ÖÀ?¢ûc*R-@PÀ?ü¬€”Ò&Ö@.@@$?¢ûc*R-À@"À?ü¬€”Ò&ÖÀ?ꆗ ñß—À?¢ûc*R-À?ꆗ ñß—?¢ûc*R-@ ?¢ûc*R-À&@?¢ûc*R-À@?¢ûc*R-?ꆗ ñß—?¢ûc*R-À4?ꆗ ñß—@$?¢ûc*R-ÀI?ꆗ ñß—?¢ûc*R-À?@?ꆗ ñß—?¢ûc*R-?ü¬€”Ò&Ö@@(?¢ûc*R-À@0@<?¢ûc*R-@E?ꆗ ñß—@"?ü¬€”Ò&Ö@[€?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@R@?ü¬€”Ò&Ö@>@ @2?¢ûc*R-À?ꆗ ñß—@O€?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@D?¢ûc*R-@?¢ûc*R-@E€@@?¢ûc*R-À@D€@@?ü¬€”Ò&Ö?¢ûc*R-À@@€@@@?¢ûc*R-À@Q@@?¢ûc*R-À@=@@?ꆗ ñß—@?¢ûc*R-@@€@"?ü¬€”Ò&Ö?¢ûc*R-@?ꆗ ñß—@ ?¢ûc*R-?ꆗ ñß—?¢ûc*R-@$@?ꆗ ñß—@ @&@B€?¢ûc*R-À@@€@T@?ü¬€”Ò&Ö@V?¢ûc*R-À?ꆗ ñß—@"?ꆗ ñß—@"?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@@(?¢ûc*R-@U€@@Q@*@<?¢ûc*R-À@@[€?¢ûc*R-À?ꆗ ñß—@?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-@J€?¢ûc*R-À?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@2?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@*?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À@gÀ?¢ûc*R-À@?¢ûc*R-À.?ꆗ ñß—?¢ûc*R-À@@@?¢ûc*R-@@0@R€?¢ûc*R-À@P@@a ?ü¬€”Ò&Ö@A?¢ûc*R-À?ü¬€”Ò&Ö@A€?ꆗ ñß—À?¢ûc*R-?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—@7?¢ûc*R-@R@@,@U@@I€?¢ûc*R-À?ꆗ ñß—@P€?¢ûc*R-À?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-ÀJ?ꆗ ñß—?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-?ꆗ ñß—?¢ûc*R-À4?ꆗ ñß—?ü¬€”Ò&Ö?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-ÀD€?ꆗ ñß—?¢ûc*R-ÀZÀ?ꆗ ñß—?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-ÀN€?ü¬€”Ò&Ö?¢ûc*R-ÀL€?ü¬€”Ò&Ö?¢ûc*R-ÀK€?ü¬€”Ò&Ö?¢ûc*R-ÀK€?ꆗ ñß—?¢ûc*R-À@&?¢ûc*R-À6?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—À?¢ûc*R-À"@?¢ûc*R-Àc`?ꆗ ñß—?¢ûc*R-À3@"?¢ûc*R-À2?ꆗ ñß—@?¢ûc*R-ÀL€?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-ÀXÀ@?ü¬€”Ò&Ö@4?¢ûc*R-À@(?¢ûc*R-@$?¢ûc*R-À?ü¬€”Ò&Ö@?¢ûc*R-À@?¢ûc*R-@?¢ûc*R-@0?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—@:?¢ûc*R-À?ꆗ ñß—À?¢ûc*R-Àqp?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À``?ꆗ ñß—?¢ûc*R-À1?ꆗ ñß—?¢ûc*R-À3?ꆗ ñß—?¢ûc*R-À@?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-À;?ꆗ ñß—?¢ûc*R-ÀG€?ꆗ ñß—?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À ?ꆗ ñß—@?¢ûc*R-ÀL€?ꆗ ñß—?¢ûc*R-ÀX@@?¢ûc*R-À*?ꆗ ñß—?¢ûc*R-À@"?¢ûc*R-À&?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-ÀE€?ꆗ ñß—?¢ûc*R-À*?ꆗ ñß—?¢ûc*R-À"?ü¬€”Ò&Ö?¢ûc*R-ÀE€?ꆗ ñß—?¢ûc*R-ÀK€?ꆗ ñß—À@$?¢ûc*R-?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—@(?¢ûc*R-@?¢ûc*R-À@"?¢ûc*R-?ꆗ ñß—?¢ûc*R-À@*?¢ûc*R-?ꆗ ñß—?¢ûc*R-@(?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—@(?¢ûc*R-Àl ?ꆗ ñß—?¢ûc*R-À3?ꆗ ñß—?¢ûc*R-ÀC€@?¢ûc*R-À@,?¢ûc*R-À1?ꆗ ñß—?¢ûc*R-ÀE€@?¢ûc*R-À[@?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-ÀO€?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-@?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À$?ꆗ ñß—@?¢ûc*R-Àc ?ꆗ ñß—?¢ûc*R-@?¢ûc*R-@?¢ûc*R-À1@B€?¢ûc*R-À(?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À?ü¬€”Ò&Ö?¢ûc*R-À*?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À"?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À7?ü¬€”Ò&Ö?¢ûc*R-ÀZÀ?ꆗ ñß—?¢ûc*R-À?ꆗ ñß—?¢ûc*R-À*@"?¢ûc*R-ÀJ€?ꆗ ñß—?¢ûc*R-À†P?ꆗ ñß—?¢ûc*R-À*?ꆗ ñß—?¢ûc*R-ÀB€?ꆗ ñß—?¢ûc*R-ÀuPfestival/lib/sec.B.hept.ngrambin0000644000175000017500000000104107262605650015161 0ustar awbawbNgramBin_2 7 B NB B NB ?­pΤτÅÀ>?ã>wâN'×?­pΤτÅÀ(@?­pΤτÅ@(@B€@[?­pΤτÅÀ@,@]@@‚p?­pΤτÅÀ@B@H?­pΤτÅ@’|@0@¡8?­pΤτÅÀ.?ã>wâN'×?­pΤτÅÀ@@(@b ?­pΤτÅÀ@,@†?­pΤτÅ@Ø@’x@©?­pΤτÅÀ?ã>wâN'×?­pΤτÅÀ@@c ?­pΤτÅ@†(@D@°´?­pΤτÅÀ?ã>wâN'×?­pΤτÅ@aà@‚€@²w?­pΤτÅ?ã>wâN'×@ZÀ@°µ?ã>wâN'×@©@¡8@®¨festival/lib/init.scm0000644000175000017500000001342110077563027013224 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Initialisation file -- loaded before anything else ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Basic siod library (need this before load_library or require works) (load (path-append libdir "siod.scm")) (defvar home-directory (or (getenv "HOME") "/") "home-directory Place looked at for .festivalrc etc.") ;;; User startup initialization, can be used to override load-path ;;; to allow alternate basic modules to be loaded. (if (probe_file (path-append home-directory ".siodvarsrc")) (load (path-append home-directory ".siodvarsrc"))) (if (probe_file (path-append home-directory ".festivalvarsrc")) (load (path-append home-directory ".festivalvarsrc"))) ;;; A chance to set various variables to a local setting e.g. ;;; lexdir, voices_dir audio etc etc. (if (probe_file (path-append libdir "sitevars.scm")) (load (path-append libdir "sitevars.scm"))) ;;; CSTR siod extensions (require 'cstr) ;;; Festival specific definitions (require 'festival) ;;; Dealing with module descriptions (require 'module_description) ;;; Web related definitions (require 'web) ;;; Utterance types and support (require 'synthesis) ;;; Some default parameters (Parameter.def 'Wavefiletype 'riff) ;;; Set default audio method (cond ((member 'nas *modules*) (Parameter.def 'Audio_Method 'netaudio)) ((member 'esd *modules*) (Parameter.def 'Audio_Method 'esdaudio)) ((member 'sun16audio *modules*) (Parameter.def 'Audio_Method 'sun16audio)) ((member 'freebsd16audio *modules*) (Parameter.def 'Audio_Method 'freebsd16audio)) ((member 'linux16audio *modules*) (Parameter.def 'Audio_Method 'linux16audio)) ((member 'irixaudio *modules*) (Parameter.def 'Audio_Method 'irixaudio)) ((member 'macosxaudio *modules*) (Parameter.def 'Audio_Method 'macosxaudio)) ((member 'win32audio *modules*) (Parameter.def 'Audio_Method 'win32audio)) ((member 'os2audio *modules*) (Parameter.def 'Audio_Method 'os2audio)) ((member 'mplayeraudio *modules*) (Parameter.def 'Audio_Method 'mplayeraudio)) (t ;; can't find direct support so guess that /dev/audio for 8k ulaw exists (Parameter.def 'Audio_Method 'sunaudio))) ;;; If you have an external program to play audio add its definition ;;; in siteinit.scm ;;; The audio spooler doesn't work under Windows so redefine audio_mode (if (member 'mplayeraudio *modules*) (define (audio_mode param) param) ) ;;; Intonation (require 'intonation) ;;; Duration (require 'duration) ;;; A large lexicon (require 'lexicons) (require 'pauses) ;;; Part of speech prediction (require 'pos) ;;; Phrasing (dependent on pos) (require 'phrase) ;;; POstlexical rules (require 'postlex) ;;; Different voices (require 'voices) ;; sets voice_default (require 'languages) ;;; Some higher level functions (require 'token) (require 'tts) ;;; ;;; Local site initialization, if the file exists load it ;;; (if (probe_file (path-append libdir "siteinit.scm")) (load (path-append libdir "siteinit.scm"))) ;;; User initialization, if a user has a personal customization ;;; file loaded it (if (probe_file (path-append home-directory ".siodrc")) (load (path-append home-directory ".siodrc"))) (if (probe_file (path-append home-directory ".festivalrc")) (load (path-append home-directory ".festivalrc"))) ;;; Default voice (have to do something cute so autoloads still work) (eval (list voice_default)) (provide 'init) festival/lib/synthesis.scm0000644000175000017500000003475307262616703014327 0ustar awbawb ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; Author: Richard Caley (rjc@cstr.ed.ac.uk) ;; ;; Date: Fri Aug 15 1997 ;; ;; ------------------------------------------------------------------- ;; ;; New synthesis mainline. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; Hooks to add to the synthesis process. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (defvar default_before_synth_hooks nil "default_before_synth_hooks The default list of functions to be run on all synthesized utterances before synthesis starts.") (defvar before_synth_hooks default_before_synth_hooks "before_synth_hooks List of functions to be run on synthesised utterances before synthesis starts.") (defvar default_after_analysis_hooks nil "default_after_analysis_hooks The default list of functions to be run on all synthesized utterances after analysis but before synthesis.") (defvar after_analysis_hooks default_after_analysis_hooks "after_analysis_hooks List of functions to be applied after analysis and before synthesis.") (defvar default_after_synth_hooks nil "default_after_synth_hooks The default list of functions to be run on all synthesized utterances after Wave_Synth. This will normally be nil but if for some reason you need to change the gain or rescale *all* waveforms you could set the function here, in your siteinit.scm.") (defvar after_synth_hooks default_after_synth_hooks "after_synth_hooks List of functions to be applied after all synthesis modules have been applied. This is primarily designed to allow waveform manipulation, particularly resampling and volume changes.") (defvar default_access_strategy 'ondemand "default_access_strategy How to access units from databases.") ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; Macro to define utterance types. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (defmac (defUttType form) (list 'defUttType_real (list 'quote (cadr form)) (list 'quote (cddr form)))) (defvar UttTypes nil "UttTypes List of types and functions used by the utt.synth function to call appropriate methods.") (define (defUttType_real type form) "(defUttType TYPE . BODY) Define a new utterance type. TYPE is an atomic type that is specified as the first argument to the function Utterance. BODY is evaluated with argument utt, when utt.synth is called with an utterance of type TYPE. You almost always require the function Initialize first. [see Utterance types]" ;;; Yes I am cheating a bit with the macro/function name. ;;; should check about redefining and the syntax of the forms (set! UttTypes (cons (cons type form) UttTypes)) type) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; Macro to define synthesis types. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (defmac (defSynthType form) (list 'defSynthType_real (list 'quote (cadr form)) (list 'quote (cddr form)))) (defvar SynthTypes nil "SynthTypes List of synthesis types and functions used by the utt.synth function to call appropriate methods for wave synthesis.") (define (defSynthType_real type form) "(defSynthType TYPE . BODY) Define a new wave synthesis type. TYPE is an atomic type that identifies the type of synthesis. BODY is evaluated with argument utt, when utt.synth is called with an utterance of type TYPE. [see Utterance types]" (set! SynthTypes (cons (cons type form) SynthTypes)) type) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Some actual Utterance type definitions ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (defUttType Words (Initialize utt) (POS utt) (Phrasify utt) (Word utt) (Pauses utt) (Intonation utt) (PostLex utt) (Duration utt) (Int_Targets utt) (Wave_Synth utt) ) (defUttType Text (Initialize utt) (Text utt) (Token_POS utt) (Token utt) (POS utt) (Phrasify utt) (Word utt) (Pauses utt) (Intonation utt) (PostLex utt) (Duration utt) (Int_Targets utt) (Wave_Synth utt) ) (defUttType Tokens ;; This is used in tts_file, Tokens will be preloaded (Token_POS utt) ;; when utt.synth is called (Token utt) (POS utt) (Phrasify utt) (Word utt) (Pauses utt) (Intonation utt) (PostLex utt) (Duration utt) (Int_Targets utt) (Wave_Synth utt) ) (defUttType Concept ;; rather gradious name for when information has (POS utt) ;; been preloaded (probably XML) to give a word (Phrasify utt) ;; relation (SOLE uses this) (Word utt) (Pauses utt) (Intonation utt) (PostLex utt) (Duration utt) (Int_Targets utt) (Wave_Synth utt) ) (defUttType Phrase (Initialize utt) (Token_POS utt) (Token utt) (POS utt) (Phrasify utt) (Word utt) (Pauses utt) (Intonation utt) (PostLex utt) (Duration utt) (Int_Targets utt) (Wave_Synth utt) ) (defUttType Segments (Initialize utt) (Wave_Synth utt) ) (defUttType Phones (Initialize utt) (Fixed_Prosody utt) (Wave_Synth utt) ) (defUttType SegF0 (Wave_Synth utt) ) (defUttType Wave (Initialize utt)) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; And some synthesis types. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (defSynthType Taylor (Taylor_Synthesize utt) ) (defSynthType UniSyn (defvar UniSyn_module_hooks nil) (Param.def "unisyn.window_name" "hanning") (Param.def "unisyn.window_factor" 1.0) (Parameter.def 'us_sigpr 'lpc) (apply_hooks UniSyn_module_hooks utt) ;; for processing of diphone names (us_get_diphones utt) (us_unit_concat utt) (if (not (member 'f0 (utt.relationnames utt))) (targets_to_f0 utt)) ;; temporary fix (if (utt.relation.last utt 'Segment) (set! pm_end (+ (item.feat (utt.relation.last utt 'Segment) "end") 0.02)) (set! pm_end 0.02)) (us_f0_to_pitchmarks utt 'f0 'TargetCoef pm_end) (us_mapping utt 'segment_single) (cond ((string-equal "td_psola" (Parameter.get 'us_sigpr)) ;; Not in standard distribution, so has to be separate function (us_tdpsola_synthesis utt 'analysis_period)) (t ;; All the rest (us_generate_wave utt (Parameter.get 'us_sigpr) 'analysis_period))) ) (defSynthType None ;; do nothing utt ) (defSynthType Standard (print "synth method: Standard") (let ((select (Parameter.get 'SelectionMethod))) (if select (progn (print "select") (apply select (list utt)) ) ) ) (let ((join (Parameter.get 'JoiningMethod))) (if join (progn (print "join") (apply join (list utt)) ) ) ) (let ((impose (Parameter.get 'ImposeMethod))) (if impose (progn (print "impose") (apply impose (list utt)) ) ) ) (let ((power (Parameter.get 'PowerSmoothMethod))) (if power (progn (print "power") (apply power (list utt)) ) ) ) (let ((wavesynthesis (Parameter.get 'WaveSynthesisMethod))) (if wavesynthesis (progn (print "synthesis") (apply wavesynthesis (list utt)) ) ) ) ) (defSynthType Minimal (print "synth method: Minimal") (let ((select (Parameter.get 'SelectionMethod))) (if select (progn (print "select") (apply select (list utt)) ) ) ) (let ((wavesynthesis (Parameter.get 'WaveSynthesisMethod))) (if wavesynthesis (progn (print "synthesis") (apply wavesynthesis (list utt "Unit" "Join" "Wave")) ) ) ) ) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; Finally the actual driver function. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (utt.synth utt) "(utt.synth UTT) The main synthesis function. Given UTT it will apply the functions specified for UTT's type, as defined with deffUttType and then those demanded by the voice. After modules have been applied synth_hooks are applied to allow extra manipulation. [see Utterance types]" (apply_hooks before_synth_hooks utt) (let ((type (utt.type utt))) (let ((definition (assoc type UttTypes))) (if (null? definition) (error "Unknown utterance type" type) (let ((body (eval (cons 'lambda (cons '(utt) (cdr definition)))))) (body utt))))) (apply_hooks after_synth_hooks utt) utt) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; And a couple of utility expressions. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (SayText text) "(SayText TEXT) TEXT, a string, is rendered as speech." (utt.play (utt.synth (eval (list 'Utterance 'Text text))))) (define (SynthText text) "(SynthText TEXT) TEXT, a string, is rendered as speech." (utt.synth (eval (list 'Utterance 'Text text)))) (define (SayPhones phones) "(SayPhones PHONES) PHONES is a list of phonemes. This uses the Phones type utterance to synthesize and play the given phones. Fixed duration specified in FP_duration and fixed monotone duration (FP_F0) are used to generate prosody." (utt.play (utt.synth (eval (list 'Utterance 'Phones phones))))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; This is the standard synthesis function. The Wave Synthesis may be ;; ;; more than a simple module ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (Wave_Synth utt) "(Wave_Synth UTT) Generate waveform from information in UTT, at least a Segment stream must exist. The actual form of synthesis used depends on the Parameter Synth_Method. If it is a function that is applied. If it is atom it should be a SynthType as defined by defSynthType [see Utterance types]" (apply_hooks after_analysis_hooks utt) (let ((method_val (Parameter.get 'Synth_Method))) (cond ((null method_val) (error "Undefined Synth_Method")) ((and (symbol? method_val) (symbol-bound? method_val)) ;; Wish there was a function? (apply (symbol-value method_val) (list utt))) ((member (typeof method_val) '(subr closure)) (apply method_val (list utt))) (t ;; its a defined synthesis type (let ((synthesis_modules (assoc_string method_val SynthTypes))) (if (null? synthesis_modules) (error (format nil "Undefined SynthType %s\n" method_val)) (let ((body (eval (cons 'lambda (cons '(utt) (cdr synthesis_modules)))))) (body utt))))))) utt) (provide 'synthesis) festival/lib/module_description.scm0000644000175000017500000001071507262605650016155 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Handle module descriptions. ;;; (defvar *module-descriptions* nil "*module-descriptions* An association list recording the description objects for proclaimed modules.") (define (set_module_description mod desc) "(set_module_description MOD DESC) Set the description for the module named MOD." (let ((entry (assoc mod *module-descriptions*))) (if entry (set-cdr! entry (cons desc nil)) (set! *module-descriptions* (cons (cons mod (cons desc nil)) *module-descriptions*)) ) ) ) (define (module_description mod) "(module_description MOD) Returns the description record of the module named by symbol MOD" (let ((entry (assoc mod *module-descriptions*))) (if entry (car (cdr entry)) nil ) ) ) (defmac (proclaim form) "(proclaim NAME &opt DESCRIPTION...) Anounce the availability of a module NAME. DESCRIPTION is a description in a fixed format." (let ((name (car (cdr form))) (description (cdr form)) ) (list 'proclaim-real (list 'quote name) (list 'quote description)) ) ) (define (proclaim-real name description) (set! *modules* (cons name *modules*)) ; (if description ; (set_module_description name (create_module_description description)) ; ) ) (define (describe_module mod) "(describe_module MOD) Describe the module named by the symbol MOD." (let ((entry (module_description mod))) (format t "---------------------\n") (if entry (print_module_description entry) (format t "No description for %l\n" mod) ) (format t "---------------------\n") ) ) (define (describe_all_modules) "(describe_all_modules) Print descriptions of all proclaimed modules" (format t "---------------------\n") (let ((p *module-descriptions*)) (while p (print_module_description (car (cdr (car p)))) (format t "---------------------\n") (set! p (cdr p)) ) ) ) (proclaim module_description 1.1 "CSTR" "Richard Caley " ( "Handle module descriptions from C++ and from Scheme." ) ) (provide 'module_description) festival/lib/lexicons.scm0000644000175000017500000002504711223441730014102 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Definition of various lexicons ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; If there exists a sudirectory of the lib-path called dicts then that ;;; is used as the lexicon directory by default. If it doesn't exist ;;; we set lexdir to the directory in CSTR where our lexicons are. ;;; In non-CSTR installations where lexicons are not in lib/dicts, ;;; you should set lexdir in sitevars.scm (defvar lexdir (if (probe_file (path-append libdir "dicts")) (path-append libdir "dicts/") ;; else we'll guess we're in the CSTR filespace (path-as-directory "/projects/festival/lib/dicts/")) "lexdir The directory where the lexicon(s) are, by default.") (require 'pos) ;; for part of speech mapping (define (setup_cstr_lex) "(setup_cstr_lexicon) Define and setup the CSTR lexicon. The CSTR lexicon consists of about 25,000 entries in the mrpa phone set. A large number of specific local entries are also added to the addenda." (if (not (member_string "mrpa" (lex.list))) (begin (lex.create "mrpa") (lex.set.compile.file (path-append lexdir "cstrlex.out")) (lex.set.phoneset "mrpa") (lex.set.lts.method 'lts_rules) (lex.set.lts.ruleset 'nrl) (lex.set.pos.map english_pos_map_wp39_to_wp20) (mrpa_addenda) (lex.add.entry '("previous" nil (((p r ii) 1) ((v ii) 0) ((@ s) 0)))) (lex.add.entry '("audio" () (((oo d) 1) ((ii) 0) ((ou) 0)))) (lex.add.entry '("modules" () (((m o d) 1) ((uu l s) 0)))) ))) (define (setup_oald_lex) "(setup_oald_lexicon) Define and setup the CUVOALD lexicon. This is derived from the Computer Users Version of the Oxford Advanced Learners' Dictionary of Current English. This version includes a trained set of letter to sound rules which have also been used to reduce the actual lexicon size by over half, for those entries that the lts model gets exactly the same." (if (not (member_string "oald" (lex.list))) (load (path-append lexdir "oald/oaldlex.scm")))) (define (setup_cmu_lex) "(setup_cmu_lex) Lexicon derived from the CMU lexicon (cmudict-0.4), around 100,000 entries, in the radio phoneset (sort of darpa-like). Includes letter to sound rule model trained from this data, and uses the lexical stress predictor from OALD." (if (not (member_string "cmu" (lex.list))) (load (path-append lexdir "cmu/cmulex.scm")))) (define (setup_cmumt_lex) "(setup_cmumt_lex) Lexicon derived from the CMU lexicon (cmudict-0.4), around 100,000 entries, in the radio phoneset (sort of darpa-like). Includes letter to sound rule model trained from this data, and uses the lexical stress predictor from OALD." (if (not (member_string "cmumt" (lex.list))) (load (path-append lexdir "cmu_mt/cmumtlex.scm")))) (define (setup_cmu6_lex) "(setup_cmu6_lex) Lexicon derived from the CMU lexicon (cmudict-0.6), around 100,000 entries, in the radio phoneset (sort of darpa-like). Includes letter to sound rule model trained from this data, the format of this lexicon is suitable for the UniSyn metrical phonology modules. That is the entries are not syllabified," (if (not (member_string "cmu6" (lex.list))) (load (path-append lexdir "cmu6/cmu6lex.scm")))) (define (setup_moby_lex) "(setup_moby_lexicon) Define and setup the MOBY lexicon. This is derived from the public domain version of the Moby (TM) Pronunciator II lexicon. It can be converted automatically to British English mrpa phoneset which of course is sub-optimal. It contains around 120,000 entries and has part of speech information for homographs." (if (not (member_string "moby" (lex.list))) (begin (lex.create "moby") ; (lex.set.compile.file (path-append lexdir "mobylex.out")) (lex.set.compile.file "/home/awb/src/mobypron/mobylex.out") (lex.set.phoneset "mrpa") (lex.set.lts.method 'lts_rules) (lex.set.lts.ruleset 'nrl) (lex.set.pos.map english_pos_map_wp39_to_wp20) (lex.add.entry '("a" dt (((@) 0)))) (lex.add.entry '("the" dt (((dh @) 0)))) (lex.add.entry '("taylor" n (((t ei) 1) ((l @) 0)))) (lex.add.entry '("who" prp ((( h uu ) 0)))) (mrpa_addenda)))) (define (setup_beep_lex) "(setup_beep_lex) Lexicon derived from the British English Example Pronunciation dictionary (BEEP) from Tony Robinson ajr@eng.cam.ac.uk. Around 160,000 entries." (if (not (member_string "beep" (lex.list))) (begin (lex.create "beep") (lex.set.compile.file (path-append lexdir "beep_lex.out")) (lex.set.phoneset "mrpa") (lex.set.lts.method 'lts_rules) (lex.set.lts.ruleset 'nrl) (lex.set.pos.map english_pos_map_wp39_to_wp20) (lex.add.entry '("taylor" nil (((t ei) 1) ((l @) 0)))) (mrpa_addenda)))) ;;; The nrl letter to sound rules produce mrpa phone set so we need ;;; to do some fancy things to make them work for American English (define (f2b_lts word features) "(f2b_lts WORD FEATURES) Letter to sound rule system for f2b (American English), uses the NRL LTS ruleset and maps the result to the radio phone set." '("unknown" nil (((ah n) 0) ((n ow n) 1))) ) ;;; A CART tree for predicting lexical stress for strings of phones ;;; generated by the LTS models. This was actually trained from ;;; OALD as that's the only lexicon with stress and part of speech information ;;; It trained in a phoneset independent way and may be used be either ;;; OALD or CMU models (and probably MOBY and OGI lex too). ;;; On held out data it gets ;;; 07390 378 7768 [7390/7768] 95.134 ;;; 1 512 8207 8719 [8207/8719] 94.128 ;;; 7902 8585 ;;; total 16487 correct 15597.000 94.602% ;;; (set! english_stress_tree '((sylpos < 1.7) ((1)) ((ph_vlng is a) ((0)) ((ph_vheight is 1) ((num2end < 1.5) ((ph_vfront is 1) ((ph_vlng is s) ((0)) ((pos is v) ((1)) ((0)))) ((pos is n) ((0)) ((sylpos < 2.2) ((1)) ((0))))) ((ph_vlng is l) ((1)) ((ph_vfront is 1) ((num2end < 2.4) ((0)) ((pos is a) ((num2end < 3.3) ((sylpos < 2.3) ((1)) ((0))) ((0))) ((sylpos < 3.2) ((num2end < 3.3) ((0)) ((pos is v) ((1)) ((0)))) ((0))))) ((0))))) ((num2end < 1.5) ((pos is n) ((0)) ((sylpos < 2.4) ((pos is v) ((1)) ((ph_vlng is d) ((ph_vheight is 2) ((ph_vfront is 1) ((1)) ((0))) ((0))) ((1)))) ((ph_vlng is d) ((sylpos < 3.3) ((pos is v) ((ph_vheight is 2) ((ph_vfront is 1) ((0)) ((1))) ((0))) ((0))) ((0))) ((ph_vheight is 2) ((1)) ((ph_vrnd is +) ((1)) ((ph_vlng is l) ((0)) ((1)))))))) ((ph_vlng is d) ((pos is v) ((sylpos < 2.4) ((1)) ((0))) ((ph_vfront is 2) ((pos is n) ((num2end < 2.4) ((ph_vrnd is +) ((0)) ((sylpos < 2.2) ((1)) ((ph_vheight is 2) ((1)) ((0))))) ((sylpos < 2.4) ((ph_vheight is 2) ((0)) ((1))) ((0)))) ((1))) ((ph_vheight is 2) ((1)) ((ph_vfront is 1) ((0)) ((1)))))) ((pos is n) ((num2end < 2.4) ((ph_vfront is 3) ((sylpos < 2.3) ((1)) ((ph_vlng is l) ((1)) ((0)))) ((1))) ((1))) ((1))))))))) (define (lex_user_unknown_word word feats) "(lex_user_unknown_word WORD FEATS) Function called by lexicon when 'function type letter to sound rules is defined. It is the user's responsibility to defined this function themselves when they want to deal with unknown words themselves." (error "lex_user_unknown_word: has not been defined by user")) (define (Word utt) "(Word utt) Construct (synthesis specific) syllable/segments from Word relation using current lexicon and specific module." (let ((rval (apply_method 'Word_Method utt))) (cond (rval rval) ;; new style (t (Classic_Word utt))))) (define (find_oovs vocab oovs) (let ((fd (fopen vocab "r")) (ofd (fopen oovs "w")) (e 0) (oov 0) (entry)) (while (not (equal? (set! entry (readfp fd)) (eof-val))) (set! e (+ 1 e)) (if (not (lex.lookup_all entry)) (begin (set! oov (+ 1 oov)) (format ofd "%l\n" (lex.lookup entry nil)))) ) (format t ";; %d words %d oov %2.2f oov_rate\n" e oov (/ (* oov 100.0) e)) ) ) (provide 'lexicons) festival/lib/festival.scm0000644000175000017500000005730612444573540014111 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; General Festival Scheme specific functions ;;; Including definitions of various standard variables. ;; will be set automatically on start-up (defvar festival_version "unknown" "festival_version A string containing the current version number of the system.") ;; will be set automatically on start-up (defvar festival_version_number '(x x x) "festival_version_number A list of major, minor and subminor version numbers of the current system. e.g. (1 0 12).") (define (apply_method method utt) "(apply_method METHOD UTT) Apply the appropriate function to utt defined in parameter." (let ((method_val (Parameter.get method))) (cond ((null method_val) nil) ;; should be an error, but I'll let you off at present ((and (symbol? method_val) (symbol-bound? method_val)) (apply (symbol-value method_val) (list utt))) ((member (typeof method_val) '(subr closure)) (apply method_val (list utt))) (t ;; again is probably an error nil)))) (define (require_module l) "(require_module l) Check that certain compile-time modules are included in this installation. l may be a single atom or list of atoms. Each item in l must appear in *modules* otherwise an error is throw." (if (consp l) (mapcar require_module l) (if (not (member_string l *modules*)) (error (format nil "module %s required, but not compiled in this installation\n" l)))) t) ;;; Feature Function Functions (define (utt.features utt relname func_list) "(utt.features UTT RELATIONNAME FUNCLIST) Get vectors of feature values for each item in RELATIONNAME in UTT. [see Features]" (mapcar (lambda (s) (mapcar (lambda (f) (item.feat s f)) func_list)) (utt.relation.items utt relname))) (define (utt.type utt) "(utt.type UTT) Returns the type of UTT." (intern (utt.feat utt 'type))) (define (utt.save.segs utt filename) "(utt.save.segs UTT FILE) Save segments of UTT in a FILE in xlabel format." (let ((fd (fopen filename "w"))) (format fd "#\n") (mapcar (lambda (info) (format fd "%2.4f 100 %s\n" (car info) (car (cdr info)))) (utt.features utt 'Segment '(segment_end name))) (fclose fd) utt)) (define (utt.save.words utt filename) "(utt.save.words UTT FILE) Save words of UTT in a FILE in xlabel format." (let ((fd (fopen filename "w"))) (format fd "#\n") (mapcar (lambda (info) (format fd "%2.4f 100 %s\n" (car info) (car (cdr info)))) (utt.features utt 'Word '(word_end name))) (fclose fd) utt)) (define (utt.resynth labfile f0file) "(utt.resynth LABFILE F0FILE) Resynthesize an utterance from a label file and F0 file (in any format supported by the Speech Tool Library). This loads, synthesizes and plays the utterance." (let (u f0 f0_item) (set! u (Utterance SegF0)) ; need some u to start with (utt.relation.load u 'Segment labfile) (utt.relation.create u 'f0) (set! f0 (track.load f0file)) (set! f0_item (utt.relation.append u 'f0)) (item.set_feat f0_item "name" "f0") (item.set_feat f0_item "f0" f0) ;; emulabel may have flipped pau to H# (mapcar (lambda (s) (cond ((string-matches (item.name s) "[hH]#") (item.set_feat s "name" "pau")) ((string-matches (item.name s) "#.*") (item.set_feat s "name" (string-after (item.name s) "#"))))) (utt.relation.items u 'Segment)) (Wave_Synth u) (utt.play u) u)) (define (utt.relation.present utt relation) "(utt.relation.present UTT RELATIONNAME) Returns t if UTT caontains a relation called RELATIONNAME, nil otherwise." (if (member_string relation (utt.relationnames utt)) t nil)) (define (utt.relation.leafs utt relation) "(utt.relation.leafs UTT RELATIONNAME) Returns a list of all the leafs in this relation." (let ((leafs nil)) (mapcar (lambda (i) (if (not (item.down (item.relation i relation))) (set! leafs (cons i leafs)))) (utt.relation.items utt relation)) (reverse leafs))) (define (utt.relation.first utt relation) "(utt.relation.first UTT RELATIONNAME) Returns a the first item in this relation." (utt.relation utt relation)) (define (utt.relation.last utt relation) "(utt.relation.last UTT RELATIONNAME) Returns a the last item in this relation." (let ((i (utt.relation.first utt relation))) (while (item.next i) (set! i (item.next i))) i)) (define (item.feat.present item feat) "(item.feat.present item feat) nil if feat doesn't existing in this item, non-nil otherwise." (and item (assoc_string feat (item.features item)))) (define (item.relation.append_daughter parent relname daughter) "(item.relation.append_daughter parent relname daughter) Make add daughter to parent as a new daughter in relname." (item.append_daughter (item.relation parent relname) daughter)) (define (item.relation.insert si relname newsi direction) "(item.relation.insert si relname newsi direction) Insert newsi in relation relname with respect to direction. If direction is ommited after is assumed, valid directions are after before, above and below. Note you should use item.relation.append_daughter for tree adjoining. newsi maybe a item itself of a LISP description of one." (item.insert (item.relation si relname) newsi direction)) (define (item.relation.daughters parent relname) "(item.relation.daughters parent relname) Return a list of all daughters of parent by relname." (let ((d1 (item.daughter1 (item.relation parent relname))) (daughters)) (while d1 (set! daughters (cons d1 daughters)) (set! d1 (item.next d1))) (reverse daughters))) (define (item.daughters p) "(item.daughters parent) Return a list of all daughters of parent." (item.relation.daughters p (item.relation.name p))) (define (item.relation.parent si relname) "(item.relation.parent item relname) Return the parent of this item in this relation." (item.parent (item.relation si relname))) (define (item.relation.daughter1 si relname) "(item.relation.daughter1 item relname) Return the first daughter of this item in this relation." (item.daughter1 (item.relation si relname))) (define (item.relation.daughter2 si relname) "(item.relation.daughter2 item relname) Return the second daughter of this item in this relation." (item.daughter2 (item.relation si relname))) (define (item.relation.daughtern si relname) "(item.relation.daughtern item relname) Return the final daughter of this item in this relation." (item.daughtern (item.relation si relname))) (define (item.relation.next si relname) "(item.relation.next item relname) Return the next item in this relation." (item.next (item.relation si relname))) (define (item.relation.prev si relname) "(item.relation.prev item relname) Return the previous item in this relation." (item.prev (item.relation si relname))) (define (item.relation.first si relname) "(item.relation.first item relname) Return the most previous item from this item in this relation." (let ((n (item.relation si relname))) (while (item.prev n) (set! n (item.prev n))) n)) (define (item.leafs si) "(item.relation.leafs item relname) Return a list of the leafs of this item in this relation." (let ((ls nil) (pl (item.first_leaf si)) (ll (item.next_leaf (item.last_leaf si)))) (while (and pl (not (equal? pl ll))) (set! ls (cons pl ls)) (set! pl (item.next_leaf pl))) (reverse ls))) (define (item.relation.leafs si relname) "(item.relation.leafs item relname) Return a list of the leafs of this item in this relation." (item.leafs (item.relation si relname))) (define (item.root s) "(item.root s) Follow parent link until s has no parent." (cond ((item.parent s) (item.root (item.parent s))) (t s))) (define (item.parent_to s relname) "(item.parent_to s relname) Find the first ancestor of s in its current relation that is also in relname. s is treated as an ancestor of itself so if s is in relname it is returned. The returned value is in will be in relation relname or nil if there isn't one." (cond ((null s) s) ((member_string relname (item.relations s)) (item.relation s relname)) (t (item.parent_to (item.parent s) relname)))) (define (item.daughter1_to s relname) "(item.daughter1_to s relname) Follow daughter1 links of s in its current relation until an item is found that is also in relname, is s is in relname it is returned. The return item is returned in relation relname, or nil if there is nothing in relname." (cond ((null s) s) ((member_string relname (item.relations s)) (item.relation s relname)) (t (item.daughter1_to (item.daughter1 s) relname)))) (define (item.daughtern_to s relname) "(item.daughter1_to s relname) Follow daughtern links of s in its current relation until an item is found that is also in relname, is s is in relname it is returned. The return item is returned in relation relname, or nil if there is nothing in relname." (cond ((null s) s) ((member_string relname (item.relations s)) (item.relation s relname)) (t (item.daughtern_to (item.daughtern s) relname)))) (define (item.name s) "(item.name ITEM) Returns the name of ITEM. [see Accessing an utterance]" (item.feat s "name")) (define (utt.wave utt) "(utt.wave UTT) Get waveform from wave (R:Wave.first.wave)." (item.feat (utt.relation.first utt "Wave") "wave")) (define (utt.wave.rescale . args) "(utt.wave.rescale UTT FACTOR NORMALIZE) Modify the gain of the waveform in UTT by GAIN. If NORMALIZE is specified and non-nil the waveform is maximized first." (wave.rescale (utt.wave (nth 0 args)) (nth 1 args) (nth 2 args)) (nth 0 args)) (define (utt.wave.resample utt rate) "(utt.wave.resample UTT RATE)\ Resample waveform in UTT to RATE (if it is already at that rate it remains unchanged)." (wave.resample (utt.wave utt) rate) utt) (define (utt.import.wave . args) "(utt.import.wave UTT FILENAME APPEND) Load waveform in FILENAME into UTT in R:Wave.first.wave. If APPEND is specified and non-nil append this to the current waveform." (let ((utt (nth 0 args)) (filename (nth 1 args)) (append (nth 2 args))) (if (and append (member 'Wave (utt.relationnames utt))) (wave.append (utt.wave utt) (wave.load filename)) (begin (utt.relation.create utt 'Wave) (item.set_feat (utt.relation.append utt 'Wave) "wave" (wave.load filename)))) utt)) (define (utt.save.wave . args) "(utt.save.wave UTT FILENAME FILETYPE) Save waveform in UTT in FILENAME with FILETYPE (if specified) or using global parameter Wavefiletype." (wave.save (utt.wave (nth 0 args)) (nth 1 args) (nth 2 args)) (nth 0 args)) (define (utt.play utt) "(utt.play UTT) Play waveform in utt by current audio method." (wave.play (utt.wave utt)) utt) (define (utt.save.track utt filename relation feature) "(utt.save.track utt filename relation feature) DEPRICATED use trace.save instead." (format stderr "utt.save.track: DEPRICATED use track.save instead\n") (track.save (item.feat (utt.relation.first utt relation) feature) filename) utt) (define (utt.import.track utt filename relation fname) "(utt.import.track UTT FILENAME RELATION FEATURE_NAME) Load track in FILENAME into UTT in R:RELATION.first.FEATURE_NAME. Deletes RELATION if it already exists. (you maybe want to use track.load directly rather than this legacy function." (utt.relation.create utt relation) (item.set_feat (utt.relation.append utt relation) fname (track.load filename)) utt) (define (wagon_predict item tree) "(wagon_predict ITEM TREE) Predict with given ITEM and CART tree and return the prediction (the last item) rather than whole probability distribution." (car (last (wagon item tree)))) (define (phone_is_silence phone) (member_string phone (car (cdr (car (PhoneSet.description '(silences))))))) (define (phone_feature phone feat) "(phone_feature phone feat) Return the feature for given phone in current phone set, or 0 if it doesn't exist." (let ((ph (intern phone))) (let ((fnames (cadr (assoc 'features (PhoneSet.description)))) (fvals (cdr (assoc ph (cadr (assoc 'phones (PhoneSet.description))))))) (while (and fnames (not (string-equal feat (car (car fnames))))) (set! fvals (cdr fvals)) (set! fnames (cdr fnames))) (if fnames (car fvals) 0)))) (defvar server_max_clients 10 "server_max_clients In server mode, the maximum number of clients supported at any one time. When more that this number of clients attach simulaneous the last ones are denied access. Default value is 10. [see Server/client API]") (defvar server_port 1314 "server_port In server mode the inet port number the server will wait for connects on. The default value is 1314. [see Server/client API]") (defvar server_log_file t "server_log_file If set to t server log information is printed to standard output of the server process. If set to nil no output is given. If set to anything else the value is used as the name of file to which server log information is appended. Note this value is checked at server start time, there is no way a client may change this. [see Server/client API]") (defvar server_passwd nil "server_passwd If non-nil clients must send this passwd to the server followed by a newline before they can get a connection. It would be normal to set this for the particular server task. [see Server/client API]") (defvar server_access_list '(localhost) "server_access_list If non-nil this is the exhaustive list of machines and domains from which clients may access the server. This is a list of REGEXs that client host must match. Remember to add the backslashes before the dots. [see Server/client API]") (defvar server_deny_list nil "server_deny_list If non-nil this is a list of machines which are to be denied access to the server absolutely, irrespective of any other control features. The list is a list of REGEXs that are used to matched the client hostname. This list is checked first, then server_access_list, then passwd. [see Server/client API]") (define (def_feature_docstring fname fdoc) "(def_feature_docstring FEATURENAME FEATUREDOC) As some feature are used directly of stream items with no accompanying feature function, the features are just values on the feature list. This function also those features to have an accompanying documentation string." (let ((fff (assoc fname ff_docstrings))) (cond (fff ;; replace what's already there (set-cdr! fff fdoc)) (t (set! ff_docstrings (cons (cons fname fdoc) ff_docstrings)))) t)) (define (linear_regression item model) "(linear_regression ITEM MODEL) Use linear regression MODEL on ITEM. MODEL consists of a list of features, weights and optional map list. E.g. ((Intercept 100) (tobi_accent 10 (H* !H*)))." (let ((intercept (if (equal? 'Intercept (car (car model))) (car (cdr (car model))) 0)) (mm (if (equal? 'Intercept (car (car model))) (cdr model) model))) (apply + (cons intercept (mapcar (lambda (f) (let ((ff (item.feat item (car f)))) (if (car (cdr (cdr f))) (if (member_string ff (car (cdr (cdr f)))) (car (cdr f)) 0) (* (parse-number ff) (car (cdr f)))))) mm))))) (defvar help "The Festival Speech Synthesizer System: Help Getting Help (doc ') displays help on (manual nil) displays manual in local netscape C-c return to top level C-d or (quit) Exit Festival (If compiled with editline) M-h displays help on current symbol M-s speaks help on current symbol M-m displays relevant manula page in local netscape TAB Command, symbol and filename completion C-p or up-arrow Previous command C-b or left-arrow Move back one character C-f or right-arrow Move forward one character Normal Emacs commands work for editing command line Doing stuff (SayText TEXT) Synthesize text, text should be surrounded by double quotes (tts FILENAME nil) Say contexts of file, FILENAME should be surrounded by double quotes (voice_rab_diphone) Select voice (Britsh Male) (voice_kal_diphone) Select voice (American Male) ") (define (festival_warranty) "(festival_warranty) Display Festival's copyright and warranty. [see Copying]" (format t (string-append " The Festival Speech Synthesis System: " festival_version " Centre for Speech Technology Research University of Edinburgh, UK Copyright (c) 1996-2014 All Rights Reserved. Permission is hereby granted, free of charge, to use and distribute this software and its documentation without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of this work, and to permit persons to whom this work is furnished to do so, subject to the following conditions: 1. The code must retain the above copyright notice, this list of conditions and the following disclaimer. 2. Any modifications must be clearly marked as such. 3. Original authors' names are not deleted. 4. The authors' names are not used to endorse or promote products derived from this software without specific prior written permission. THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. "))) (define (intro) "(intro) Synthesize an introduction to the Festival Speech Synthesis System." (tts (path-append libdir "../examples/intro.text") nil)) (define (intro-spanish) "(intro-spanish) Synthesize an introduction to the Festival Speech Synthesis System in spanish. Spanish voice must already be selected for this." (tts (path-append libdir "../examples/spintro.text") nil)) (define (na_play FILENAME) "(play_wave FILENAME) Play given wavefile" (utt.play (utt.synth (eval (list 'Utterance 'Wave FILENAME))))) ;;; Some autoload commands (autoload manual-sym "festdoc" "Show appropriate manual section for symbol.") (autoload manual "festdoc" "Show manual section.") (autoload display "display" "Graphically display utterance.") (autoload festtest "festtest" "Run tests of Festival.") (defvar diphone_module_hooks nil "diphone_module_hooks A function or list of functions that will be applied to the utterance at the start of the diphone module. It can be used to map segment names to those that will be used by the diphone database itself. Typical use specifies _ and $ for consonant clusters and syllable boundaries, mapping to dark ll's etc. Reduction and tap type phenomena should probabaly be done by post lexical rules though the distinction is not a clear one.") (def_feature_docstring 'Segment.diphone_phone_name "Segment.diphone_phone_name This is produced by the diphone module to contain the desired phone name for the desired diphone. This adds things like _ if part of a consonant or $ to denote syllable boundaries. These are generated on a per voice basis by function(s) specified by diphone_module_hooks. Identification of dark ll's etc. may also be included. Note this is not necessarily the name of the diphone selected as if it is not found some of these characters will be removed and fall back values will be used.") (def_feature_docstring 'Syllable.stress "Syllable.stress The lexical stress of the syllable as specified from the lexicon entry corresponding to the word related to this syllable.") ;;; ;;; I tried some tests on the resulting speed both runtime and loadtime ;;; but compiled files don't seem to make any significant difference ;;; (define (compile_library) "(compile_library) Compile all the scheme files in the library directory." (mapcar (lambda (file) (format t "compile ... %s\n" file) (compile-file (string-before file ".scm"))) (list "synthesis.scm" "siod.scm" "init.scm" "lexicons.scm" "festival.scm" "gsw_diphone.scm" "intonation.scm" "duration.scm" "pos.scm" "phrase.scm" "don_diphone.scm" "rab_diphone.scm" "voices.scm" "tts.scm" "festdoc.scm" "languages.scm" "token.scm" "mbrola.scm" "display.scm" "postlex.scm" "tokenpos.scm" "festtest.scm" "cslush.scm" "ducs_cluster.scm" "sucs.scm" "web.scm" "cart_aux.scm" "lts_nrl.scm" "lts_nrl_us.scm" "email-mode.scm" "mrpa_phones.scm" "radio_phones.scm" "holmes_phones.scm" "mrpa_durs.scm" "klatt_durs.scm" "gswdurtreeZ.scm" "tobi.scm" "f2bf0lr.scm")) t) ;;; For mlsa resynthesizer (defvar mlsa_alpha_param 0.42) (defvar mlsa_beta_param 0.0) (provide 'festival) festival/lib/intonation.scm0000644000175000017500000001736212306075346014452 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Basic Intonation modules. These call appropriate sub-modules ;;; depending on the chosen intonation methods ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; These modules should predict intonation events/labels ;;; based on information in the phrase and word streams ; to detect prespecified accents (feature "accent" in 'Word relation) ; AS 5/29/00 (define (tobi_accent_prespecified utt) (let ((tobi_found nil) (words (utt.relation.items utt 'Word))) (while (and words (not tobi_found)) ; feature "accent" might be prespecified on words or tokens, AS 05/29/00 (if (item.feat.present (car words) 'accent) (set! tobi_found t) ; if Token relation exists, check tokens as well (if (not (null (item.parent (item.relation (car words) 'Token)))) (if (item.feat.present (item.parent (item.relation (car words) 'Token)) 'accent) (set! tobi_found t) (set! words (cdr words))) (set! words (cdr words))))) tobi_found)) (set! int_accent_cart_tree_no_accent '((NONE))) (define (Intonation utt) "(Intonation utt) Select between different intonation modules depending on the Parameter Int_Method. Currently offers three types: Simple, hats on each content word; ToBI, a tree method for predicting ToBI accents; and Default a really bad method with a simple downward sloping F0. This is the first of a two-stage intonation prediction process. This adds accent-like features to syllables, the second, Int_Targets generates the F0 contour itself. [see Intonation]" ; AS 5/29/00: Hack to avoid prediction of further accent labels ; on utterance chunks that have already been annotated with ; accent labels ; use CART that doesn't assign any labels when using Intonation_Tree (if (tobi_accent_prespecified utt) (progn (set! int_accent_cart_tree_save int_accent_cart_tree) (set! int_accent_cart_tree int_accent_cart_tree_no_accent) (Intonation_Tree utt) (set! int_accent_cart_tree int_accent_cart_tree_save)) (let ((rval (apply_method 'Int_Method utt))) (Parameter.get 'Int_Method) (cond (rval rval) ;; new style ((eq 'Simple (Parameter.get 'Int_Method)) (Intonation_Simple utt)) ((eq 'ToBI (Parameter.get 'Int_Method)) (format t "Using Intonation_Tree") (Intonation_Tree utt)) ((eq 'General (Parameter.get 'Int_Method)) (Intonation_Simple utt)) ;; yes this is a duplication (t (Intonation_Default utt)))))) ;;; These modules should create an actual F0 contour based on the ;;; the existing intonational events/labels etc ;;; Specifically this is called after durations have been predicted (define (Int_Targets utt) "(Int_Targets utt) The second stage in F0 prediction. This generates F0 targets related to segments using one of three methods, a simple hat, linear regression based on ToBI markings, and a simple declining slope. This second part deals with actual F0 values and durations, while the previous section only deals with accent (and boundary tone) assignment. [see Intonation]" (let ((rval (apply_method 'Int_Target_Method utt))) (cond (rval rval) ;; new style ((eq 'Simple (Parameter.get 'Int_Method)) (Int_Targets_Simple utt)) ((eq 'ToBI (Parameter.get 'Int_Method)) (Int_Targets_LR utt)) ((eq 'General (Parameter.get 'Int_Method)) (Int_Targets_General utt)) (t (Int_Targets_Default utt))))) ;;; ;;; A tree that adds accents (H) to all content words ;;; simple but better than nothing at all ;;; (set! simple_accent_cart_tree ' ((R:SylStructure.parent.gpos is content) ((stress is 1) ((Accented)) ((position_type is single) ((Accented)) ((NONE)))) ((NONE)))) (defvar duffint_params '((start 130) (end 110)) "duffint_params Default parameters for Default (duff) intonation target generation. This is an assoc list of parameters. Two parameters are supported start specifies the start F0 in Hertz for an utterance, and end specifies the end.") ;;; ;;; For simple testing, this function adds fixed duration and ;;; monotone intonation to a set of phones ;;; (defvar FP_F0 120 "FP_F0 In using Fixed_Prosody as used in Phones type utterances and hence SayPhones, this is the value in Hertz for the monotone F0.") (defvar FP_duration 100 "FP_duration In using Fixed_Prosody as used in Phones type utterances and hence SayPhones, this is the fix value in ms for phone durations.") (define (Fixed_Prosody utt) "(Fixed_Prosody UTT) Add fixed duration and fixed monotone F0 to the sgements in UTT. Uses values of FP_duration and FP_F0 as fixed values." (let (utt1 (dur_stretch (Parameter.get 'Duration_Stretch)) (orig_duffint_params duffint_params)) (Parameter.set 'Duration_Stretch (/ FP_duration 100.0)) (set! duffint_params (list (list 'start FP_F0) (list 'end FP_F0))) (set! utt1 (Duration_Default utt)) (set! utt1 (Int_Targets_Default utt1)) ;; Reset Parameter values back (Parameter.set 'Duration_Stretch dur_stretch) (set! duffint_params orig_duffint_params) utt1 ) ) (define (segment_dpitch seg) "(segment_dpitch UTT SEG) Returns delta pitch, this pitch minus previous pitch." (- (parse-number (item.feat utt seg 'seg_pitch)) (parse-number (item.feat utt seg 'R:Segment.p.seg_pitch)))) (provide 'intonation) festival/lib/duration.scm0000644000175000017500000001647407262605650014122 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Basic Duration module which will call appropriate duration ;;; (C++) modules based on set parameter ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; These modules should predict intonation events/labels ;;; based on information in the phrase and word streams (define (Duration utt) "(Duration utt) Predict segmental durations using Duration_Method defined in Parameters. Four methods are currently available: averages, Klatt rules, CART tree based, and fixed duration." (let ((rval (apply_method 'Duration_Method utt))) (cond (rval rval) ;; new style ;; 1.1.1 voices still use other names ((eq 'Averages (Parameter.get 'Duration_Method)) (Duration_Averages utt)) ((eq 'Klatt (Parameter.get 'Duration_Method)) (Duration_Klatt utt)) ((eq 'Tree_ZScores (Parameter.get 'Duration_Method)) (Duration_Tree_ZScores utt)) ((eq 'Tree (Parameter.get 'Duration_Method)) (Duration_Tree utt)) (t (Duration_Default utt))))) (define (Duration_LogZScores utt) "(Duration_LogZScores utt) Predicts duration to segments using the CART tree in duration_logzscore_tree and duration_logzscore_tree_silence which produces a zscore of the log duration. The variable duration_logzscore_ph_info contains (log) means and std for each phone in the set." (let ((silence (car (car (cdr (assoc 'silences (PhoneSet.description)))))) ldurinfo) (mapcar (lambda (s) (if (string-equal silence (item.name s)) (set! ldurinfo (wagon s duration_logzscore_tree_silence)) (set! ldurinfo (wagon s duration_logzscore_tree))) (set! dur (exp (duration_unzscore (item.name s) (car (last ldurinfo)) duration_logzscore_ph_info))) (set! dur (* dur (duration_find_stretch s))) (item.set_feat s "end" (+ dur (item.feat s "start_segment")))) (utt.relation.items utt 'Segment)) utt)) (define (duration_unzscore phname zscore table) "(duration_unzscore phname zscore table) Look up phname in table and convert xscore back to absolute domain." (let ((phinfo (assoc phname table)) mean std) (if phinfo (begin (set! mean (car (cdr phinfo))) (set! std (car (cdr (cdr phinfo))))) (begin (format t "Duration: unzscore no info for %s\n" phname) (set! mean 0.100) (set! std 0.25))) (+ mean (* zscore std)))) (define (duration_find_stretch seg) "(duration_find_stretch utt seg) Find any relavant duration stretch." (let ((global (Parameter.get 'Duration_Stretch)) (local (item.feat seg "R:SylStructure.parent.parent.R:Token.parent.dur_stretch"))) (if (or (not global) (equal? global 0.0)) (set! global 1.0)) (if (string-equal local 0.0) (set! local 1.0)) (* global local))) ;; These provide lisp level functions, some of which have ;; been converted in C++ (in festival/src/modules/base/ff.cc) (define (onset_has_ctype seg type) ;; "1" if onset contains ctype (let ((syl (item.relation.parent seg 'SylStructure))) (if (not syl) "0" ;; a silence (let ((segs (item.relation.daughters syl 'SylStructure)) (v "0")) (while (and segs (not (string-equal "+" (item.feat (car segs) "ph_vc")))) (if (string-equal type (item.feat (car segs) "ph_ctype")) (set! v "1")) (set! segs (cdr segs))) v)))) (define (coda_has_ctype seg type) ;; "1" if coda contains ctype (let ((syl (item.relation.parent seg 'SylStructure))) (if (not syl) "0" ;; a silence (let ((segs (reverse (item.relation.daughters syl 'SylStructure))) (v "0")) (while (and segs (not (string-equal "+" (item.feat (car segs) "ph_vc")))) (if (string-equal type (item.feat (car segs) "ph_ctype")) (set! v "1")) (set! segs (cdr segs))) v)))) (define (onset_stop seg) (onset_has_ctype seg "s")) (define (onset_fric seg) (onset_has_ctype seg "f")) (define (onset_nasal seg) (onset_has_ctype seg "n")) (define (onset_glide seg) (let ((l (onset_has_ctype seg "l"))) (if (string-equal l "0") (onset_has_ctype seg "r") "1"))) (define (coda_stop seg) (coda_has_ctype seg "s")) (define (coda_fric seg) (coda_has_ctype seg "f")) (define (coda_nasal seg) (coda_has_ctype seg "n")) (define (coda_glide seg) (let ((l (coda_has_ctype seg "l"))) (if (string-equal l "0") (coda_has_ctype seg "r") "1"))) (define (Unisyn_Duration utt) "(UniSyn_Duration utt) predicts Segment durations is some speficied way but holds the result in a way necessary for other Unisyn code." (let ((end 0)) (mapcar (lambda (s) (item.get_utt s) (let ((dur (wagon_predict s duration_cart_tree))) (set! dur (* (Parameter.get 'Duration_Stretch) dur)) (set! end (+ dur end)) (item.set_feat s "target_dur" dur) (item.set_function s "start" "unisyn_start") (item.set_feat s "end" end) (item.set_feat s "dur" dur) )) (utt.relation.items utt 'Segment)) utt)) (provide 'duration) festival/lib/pos.scm0000644000175000017500000002203110643201763013052 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; A part of speech tagger ;;; (set! english_guess_pos '((in of for in on that with by at from as if that against about before because if under after over into while without through new between among until per up down) (to to) (det the a an no some this that each another those every all any these both neither no many) (md will may would can could should must ought might) (cc and but or plus yet nor) (wp who what where how when) (pps her his their its our their its mine) (aux is am are was were has have had be) (punc "." "," ":" ";" "\"" "'" "(" "?" ")" "!") )) (defvar guess_pos english_guess_pos "guess_pos An assoc-list of simple part of speech tag to list of words in that class. This basically only contains closed class words all other words may be assumed to be content words. This was built from information in the f2b database and is used by the ffeature gpos.") ;;; A more elaborate part of speech tagger using ngrams works but ;;; at present requires a large list of a priori probabilities ;;; to work. If that file exists on your system we'll use it otherwise ;;; POS is guessed by the lexicon ;;; These models were build from the Penn TreeBank, WSJ corpus (defvar pos_model_dir lexdir "pos_model_dir The directory contains the various models for the POS module. By default this is the same directory as lexdir. The directory should contain two models: a part of speech lexicon with reverse log probabilities and an ngram model for the same part of speech tag set.") (defvar pos_p_start_tag "punc" "pos_p_start_tag This variable's value is the tag most likely to appear before the start of a sentence. It is used when looking for pos context before an utterance. Typically it should be some type of punctuation tag.") (defvar pos_pp_start_tag "n" "pos_pp_start_tag This variable's value is the tag most likely to appear before pos_p_start_tag and any position preceding that. It is typically some type of noun tag. This is used to provide pos context for early words in an utterance.") (defvar pos_supported nil "pos_supported If set to non-nil use part of speech prediction, if nil just get pos information from the lexicon.") (defvar pos_ngram_name nil "pos_ngram_name The name of a loaded ngram containing the a posteriori ngram model for predicting part of speech. The a priori model is held as a lexicon call poslex.") (defvar pos_map nil "pos_map If set this should be a reverse assoc-list mapping on part of speech tag set to another. It is used after using the defined POS models to map the pos feature on each word to a new tagset.") ;;; ;;; All the names here don't really allow multiple versions ;;; they should be prefixed with english_ ;;; (if (probe_file (path-append pos_model_dir "wsj.wp39.poslexR")) (begin (lex.create "english_poslex") (lex.set.compile.file (path-append pos_model_dir "wsj.wp39.poslexR")) (lex.set.phoneset "mrpa") (lex.set.lts.method nil) (set! pos_lex_name "english_poslex") (set! pos_p_start_tag "punc") (set! pos_pp_start_tag "nn") ;; wp39 (lex.add.entry '("_OOV_" ((nnp -2.9144) (jj -2.7357) (nn -3.5787) (nns -3.4933) (vbn -3.2486) (vbg -2.9419) (vb -3.5471) (vbd -3.7896) (vbz -3.7820) (rb -4.1940) (vbp -3.2755) (nnps -2.1605)) ())) (lex.add.entry '("_number_" ((cd -0.35202) (jj -4.1083) (nns -6.4488) (nnp -7.3595)) () )) (lex.add.entry '("," ((punc -0.88488)) () )) (lex.add.entry '("." ((punc -1.1104)) () )) (lex.add.entry '(":" ((punc -4.4236)) () )) (lex.add.entry '("``" ((punc -2.7867)) () )) (lex.add.entry '("`" ((punc -2.7867)) () )) (lex.add.entry '("'" ((punc -2.7867)) () )) (lex.add.entry '("\"" ((punc -2.7867)) () )) ;; wp17 ;; (lex.add.entry '("_OOV_" ((n -3.4109) (j -2.7892) (v -3.7426)) ())) ; (lex.add.entry '("_OOV_" ((n -1.968) (j -2.351) (v -2.287)) ())) ; (lex.add.entry '("_number_" ((j -0.35202)) ())) ; (lex.add.entry '("," ((punc -0.88359)) () )) ; (lex.add.entry '("." ((punc -1.1101)) () )) ; (lex.add.entry '(":" ((punc -4.4236)) () )) ; (lex.add.entry '("``" ((punc -2.7867)) () )) ; (lex.add.entry '("`" ((punc -2.7867)) () )) ; (lex.add.entry '("'" ((punc -2.7867)) () )) ; (lex.add.entry '("\"" ((punc -2.7867)) () )) ;; wp22 ; (lex.add.entry '("_OOV_" ((n -3.4109) (j -2.7892) (v -3.7426)) ())) ; (lex.add.entry '("_number_" ((cd -0.35202) (j -4.1908) (n -7.3890)) ())) ; (lex.add.entry '("," ((punc -0.88359)) () )) ; (lex.add.entry '("." ((punc -1.1101)) () )) ; (lex.add.entry '(":" ((punc -4.4236)) () )) ; (lex.add.entry '("``" ((punc -2.7867)) () )) ;; wp18 ; (lex.add.entry '("_OOV_" ((n -3.4109) (j -2.7892) (v -3.7426)) ())) ; (lex.add.entry '("_number_" ((j -0.35202)) ())) ; (lex.add.entry '("`" ((punc -6.539) ) () )) ; (lex.add.entry '("``" ((punc -2.399) ) () )) ; (lex.add.entry '("," ((punc -0.480) ) () )) ; (lex.add.entry '("." ((fpunc -0.012) ) () )) ; (lex.add.entry '(":" ((punc -4.100) ) () )) (ngram.load 'english_pos_ngram (path-append pos_model_dir "wsj.wp39.tri.ngrambin")) ; (ngram.load 'english_pos_ngram ; (path-append pos_model_dir "wsj.wp45.tri.ngram")) (set! pos_supported t) ) (set! pos_supported nil)) (setq english_pos_map_wp39_to_wp20 '( (( vbd vb vbn vbz vbp vbg ) v) (( nn nnp nns nnps fw sym ls ) n) (( dt ) dt) (( punc fpunc ) punc) (( in ) in) (( jj jjr jjs 1 2 ) j) (( prp ) prp) (( rb rp rbr rbs ) r) (( cc ) cc) (( of ) of) (( to ) to) (( cd ) cd) (( md ) md) (( pos ) pos) (( wdt ) wdt) (( wp ) wp) (( wrb ) wrb) (( ex ) ex) (( uh ) uh) (( pdt ) pdt) )) (defvar pos_map nil "pos_map A reverse assoc list of predicted pos tags to some other tag set. Note using this changes the pos tag loosing the actual predicted value. Rather than map here you may find it more appropriate to map tags sets locally in the modules that use them (e.g. phrasing and lexicons).") ;;(setq pos_map_remap ;; '( ;; (( fpunc ) punc) ;; (( of ) in))) (def_feature_docstring 'Word.pos "Word.pos Part of speech tag value returned by the POS tagger module.") (def_feature_docstring 'Word.pos_score "Word.pos_score Part of speech tag log likelihood from Viterbi search.") (define (POS utt) "(POS utt) Apply part of speech tagging (and possible parsing too) to Word relation." (let ((rval (apply_method 'POS_Method utt))) (cond (rval rval) ;; new style (t (Classic_POS utt))))) (provide 'pos) festival/lib/phrase.scm0000644000175000017500000001534107262605650013547 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Phrase boundary prediction. ;;; ;;; Two methods supported, if POS is enabled we use ngrams for that ;;; otherwise we use a CART tree ;;; ;;; Models trained from the IBM/Lancaster Spoken English Corpus and ;;; Boston University's FM Radio Corpus. ;;; ;;; Here's a very simple CART tree for predicting phrase breaks ;;; based on punctuation only ;;; (set! simple_phrase_cart_tree ' ((lisp_token_end_punc in ("?" "." ":")) ((BB)) ((lisp_token_end_punc in ("'" "\"" "," ";")) ((B)) ((n.name is 0) ;; end of utterance ((BB)) ((NB)))))) (define (token_end_punc word) "(token_end_punc UTT WORD) If punctuation at end of related Token and if WORD is last word in Token return punc, otherwise 0." (if (item.relation.next word "Token") "0" (item.feat word "R:Token.parent.punc"))) ;;; This is a simple CART tree used after boundaries are predicted ;;; by the probabilistic method to get two levels of break (set! english_phrase_type_tree '((pbreak is NB) ((num_break is 1) ((mB)) ((R:Token.parent.EMPH is 1) ((NB)) ((n.R:Token.parent.EMPH is 1) ((NB)) ((NB))))) ((pbreak is BB) ((BB)) ((pbreak is mB) ((mB)) ((name in ("." "!" "?"));; only (potentially) change Bs to BBs ((BB)) ((B))))))) (set! f2b_phrase_cart_tree ' ((gpos is punc) (((1 0.00238095) (3 0) (4 0.997619) B)) (((4 0.00238095) (3 0) (1 0.997619) NB)))) ;;; For more detailed prediction of phrase breaks we use POS and ;;; probability distribution of breaks ;;; These models were trained using data from the Lancaster/IBM ;;; Spoken English Corpus (require 'pos) ;; for part of speech map (defvar pbreak_ngram_dir libdir "pbreak_ngram_dir The directory containing the ngram models for predicting phrase breaks. By default this is the standard library directory.") (defvar english_phr_break_params (list ;; The name and filename off the ngram with the a priori ngram model ;; for predicting phrase breaks in the Phrasify module. This model should ;; predict probability distributions for B and NB given some context of ;; part of speech tags. (list 'pos_ngram_name 'english_break_pos_ngram) (list 'pos_ngram_filename (path-append pbreak_ngram_dir "sec.ts20.quad.ngrambin")) ;; The name and filename of the ngram containing the a posteriori ngram ;; for predicting phrase breaks in the Phrasify module. This module should ;; predict probability distributions for B and NB given previous B and ;; NBs. (list 'break_ngram_name 'english_break_ngram) (list 'break_ngram_filename (path-append pbreak_ngram_dir "sec.B.hept.ngrambin")) ;; A weighting factor for breaks in the break/non-break ngram. (list 'gram_scale_s 0.59) ;; When Phrase_Method is prob_models, this tree, if set is used to ;; potentially predict phrase type. At least some prob_models only ;; predict B or NB, this tree may be used to change some Bs into ;; BBs. If it is nil, the pbreak value predicted by prob_models ;; remains the same. (list 'phrase_type_tree english_phrase_type_tree) ;; A list of tags used in identifying breaks. Typically B and NB (and ;; BB). This should be the alphabet of the ngram identified in ;; break_ngram_name (list 'break_tags '(B NB)) (list 'pos_map english_pos_map_wp39_to_wp20) ) "english_phr_break_params Parameters for English phrase break statistical model.") (defvar phr_break_params nil "phr_break_params Parameters for phrase break statistical model. This is typcal set by a voice selection function to the parameters for a particular model.") ;;; ;;; Declaration of some features ;;; (def_feature_docstring 'Word.pbreak "Word.pbreak Result from statistical phrasing module, may be B or NB denoting phrase break or non-phrase break after the word.") (def_feature_docstring 'Word.pbreak_score "Word.pbreak_score Log likelihood score from statistical phrasing module, for pbreak value.") (def_feature_docstring 'Word.blevel "Word.blevel A crude translation of phrase break into ToBI like phrase level. Values may be 0,1,2,3,4.") (define (Phrasify utt) "(Phrasify utt) Construct phrasify over Words module." (let ((rval (apply_method 'Phrasify_Method utt))) (cond (rval rval) ;; new style (t (Classic_Phrasify utt))))) (provide 'phrase) festival/lib/voices.scm0000644000175000017500000003215112437426263013554 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Preapre to access voices. Searches down a path of places. ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define current-voice nil "current-voice The name of the current voice.") ;; The path to search for voices is created from the load-path with ;; an extra list of directories appended. (defvar system-voice-path '( ) "system-voice-path Additional directory not near the load path where voices can be found, this can be redefined in lib/sitevars.scm if desired.") (defvar system-voice-path-multisyn '( ) "system-voice-path-multisyn Additional directory not near the load path where multisyn voices can be found, this can be redefined in lib/sitevars.scm if desired.") (defvar voice-path (remove-duplicates (append (mapcar (lambda (d) (path-append d "voices/")) load-path) (mapcar (lambda (d) (path-as-directory d)) system-voice-path) )) "voice-path List of places to look for voices. If not set it is initialised from load-path by appending \"voices/\" to each directory with system-voice-path appended.") (defvar voice-path-multisyn (remove-duplicates (append (mapcar (lambda (d) (path-append d "voices-multisyn/")) load-path) (mapcar (lambda (d) (path-as-directory d)) system-voice-path-multisyn) )) "voice-path-multisyn List of places to look for multisyn voices. If not set it is initialised from load-path by appending \"voices-multisyn/\" to each directory with system-voice-path-multisyn appended.") ;; Declaration of voices. When we declare a voice we record the ;; directory and set up an autoload for the vocie-selecting function (defvar voice-locations () "voice-locations Association list recording where voices were found.") (defvar voice-location-trace nil "voice-location-trace Set t to print voice locations as they are found") (define (voice-location name dir doc) "(voice-location NAME DIR DOCSTRING) Record the location of a voice. Called for each voice found on voice-path. Can be called in site-init or .festivalrc for additional voices which exist elsewhere." (let ((func_name (intern (string-append "voice_" name))) ) (set! name (intern name)) (set! voice-locations (cons (cons name dir) voice-locations)) (eval (list 'autoload func_name (path-append dir "festvox/" name) doc)) (if voice-location-trace (format t "Voice: %s %s\n" name dir) ) ) ) (define (voice-location-multisyn name rootname dir doc) "(voice-location NAME ROOTNAME DIR DOCSTRING) Record the location of a voice. Called for each voice found on voice-path. Can be called in site-init or .festivalrc for additional voices which exist elsewhere." (let ((func_name (intern (string-append "voice_" name))) ) (set! name (intern name)) (set! voice-locations (cons (cons name dir) voice-locations)) (eval (list 'autoload func_name (path-append dir "festvox/" rootname) doc)) (if voice-location-trace (format t "Voice: %s %s\n" name dir) ) ) ) (define (current_voice_reset) "(current_voice_reset) This function is called at the start of defining any new voice. It is design to allow the previous voice to reset any global values it has messed with. If this variable value is nil then the function wont be called.") (define (voice_reset) "(voice_reset) This resets all variables back to acceptable values that may affect voice generation. This function should always be called at the start of any function defining a voice. In addition to reseting standard variables the function current_voice_reset will be called. This should always be set by the voice definition function (even if it does nothing). This allows voice specific changes to be reset when a new voice is selection. Unfortunately I can't force this to be used." (Parameter.set 'Duration_Stretch 1.0) (set! after_synth_hooks default_after_synth_hooks) ;; The follow are reset to allow existing voices to continue ;; to work, new voices should be setting these explicitly (Parameter.set 'Token_Method 'Token_English) (Parameter.set 'POS_Method Classic_POS) (Parameter.set 'Phrasify_Method Classic_Phrasify) (Parameter.set 'Word_Method Classic_Word) (Parameter.set 'Pause_Method Classic_Pauses) (Parameter.set 'PostLex_Method Classic_PostLex) (set! diphone_module_hooks nil) (set! UniSyn_module_hooks nil) (if current_voice_reset (current_voice_reset)) (set! current_voice_reset nil) ) (defvar Voice_descriptions nil "Internal variable containing list of voice descriptions as decribed by proclaim_voice.") (define (proclaim_voice name description) "(proclaim_voice NAME DESCRIPTION) Describe a voice to the systen. NAME should be atomic name, that conventionally will have voice_ prepended to name the basic selection function. OPTIONS is an assoc list of feature and value and must have at least features for language, gender, dialect and description. The first there of these are atomic, while the description is a text string describing the voice." (let ((voxdesc (assoc name Voice_descriptions))) (if voxdesc (set-car! (cdr voxdesc) description) (set! Voice_descriptions (cons (list name description) Voice_descriptions)))) ) (define (voice.description name) "(voice.description NAME) Output description of named voice. If the named voice is not yet loaded it is loaded." (let ((voxdesc (assoc name Voice_descriptions)) (cv current-voice)) (if (null voxdesc) (unwind-protect (begin (voice.select name) (voice.select cv) ;; switch back to current voice (set! voxdesc (assoc name Voice_descriptions))))) (if voxdesc voxdesc (begin (format t "SIOD: unknown voice %s\n" name) nil)))) (define (voice.select name) "(voice.select NAME) Call function to set up voice NAME. This is normally done by prepending voice_ to NAME and call it as a function." (eval (list (intern (string-append "voice_" name))))) (define (voice.describe name) "(voice.describe NAME) Describe voice NAME by saying its description. Unfortunately although it would be nice to say that voice's description in the voice itself its not going to work cross language. So this just uses the current voice. So here we assume voices describe themselves in English which is pretty anglo-centric, shitsurei shimasu." (let ((voxdesc (voice.description name))) (let ((desc (car (cdr (assoc 'description (car (cdr voxdesc))))))) (cond (desc (tts_text desc nil)) (voxdesc (SayText (format nil "A voice called %s exist but it has no description" name))) (t (SayText (format nil "There is no voice called %s defined" name))))))) (define (voice.list) "(voice.list) List of all (potential) voices in the system. This checks the voice-location list of potential voices found be scanning the voice-path at start up time. These names can be used as arguments to voice.description and voice.describe." (mapcar car voice-locations)) ;; Voices are found on the voice-path if they are in directories of the form ;; DIR/LANGUAGE/NAME (define (search-for-voices) "(search-for-voices) Search down voice-path to locate voices." (let ((dirs voice-path) (dir nil) languages language voices voicedir voice ) (while dirs (set! dir (car dirs)) (setq languages (directory-entries dir t)) (while languages (set! language (car languages)) (set! voices (directory-entries (path-append dir language) t)) (while voices (set! voicedir (car voices)) (set! voice (path-basename voicedir)) (if (string-matches voicedir ".*\\..*") nil (voice-location voice (path-as-directory (path-append dir language voicedir)) "voice found on path") ) (set! voices (cdr voices)) ) (set! languages (cdr languages)) ) (set! dirs (cdr dirs)) ) ) ) ;; A single file is allowed to define multiple multisyn voices, so this has ;; been adapted for this. Rob thinks this is just evil, but couldn't think ;; of a better way. (define (search-for-voices-multisyn) "(search-for-voices-multisyn) Search down multisyn voice-path to locate multisyn voices." (let ((dirs voice-path-multisyn) (dir nil) languages language voices voicedir voice voice-list ) (while dirs (set! dir (car dirs)) (set! languages (directory-entries dir t)) (while languages (set! language (car languages)) (set! voices (directory-entries (path-append dir language) t)) (while voices (set! voicedir (car voices)) (set! voice (path-basename voicedir)) (if (string-matches voicedir ".*\\..*") nil (begin ;; load the voice definition file, but don't evaluate it! (set! voice-def-file (load (path-append dir language voicedir "festvox" (string-append voicedir ".scm")) t)) ;; now find the "proclaim_voice" lines and register these voices. (mapcar (lambda (line) (if (string-matches (car line) "proclaim_voice") (voice-location-multisyn (intern (cadr (cadr line))) voicedir (path-append dir language voicedir) "registerd multisyn voice"))) voice-def-file) )) (set! voices (cdr voices))) (set! languages (cdr languages))) (set! dirs (cdr dirs))))) (search-for-voices) (search-for-voices-multisyn) ;; We select the default voice from a list of possibilities. One of these ;; had better exist in every installation. (define (no_voice_error) (format t "\nWARNING\n") (format t "No default voice found in %l\n" voice-path) (format t "either no voices unpacked or voice-path is wrong\n") (format t "Scheme interpreter will work, but there is no voice to speak with.\n") (format t "WARNING\n\n")) (defvar voice_default 'no_voice_error "voice_default A variable whose value is a function name that is called on start up to the default voice. [see Site initialization]") (defvar default-voice-priority-list '(kal_diphone cmu_us_slt_cg cmu_us_rms_cg cmu_us_bdl_cg cmu_us_jmk_cg cmu_us_awb_cg ; cstr_rpx_nina_multisyn ; restricted license (lexicon) ; cstr_rpx_jon_multisyn ; restricted license (lexicon) ; cstr_edi_awb_arctic_multisyn ; restricted license (lexicon) ; cstr_us_awb_arctic_multisyn ked_diphone don_diphone rab_diphone en1_mbrola us1_mbrola us2_mbrola us3_mbrola gsw_diphone ;; not publically distributed el_diphone ) "default-voice-priority-list List of voice names. The first of them available becomes the default voice.") (let ((voices default-voice-priority-list) voice) (while (and voices (eq voice_default 'no_voice_error)) (set! voice (car voices)) (if (assoc voice voice-locations) (set! voice_default (intern (string-append "voice_" voice))) ) (set! voices (cdr voices)) ) ) (provide 'voices) festival/lib/tts.scm0000644000175000017500000002565310267736007013106 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Various tts functions and hooks ;;; Once the utterance is built these functions synth and play it (defvar tts_hooks (list utt.synth utt.play) "tts_hooks Function or list of functions to be called during text to speech. The function tts_file, chunks data into Utterances of type Token and applies this hook to the utterance. This typically contains the utt.synth function and utt.play. [see TTS]") ;;; This is used to define utterance breaks in tts on files (defvar eou_tree '((lisp_max_num_tokens > 200) ((1)) ((n.whitespace matches ".*\n.*\n\\(.\\|\n\\)*");; significant break (2 nls) ((1)) ((name matches "--+") ((1)) ((punc matches ".*[\\?:!;].*") ((1)) ((punc matches ".*\\..*") ((punc matches "..+");; longer punctuation string ((punc matches "\\..*,") ;; for U.S.S.R., like tokens ((0)) ((1))) ;; This is to distinguish abbreviations vs periods ;; These are heuristics ((name matches "\\(.*\\..*\\|[A-Z][A-Za-z]?[A-Za-z]?\\|etc\\)");; an abbreviation ((n.whitespace is " ") ((0));; if abbrev single space isn't enough for break ((n.name matches "[A-Z].*") ((1)) ((0)))) ((n.whitespace is " ");; if it doesn't look like an abbreviation ((n.name matches "[A-Z].*");; single space and non-cap is no break ((1)) ((0))) ((1))))) ((0))))))) "eou_tree End of utterance tree. A decision tree used to determine if the given token marks the end of an utterance. It may look one token ahead to do this. [see Utterance chunking]") (define (max_num_tokens x) "(num_tokens x) This is probably controversial, but its good to have a maximum number of tokens in an utterance. You really dont want to wait on very long utterances, some utts can be thousands of words long, these maybe shouldn't be spoken, but we do have to deal with them." (let ((c 1) (y x)) (while y (set! c (+ 1 c)) (set! y (item.prev y))) c)) ;;; The program used to parse stml files ;;; Needs version 1.0 to allow -D option to work (defvar sgml_parse_progname "nsgmls-1.0" "sgml_parse_progname The name of the program to use to parse SGML files. Typically this is nsgml-1.0 from the sp SGML package. [see XML/SGML requirements]") ;;; When PHRASE elements are specified in an utterance in STML ;;; no other method for phrase prediction is to be used, so we ;;; use the following tree (set! stml_phrase_cart_tree '((R:Token.parent.pbreak is B) ((B)) ((n.name is 0) ((B)) ((NB))))) (define (xxml_synth utt) "(xxml_synth UTT) This applies the xxml_hooks (mode specific) and tts_hooks to the given utterance. This function should be called from xxml element definitions that signal an utterance boundary." (cond ((or (not utt) (not (utt.relation utt 'Token))) ;; no tokens nil) (t (apply_hooks xxml_hooks utt) (apply_hooks tts_hooks utt) (set! utt nil) ;; not enough ... (gc) utt)) ) (define (xxml_attval ATTNAME ATTLIST) "(xxml_attval ATTNAME ATTLIST) Returns attribute value of ATTNAME in ATTLIST or nil if it doesn't exists." (cond ((not ATTLIST) nil) ((string-equal ATTNAME (car (car ATTLIST))) (car (cdr (car ATTLIST)))) (t (xxml_attval ATTNAME (cdr ATTLIST))))) (defvar xxml_word_features nil "xxml_word_features An assoc list of features to be added to the current word when in xxml parse mode.") (defvar xxml_token_hooks nil "xxml_token_hooks Functions to apply to each token.") (defvar xxml_hooks nil "xxml_hooks Function or list of functions to be applied to an utterance when parsed with xxML, before tts_hooks.") (defvar xxml_elements nil "xxml_elements List of Scheme actions to perform on finding xxML tags.") (defvar xml_dtd_dir libdir "xml_dtd_dir The directory holding standard DTD form the xml parser.") (set! tts_fnum 1) (define (save_tts_output utt) (let ((fname (string-append "tts_file_" tts_fnum ".wav"))) (format stderr "festival: saving waveform in %s\n" fname) (utt.save.wave utt fname) (set! tts_fnum (+ 1 tts_fnum)) utt)) (define (save_waves_during_tts) "(save_waves_during_tts) Save each waveform in the current directory in files \"tts_file_XXX.wav\". use (save_waves_during_tts_STOP) to stop saving waveforms" (if (not (member save_tts_output tts_hooks)) (set! tts_hooks (append tts_hooks (list save_tts_output)))) t) (define (save_waves_during_tts_STOP) "(save_waves_during_tts_STOP) Stop saving waveforms when doing tts." (if (member save_tts_output tts_hooks) (set! tts_hooks (delq save_tts_output tts_hooks))) t) (define (tts file mode) "(tts FILE MODE) Convert FILE to speech. MODE identifies any special treatment necessary for FILE. This is simply a front end to tts_file but puts the system in async audio mode first. [see TTS]" (audio_mode 'async) (if mode (tts_file file mode) (tts_file file (tts_find_text_mode file auto-text-mode-alist))) ;; (audio_mode 'sync) ;; Hmm this is probably bad ) (define (tts_text string mode) "(tts_text STRING mode) Apply tts on given string. That is, segment it into utterances and apply tts_hooks to each utterance. This is naively done by saving the string to a file and calling tts_file on that file. This differs from SayText which constructs a single utterance for the whole given text." (let ((tmpfile (make_tmp_filename)) (fd)) (set! fd (fopen tmpfile "wb")) (format fd "%s" string) (fclose fd) (audio_mode 'async) (tts_file tmpfile mode) (delete-file tmpfile))) (define (save_record_wave utt) "Saves the waveform and records its so it can be joined into a a single waveform at the end." (let ((fn (make_tmp_filename))) (utt.save.wave utt fn) (set! wavefiles (cons fn wavefiles)) utt)) (define (combine_waves) "(combine_waves) Join all the waves together into the desired output file and delete the intermediate ones." (let ((wholeutt (Utterance Text ""))) (mapcar (lambda (d) (utt.import.wave wholeutt d t) (delete-file d)) (reverse wavefiles)) wholeutt)) (define (tts_textall string mode) "(tts_textall STRING MODE) Apply tts to STRING. This function is specifically designed for use in server mode so a single function call may synthesize the string. This function name maybe added to the server safe functions." (if (not (string-equal mode "nil")) (begin ;; a mode has been specified so do something different (let ((tmpfile (make_tmp_filename)) (fd)) (set! fd (fopen tmpfile "wb")) (format fd "%s" string) (fclose fd) (set! tts_hooks (list utt.synth save_record_wave)) (set! wavefiles nil) (tts_file tmpfile mode) (delete-file tmpfile) (utt.send.wave.client (combine_waves)) )) ;; Simple fundamental mode (utt.send.wave.client (utt.synth (eval (list 'Utterance 'Text string)))))) ;; Function to interface with app_festival for asterisk ;; See http://www.asterisk.org (define (tts_textasterisk string mode) "(tts_textasterisk STRING MODE) Apply tts to STRING. This function is specifically designed for use in server mode so a single function call may synthesize the string. This function name may be added to the server safe functions." (utt.send.wave.asterisk (utt.synth (eval (list 'Utterance 'Text string))))) (define (tts_return_to_client) "(tts_return_to_client) This function is called by clients who wish to return waveforms of their text samples asynchronously. This replaces utt.play in tts_hooks with utt.send.wave.client." (if (not (member utt.send.wave.client tts_hooks)) (set! tts_hooks (append (delq utt.play tts_hooks) (list utt.send.wave.client))))) (defvar tts_text_modes nil "tts_text_modes An a-list of text modes data for file type specific tts functions. See the manual for an example. [see Text modes]") (define (tts_find_text_mode file alist) "(find_text_mode FILE ALIST) Search through ALIST for one that matches FILE. Returns nil if nothing macthes." (cond ((null alist) nil) ;; can't find a match ((string-matches file (string-append ".*" (car (car alist)) ".*")) (cdr (car alist))) (t (tts_find_text_mode file (cdr alist))))) (defvar auto-text-mode-alist (list (cons "\\.sable$" 'sable) (cons "\\.ogi" 'ogimarkup) (cons "\\.email" 'email) (cons "" 'fundamental) ) "auto-text-mode-alist Following Emacs' auto-mode-alist thios provides a mechanism for auto selecting a TTS text mode based on the filename being analyzed. Its format is exactly the same as Emacs in that it consists of an alist of dotted pairs of regular expression and text mode name.") (provide 'tts) festival/lib/festdoc.scm0000644000175000017500000001672607262605650013724 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: August 1996 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Save documentation strings as texinfo files ;;; ;;; Finds all functions with documentation, and all variables with ;;; documentation, sorts and dumps the information in doc/festfunc.texi ;;; and doc/festvars.texi ;;; ;;; The makefile in the doc directory runs the compiled festival binary and ;;; causes these files to be created form the currently defined functions ;;; and variables ;;; ;;; Also provides function to extract manual section for documentation ;;; string and send a url to Netscape to display it ;;; (define (make-doc) "(make-doc) Find function and variable document strings and save them in texinfo format to respective files." (format t "Making function, feature and variable lists\n") ;; Need to ensure all library files are actually loaded if they contain ;; funcstions/variables which have to be put in the manual (require 'display) (require 'mbrola) (require 'tilt) (make-a-doc "festfunc.texi" 'function) (make-a-doc "festfeat.texi" 'features) (make-a-doc "festvars.texi" 'vars)) (define (make-a-doc outfile doclist) "(make-a-doc FILENAME DOCLIST) Make a texinfo document in FILENAME as a texinfo table, items are from DOCLIST. DOCLIST names which doclist to use, it may be one of 'function, 'features or 'vars." (let ((outfp (fopen outfile "wb"))) (format outfp "@table @code\n") ;; Yes I am so lazy I'm not willing to write a sort function in Scheme (sort-and-dump-docstrings doclist outfp) (format outfp "@end table\n") (fclose outfp))) ;;; ;;; Documentation string may refer to a section in the manual ;;; If it does then we can automatically go to that section in the ;;; menu using Netscape. ;;; (defvar manual-browser "netscape" "manual-browser The Unix program name of your Netscape Navigator browser. [see Getting some help]") (defvar manual-url (format nil "http://www.cstr.ed.ac.uk/projects/festival/manual-%s.%s.%s/" (car festival_version_number) (car (cdr festival_version_number)) (car (cdr (cdr festival_version_number)))) "manual-url The default URL for the Festival Manual in html format. You may reset this to a file://.../... type URL on you're local machine. [see Getting some help]") ;;; Paul got this idea from VM, the email system for emacs and ;;; I found out how to do this from their code, thanks Kyle (define (send-url-to-netscape url) "(send-url-to-netscape URL) Send given URL to netscape for display. This is primarily used to display parts of the manual referenced in documentation strings." (system (string-append manual-browser " -remote \"openURL( " url " )\" "))) (define (lastline string) "(lastline STRING) Returns the part of the string which between the last newline and the end of string." (let ((ns (string-after string "\n"))) (if (string-equal ns "") string (lastline ns)))) (define (manual-sym symbol) "(manual-sym SYMBOL) Display the section in the manual that SYMBOL's docstring has identified as the most relevant. The section is named on the last line of a documentation string with no newlines within it prefixed by \"[see \" with a \"]\" just immediately before the end of the documentation string. The manual section name is translated to the section in the HTML version of the manual and a URL is and sent to Netscape for display. [see Getting some help]" (let ((section (string-before (string-after (lastline (eval (list 'doc symbol))) "[see ") "]"))) (cond ((string-equal section "") (eval (list 'doc symbol))) ;; nothing there (t (manual section))))) (define (manual section) "(manual SECTION) Display SECTION in the manual. SECTION is a string identifying a manual section (it could be an initial substring. If SECTION is nil or unspecifed then the Manual table of contents is displayed. This uses netscape to display the manual page so you must have that (use variable manual-browser to identify it) and the variable manual-url pointing to a copy of the manual. [see Getting some help]" (let ((tmpfile (make_tmp_filename)) (manual-section)) (cond ((string-matches section "\"") (string-append "Invalid section reference containing quote: " section "\n")) ((not section) (send-url-to-netscape (string-append manual-url "festival_toc.html"))) (t ;; find section in manual (get_url (string-append manual-url "festival_toc.html") tmpfile) (system (string-append "grep -i \"^
  • .*$//' > \"" tmpfile ".out\"")) (set! manual-section (load (string-append tmpfile ".out") t)) (cond ((not manual-section) (string-append "No section called: " section)) (t (send-url-to-netscape (string-append manual-url (car manual-section))) (delete-file tmpfile) (delete-file (string-append tmpfile ".out")) "Sent manual reference url to netscape.")))))) (provide 'festdoc) festival/lib/languages.scm0000644000175000017500000001116007472506470014231 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Specification of voices and some major choices of synthesis ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; This should use some sort of database description for voices so ;;; new voices will become automatically available. ;;; (define (language_british_english) "(language_british_english) Set up language parameters for British English." (require 'voices) ;; Will get more elaborate, with different choices of voices in language (set! male1 voice_rab_diphone) (set! male2 voice_don_diphone) (if (symbol-bound? 'voice_gsw_diphone) (set! male3 voice_gsw_diphone)) (if (symbol-bound? 'voice_gsw_450) (set! male4 voice_gsw_450)) (male1) (Parameter.set 'Language 'britishenglish) ) (define (language_american_english) "(language_american_english) Set up language parameters for Aemerican English." (if (symbol-bound? 'voice_kal_diphone) (set! female1 voice_kal_diphone)) (set! male1 voice_ked_diphone) (male1) (Parameter.set 'Language 'americanenglish) ) (define (language_scots_gaelic) "(language_scots_gaelic) Set up language parameters for Scots Gaelic." (error "Scots Gaelic not yet supported.") (Parameter.set 'Language 'scotsgaelic) ) (define (language_welsh) "(language_welsh) Set up language parameters for Welsh." (set! male1 voice_welsh_hl) (male1) (Parameter.set 'Language 'welsh) ) (define (language_castillian_spanish) "(language_spanish) Set up language parameters for Castillian Spanish." (voice_el_diphone) (set! male1 voice_el_diphone) (Parameter.set 'Language 'spanish) ) (define (select_language language) (cond ((or (equal? language 'britishenglish) (equal? language 'english)) ;; we all know its the *real* English (language_british_english)) ((equal? language 'americanenglish) (language_american_english)) ((equal? language 'scotsgaelic) (language_scots_gaelic)) ((equal? language 'welsh) (language_welsh)) ((equal? language 'spanish) (language_castillian_spanish)) ((equal? language 'klingon) (language_klingon)) (t (print "Unsupported language, using English") (language_british_english)))) (defvar language_default language_british_english) (provide 'languages) festival/lib/token.scm0000644000175000017500000005676610606447120013415 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Various tokenizing functions and customization (define (Token utt) "(Token UTT) Build a Word stream from the Token stream, analyzing compound words numbers etc as tokens into words. Respects the Parameter Language to choose the appropriate token to word module." (let ((rval (apply_method 'Token_Method utt)) ;; might be defined (language (Parameter.get 'Language))) (cond (rval rval) ;; newer style ((or (string-equal "britishenglish" language) (string-equal "english" language) (string-equal "americanenglish" language)) (Token_English utt)) ((string-equal "welsh" language) (Token_Welsh utt)) (t (Token_Any utt))))) (define (remove_leadtrail_underscores name) "(remove_leadtrail_underscores name) Get rid of leading and trailing underscores that may be used for emphasis, not this is called when there are underscores at the beginning and end but there may not be an equal number of them." (let ((se (symbolexplode name))) (while (string-equal "_" (car se)) (set! se (cdr se))) (set! se (reverse se)) (while (string-equal "_" (car se)) (set! se (cdr se))) (apply string-append (reverse se)))) (define (english_token_to_words token name) "(english_token_to_words TOKEN NAME) Returns a list of words for NAME from TOKEN. This allows the user to customize various non-local, multi-word, context dependent translations of tokens into words. If this function is unset only the builtin translation rules are used, if this is set the builtin rules are not used unless explicitly called. [see Token to word rules]" (cond ((string-matches name "[A-Z]*[\\$#\\\\Y£][0-9,]+\\(\\.[0-9]+\\)?") ;; Some form of money (pounds or type of dollars) (let (amount type currency) (cond ((string-matches name ".*\\$.*") (set! amount (string-after name "$")) (set! type (string-before name "$")) (set! currency "dollar")) ((string-matches name ".*£.*") (set! amount (string-after name "£")) (set! type (string-before name "£")) (set! currency "pound")) ((string-matches name ".*#.*") (set! amount (string-after name "#")) (set! type (string-before name "#")) (set! currency "pound")) ((string-matches name ".*Y[0-9].*") (set! amount (string-after name "Y")) (set! type (string-before name "Y")) (set! currency "yen")) ((string-matches name ".*\\\\.*") (set! amount (string-after name "\\")) (set! type (string-before name "\\")) (set! currency "yen")) (t ;; who knows (set! amount (string-after name "$")) (set! type (string-before name "$")) (set! currency "dollar"))) (cond ((string-matches (item.feat token "n.name") ".*illion.?") (append ;; "billions and billions" - Sagan (builtin_english_token_to_words token amount) (list (item.feat token "n.name")) ;; illion (token_money_expand type) (list (string-append currency "s")))) ((string-matches amount ".*\\...$") (append ;; exactly two places after point (builtin_english_token_to_words token (string-before amount ".")) (token_money_expand type) (if (or (string-matches amount "1\\..*") (string-equal currency "yen")) (list currency) (list (string-append currency "s"))) (if (not (string-matches name ".*\\.00$")) (builtin_english_token_to_words token (remove_leading_zeros (string-after amount "."))) nil))) (t (append ;; nothing after point or lots after point (builtin_english_token_to_words token amount) (token_money_expand type) (if (or (string-matches amount "1") (string-equal currency "yen")) (list currency) (list (string-append currency "s")))))))) ((and (string-matches name ".*illion.?") (string-matches (item.feat token "p.name") "[A-Z]*[\\$#][0-9,]+\\(\\.[0-9]+\\)?")) nil ;; dealt with on the previous symbol ) ((string-matches name "[1-9][0-9]*/[1-9][0-9]*") (let ((numerator (string-before name "/")) (denominator (string-after name "/")) ) (cond ((string-matches name "1/2") (list "half")) ((string-matches denominator "4") (append (builtin_english_token_to_words token numerator) (list "quarter") (if (string-equal numerator "1") (list '((name "'s")(pos nnp))) nil))) (t (append (builtin_english_token_to_words token numerator) (begin (item.set_feat token "token_pos" "ordinal") (builtin_english_token_to_words token denominator)) (if (string-equal numerator "1") nil (list '((name "'s")(pos nnp))))))))) ((and (string-matches name "No") (item.next token) (string-matches (item.feat token "n.name") "[0-9]+")) (list "number")) ((string-matches name ".*%$") (append (token_to_words token (string-before name "%")) (list "percent"))) ((string-matches name "[0-9]+s") ;; e.g. 1950s (item.set_feat token "token_pos" "year") ;; reasonable guess (append (builtin_english_token_to_words token (string-before name "s")) (list '((name "'s")(pos nnp))) ;; will get assimilated by postlexical rules )) ((string-matches name "[0-9]+'s") ;; e.g. 1950's (item.set_feat token "token_pos" "year") ;; reasonable guess (append (builtin_english_token_to_words token (string-before name "'s")) (list '((name "'s")(pos nnp))) ;; will get assimilated by postlexical rules )) ((and (string-matches name ".*s$") (string-equal (item.feat token "punc") "'")) ;; potential possessive or may be end of a quote (if (token_no_starting_quote token) (item.set_feat token "punc" "")) (builtin_english_token_to_words token name)) ((and (string-equal name "A") ;; letter or determiner (or (string-matches (item.feat token "p.name") "[a-z].*") (string-matches (item.feat token "n.name") "[A-Z].*"))) (list (list '(name "a")(list 'pos token.letter_pos)))) ((member_string name english_homographs) (list (list (list 'name name) (list 'hg_pos (item.feat token "token_pos"))))) ((string-matches name "__*[^_][^_]*_*_") ;; _emphasis_ (english_token_to_words token (remove_leadtrail_underscores name) )) ((string-matches name "[0-9]?[0-9][:\\.][0-9][0-9][AaPp][Mm]") ;; time ;; must be am/pm present for . to be acceptable separator (let (hours mins half sep (ttime (downcase name))) (if (string-matches ttime ".*:.*") (set! sep ":") (set! sep ".")) (set! hours (string-before ttime sep)) (set! mins (string-after ttime sep)) (if (string-matches ttime ".*am") (set! sep "am") (set! sep "pm")) (set! mins (string-before mins sep)) (append (builtin_english_token_to_words token hours) (cond ((string-equal mins "00") nil) ((string-matches mins "0.") (cons "oh" (builtin_english_token_to_words token (string-after mins "0")))) (t (builtin_english_token_to_words token mins))) (if (string-equal sep "am") (builtin_english_token_to_words token "A.M") (builtin_english_token_to_words token "P.M"))))) ((string-matches name "[0-9]?[0-9]:[0-9][0-9]") ;; time (append (builtin_english_token_to_words token (remove_leading_zeros (string-before name ":"))) (cond ((string-equal "00" (string-after name ":")) nil) ((string-matches (string-after name ":") "0.") (cons "oh" (builtin_english_token_to_words token (remove_leading_zeros (string-after name ":"))))) (t (builtin_english_token_to_words token (string-after name ":")))))) ((string-matches name "[0-9][0-9]:[0-9][0-9]:[0-9][0-9]") ;; exact time (append (builtin_english_token_to_words token (remove_leading_zeros (string-before name ":"))) (list "hours") (builtin_english_token_to_words token (remove_leading_zeros (string-before (string-after name ":") ":"))) (list "minutes" "and") (builtin_english_token_to_words token (remove_leading_zeros (string-after (string-after name ":") ":"))) (list "seconds"))) ((string-matches name "[0-9][0-9]?/[0-9][0-9]?/[0-9][0-9]\\([0-9][0-9]\\)?") ;; date, say it as numbers to avoid American/British problem (let ((num1 (string-before name "/")) (num2 (string-before (string-after name "/") "/")) (year (string-after (string-after name "/") "/")) day month) (item.set_feat token "token_pos" "cardinal") (set! day (builtin_english_token_to_words token num1)) (set! month (builtin_english_token_to_words token num2)) (item.set_feat token "token_pos" "year") (append day month (list '((name ",")(pbreak_scale 0.9))) (builtin_english_token_to_words token year)))) ((string-matches name "[0-9][0-9][0-9]-[0-9][0-9][0-9][0-9]") (item.set_feat token "token_pos" "digits") ;; canonical phone number (append (builtin_english_token_to_words token (string-before name "-")) (list '((name ",")(pbreak_scale 1.0))) (builtin_english_token_to_words token (string-after name "-")))) ((string-matches name "[0-9]+-[0-9]+-[-0-9]+") ;; long distance number (let ((r '(dummy)) (remainder name)) (item.set_feat token "token_pos" "digits") (while (> (length remainder) 0) (if (string-matches remainder "[0-9]+") (set! r (append r (builtin_english_token_to_words token remainder))) (set! r (append r (builtin_english_token_to_words token (string-before remainder "-"))))) (set! remainder (string-after remainder "-")) (if (> (length remainder) 0) (set! r (append r (list '((name ",")(pbreak_scale 1.0))))))) (cdr r)) ) ((and (string-matches name "[0-9][0-9][0-9]") (string-matches (item.feat token "n.name") "[0-9][0-9][0-9]-[0-9][0-9][0-9][0-9]")) (item.set_feat token "token_pos" "digits") (builtin_english_token_to_words token name)) ((string-matches name "[0-9]+-[0-9]+") (let ((tokpos)) (item.set_name token (string-before name "-")) (set! tokpos (wagon token (car (cdr (assoc "[0-9]+" token_pos_cart_trees))))) (item.set_feat token "token_pos" (car tokpos)) (append (builtin_english_token_to_words token (string-before name "-")) (list "to") (builtin_english_token_to_words token (string-after name "-"))))) ((string-matches name "\\(II?I?\\|IV\\|VI?I?I?\\|IX\\|X[VIX]*\\)") ;; Roman numerals (let ((tp (item.feat token "token_pos"))) (cond ((string-matches tp "century");; always believe this (item.set_feat token "token_pos" "ordinal") (if (or (string-equal "1" (tok_rex token)) (item.feat token "p.lisp_tok_rex_names")) (append (list "the") (builtin_english_token_to_words token (tok_roman_to_numstring name))) (builtin_english_token_to_words token (tok_roman_to_numstring name)))) ((string-matches name "[IVX]");; be *very* wary of this one (if (and (string-equal "1" (item.feat token "p.lisp_tok_section_name")) (string-matches tp "number")) (builtin_english_token_to_words token (tok_roman_to_numstring name)) (tok_string_as_letters name))) ((string-matches tp "number") (item.set_feat token "token_pos" "cardinal") (builtin_english_token_to_words token (tok_roman_to_numstring name))) (t;; else its a letter (tok_string_as_letters name))))) ((and (string-matches name "pp") (string-matches (item.feat token "n.name") "[0-9]+-[0-9]+")) (list "pages")) ((and (string-matches name "ss") (string-matches (item.feat token "n.name") "[0-9]+-[0-9]+")) (list "sections")) ((string-matches name "_____+") (list "line" "of" "underscores")) ((string-matches name "=====+") (list "line" "of" "equals")) ((string-matches name "-----+") (list "line" "of" "hyphens")) ((string-matches name "\\*\\*\\*\\*\\*+") (list "line" "of" "asterisks")) ((string-matches name "--+") (list '((name ",")(pbreak_scale 1.0)))) ((string-matches name ".*--+.*") (append (builtin_english_token_to_words token (string-before name "--")) (list '((name ",")(pbreak_scale 1.0))) (builtin_english_token_to_words token (string-after name "--")))) ((string-matches name "[A-Z][A-Z]?&[A-Z][A-Z]?") (append (tok_string_as_letters (string-before name "&")) (list "and") (tok_string_as_letters (string-after name "&")))) ((and (string-equal name "Ms") (string-matches (item.feat token "n.name") "[A-Z][^A-Z]*")) (list "mizz")) ((or (string-matches name "[A-Z][A-Z]+s") (string-matches name "[BCDEFGHJKLMNOPQRSTVWXYZ]+s")) (append (builtin_english_token_to_words token (string-before name "s")) (list '((name "'s")(pos nnp))) ;; will get assimilated by postlexical rules )) ((string-matches name "<.*@.*>") ;; quoted e-mail (append (builtin_english_token_to_words token (string-after (string-before name "@") "<")) (list "at") (builtin_english_token_to_words token (string-before (string-after name "@") ">")))) ((string-matches name ".*@.*") ;; e-mail (append (builtin_english_token_to_words token (string-before name "@")) (list "at") (builtin_english_token_to_words token (string-after name "@") ">"))) ((string-matches name "\\([dD][Rr]\\|[Ss][tT]\\)") (if (string-equal (item.feat token "token_pos") "street") (if (string-matches name "[dD][rR]") (list "drive") (list "street")) (if (string-matches name "[dD][rR]") ;; default on title side (list "doctor") (list "saint")))) ((string-matches name "[Cc]alif") ;; hopelessly specific ... (list "california")) (t (builtin_english_token_to_words token name)))) ;;; This is set as the default (defvar token_to_words english_token_to_words) (defvar token.punctuation "\"'`.,:;!?(){}[]" "token.punctuation A string of characters which are to be treated as punctuation when tokenizing text. Punctuation symbols will be removed from the text of the token and made available through the \"punctuation\" feature. [see Tokenizing]") (defvar token.prepunctuation "\"'`({[" "token.prepunctuation A string of characters which are to be treated as preceding punctuation when tokenizing text. Prepunctuation symbols will be removed from the text of the token and made available through the \"prepunctuation\" feature. [see Tokenizing]") (defvar token.whitespace " \t\n\r" "token.whitespace A string of characters which are to be treated as whitespace when tokenizing text. Whitespace is treated as a separator and removed from the text of a token and made available through the \"whitespace\" feature. [see Tokenizing]") (defvar token.singlecharsymbols "" "token.singlecharsymbols Characters which have always to be split as tokens. This would be usual is standard text, but is useful in parsing some types of file. [see Tokenizing]") (defvar token.letter_pos 'nn "token.letter_pos The part of speech tag (valid for your part of speech tagger) for individual letters. When the tokenizer decide to pronounce a token as a list of letters this tag is added to each letter in the list. Note this should be from the part of speech set used in your tagger which may not be the same one that appears in the actual lexical entry (if you map them afterwards). This specifically allows \"a\" to come out as ae rather than @.") (defvar token.unknown_word_name "unknown" "token.unknown_word_name When all else fails and a pronunciation for a word or character can't be found this word will be said instead. If you make this \"\" them the unknown word will simple be omitted. This will only really be called when there is a bug in the lexicon and characters are missing from the lexicon. Note this word should be in the lexicon.") (def_feature_docstring 'Token.punc "Token.punc Succeeding punctuation symbol found after token in original string/file.") (def_feature_docstring 'Token.whitespace "Token.whitespace Whitespace found before token in original string/file.") (def_feature_docstring 'Token.prepunctuation "Token.prepunctuation Preceeding puctuation symbol found before token in original string/file.") (require 'tokenpos) ;;; ;;; Token pos are gross level part of speech tags which help decide ;;; pronunciation of tokens (particular expansion of Tokens into words) ;;; The most obvious example is identifying number types (ordinals, ;;; years, digits or numbers). ;;; (defvar english_token_pos_cart_trees '( ;; Format is (Regex Tree) ("[0-9]+" ((lisp_num_digits < 3.8) ((p.lisp_token_pos_guess is month) ((lisp_month_range is 0) ((year)) ((ordinal))) ((n.lisp_token_pos_guess is month) ((lisp_month_range is 0) ((cardinal)) ((ordinal))) ((n.lisp_token_pos_guess is numeric) ((lisp_num_digits < 2) ((p.lisp_token_pos_guess is numeric) ((pp.lisp_token_pos_guess is sym) ((digits)) ((cardinal))) ((cardinal))) ((nn.lisp_token_pos_guess is sym) ((cardinal)) ((digits)))) ((lisp_num_digits < 2) ((nn.lisp_token_pos_guess is numeric) ((n.lisp_token_pos_guess is sym) ((lisp_month_range is 0) ((digits)) ((cardinal))) ((cardinal))) ((cardinal))) ((name < 302.3) ((p.lisp_token_pos_guess is flight) ((digits)) ((n.lisp_token_pos_guess is sym) ((p.lisp_token_pos_guess is sym) ((digits)) ((cardinal))) ((cardinal)))) ((p.lisp_token_pos_guess is a) ((digits)) ((n.lisp_token_pos_guess is sym) ((nn.lisp_token_pos_guess is sym) ((name < 669.2) ((digits)) ((cardinal))) ((cardinal))) ((name < 373.2) ((cardinal)) ((name < 436.2) ((name < 392.6) ((digits)) ((cardinal))) ((name < 716.5) ((cardinal)) ((name < 773.6) ((p.lisp_token_pos_guess is _other_) ((digits)) ((cardinal))) ((cardinal))))))))))))) ((p.lisp_token_pos_guess is numeric) ((pp.lisp_token_pos_guess is month) ((year)) ((nn.lisp_token_pos_guess is numeric) ((cardinal)) ((digits)))) ((nn.lisp_token_pos_guess is numeric) ((n.lisp_token_pos_guess is month) ((cardinal)) ((n.lisp_token_pos_guess is numeric) ((digits)) ((p.lisp_token_pos_guess is _other_) ((cardinal)) ((year))))) ((p.lisp_token_pos_guess is _other_) ((lisp_num_digits < 4.4) ((name < 2959.6) ((name < 1773.4) ((cardinal)) ((year))) ((cardinal))) ((pp.lisp_token_pos_guess is _other_) ((digits)) ((cardinal)))) ((n.lisp_token_pos_guess is to) ((year)) ((p.lisp_token_pos_guess is sym) ((pp.lisp_token_pos_guess is sym) ((cardinal)) ((lisp_num_digits < 4.6) ((year)) ((digits)))) ((lisp_num_digits < 4.8) ((name < 2880) ((name < 1633.2) ((name < 1306.4) ((cardinal)) ((year))) ((year))) ((cardinal))) ((cardinal))))))))) ) ("\\(II?I?\\|IV\\|VI?I?I?\\|IX\\|X[VIX]*\\)";; Roman numerals ((p.lisp_tok_rex_names is 0) ((lisp_num_digits is 5) ((number)) ((lisp_num_digits is 4) ((number)) ((nn.lisp_num_digits is 13) ((number)) ((p.lisp_num_digits is 7) ((number)) ((p.lisp_tok_section_name is 0) ((lisp_tok_rex is 0) ((lisp_num_digits is 3) ((p.lisp_num_digits is 4) ((number)) ((nn.lisp_num_digits is 4) ((number)) ((n.lisp_num_digits is 4) ((number)) ((pp.lisp_num_digits is 3) ((number)) ((p.lisp_num_digits is 2) ((letter)) ((nn.lisp_num_digits is 2) ((letter)) ((n.cap is 0) ((letter)) ((number))))))))) ((nn.lisp_num_digits is 11) ((letter)) ((lisp_num_digits is 1) ((pp.lisp_num_digits is 9) ((letter)) ((p.lisp_num_digits is 9) ((letter)) ((n.lisp_num_digits is 6) ((letter)) ((pp.lisp_num_digits is 6) ((letter)) ((pp.cap is 0) ((n.cap is 0) ((p.lisp_num_digits is 1) ((letter)) ((n.lisp_num_digits is 4) ((letter)) ((letter)))) ((letter))) ((letter))))))) ((p.lisp_num_digits is 10) ((number)) ((n.lisp_num_digits is 8) ((number)) ((pp.lisp_num_digits is 9) ((number)) ((nn.lisp_num_digits is 5) ((number)) ((n.lisp_num_digits is 4) ((number)) ((letter)))))))))) ((letter))) ((number))))))) ((century)))) ("\\([dD][Rr]\\|[Ss][tT]\\)" ((n.name is 0) ((p.cap is 1) ((street)) ((p.name matches "[0-9]*\\(1[sS][tT]\\|2[nN][dD]\\|3[rR][dD]\\|[0-9][tT][hH]\\)") ((street)) ((title)))) ((punc matches ".*,.*") ((street)) ((p.punc matches ".*,.*") ((title)) ((n.cap is 0) ((street)) ((p.cap is 0) ((p.name matches "[0-9]*\\(1[sS][tT]\\|2[nN][dD]\\|3[rR][dD]\\|[0-9][tT][hH]\\)") ((street)) ((title))) ((pp.name matches "[1-9][0-9]+") ((street)) ((title))))))))) ("lead" ((p.name in (was were had been having has is are)) ((led)) ((liid)))) ("read" ((p.name in (to)) ((riid)) ((red)))) )) (defvar english_homographs '("lead" "read") "english_homographs A list of tokens that are dealt with by a homograph disambiguation tree in english_token_pos_cart_trees.") (defvar token_pos_cart_trees english_token_pos_cart_trees "token_pos_cart_trees This is a list of pairs or regex plus CART tree. Tokens that match the regex will have the CART tree aplied, setting the result as the token_pos feature on the token. The list is checked in order and only the first match will be applied.") (provide 'token) festival/lib/mbrola.scm0000644000175000017500000001115107472506470013537 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Support for MBROLA as an external module. ;;; ;;; You might want to set this in your sitevars.scm (defvar mbrola_progname "/cstr/external/mbrola/mbrola" "mbrola_progname The program name for mbrola.") (defvar mbrola_database "fr1" "mbrola_database The name of the MBROLA database to usde during MBROLA Synthesis.") (define (MBROLA_Synth utt) "(MBROLA_Synth UTT) Synthesize using MBROLA as external module. Basically dump the info from this utterance. Call MBROLA and reload the waveform into utt. [see MBROLA]" (let ((filename (make_tmp_filename)) ) (save_segments_mbrola utt filename) (system (string-append mbrola_progname " " mbrola_database " " filename " " filename ".au")) (utt.import.wave utt (string-append filename ".au")) (apply_hooks after_synth_hooks utt) (delete-file filename) (delete-file (string-append filename ".au")) utt)) (define (save_segments_mbrola utt filename) "(save_segments_mbrola UTT FILENAME) Save segment information in MBROLA format in filename. The format is phone duration (ms) [% position F0 target]*. [see MBROLA]" (let ((fd (fopen filename "w"))) (mapcar (lambda (segment) (save_seg_mbrola_entry (item.feat segment 'name) (item.feat segment 'segment_start) (item.feat segment 'segment_duration) (mapcar (lambda (targ_item) (list (item.feat targ_item "pos") (item.feat targ_item "f0"))) (item.relation.daughters segment 'Target)) ;; list of targets fd)) (utt.relation.items utt 'Segment)) (fclose fd))) (define (save_seg_mbrola_entry name start dur targs fd) "(save_seg_mbrola_entry ENTRY NAME START DUR TARGS FD) Entry contains, (name duration num_targs start 1st_targ_pos 1st_targ_val)." (format fd "%s %d " name (nint (* dur 1000))) (if targs ;; if there are any targets (mapcar (lambda (targ) ;; targ_pos and targ_val (let ((targ_pos (car targ)) (targ_val (car (cdr targ)))) (format fd "%d %d " (nint (* 100 (/ (- targ_pos start) dur))) ;; % pos of target (nint (parse-number targ_val))) ;; target value )) targs)) (terpri fd) (terpri fd) ) (provide 'mbrola) festival/lib/display.scm0000644000175000017500000000735707262605650013742 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: December 1996 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; An xwaves display function for utterances ;;; ;;; Requires Xwaves to be running, saves labels etc and sends ;;; messages to Xwaves to display the utterance. ;;; ;;; This can be a model for other display functions. ;;; (define (display utt) "(display utt) Display an utterance's waveform, F0 and segment labels in Xwaves. Xwaves must be running on the current machine, with a labeller for this to work." (let ((tmpname (make_tmp_filename))) (utt.save.wave utt (string-append tmpname ".wav")) (utt.save.segs utt (string-append tmpname ".lab")) (utt.save.f0 utt (string-append tmpname ".f0")) (system (format nil "send_xwaves make file %s name %s height 150" (string-append tmpname ".f0") tmpname)) (system (format nil "send_xwaves make name %s file %s height 200" tmpname (string-append tmpname ".wav"))) (system (format nil "send_xwaves send make file %s name %s color 125" (string-append tmpname ".lab") tmpname)) (system (format nil "send_xwaves send activate name %s fields 1" tmpname)) (system (format nil "send_xwaves %s align file %s" tmpname (string-append tmpname ".wav")))) ) (provide 'display) festival/lib/postlex.scm0000644000175000017500000006221211415076550013756 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Postlexical rules ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; Modifed for CSTR HTS Voice Library ;; ;; Author : Junichi Yamagishi (jyamagis@inf.ed.ac.uk) ;; ;; Date : Sept 2008 ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (PostLex utt) "(PostLex utt) Apply post lexical rules to segment stream. These may be almost arbitrary rules as specified by the particular voice, through the postlex_hooks variable. A number of standard post lexical rule sets are provided including reduction, posessives etc. These rules are also used to mark standard segments with their cluster information used in creating diphone names." (let ((rval (apply_method 'PostLex_Method utt))) (cond (rval rval) ;; new style (t ;; should only really need this one (apply_hooks postlex_rules_hooks utt))) utt )) (define (Classic_PostLex utt) "(Classic_PostLex utt) Apply post lexical rules (both builtin and those specified in postlex_rules_hooks)." (Builtin_PostLex utt) ;; haven't translated all the rules yet (apply_hooks postlex_rules_hooks utt) utt ) (defvar postlex_rules_hooks nil "postlex_rules_hooks A function or list of functions which encode post lexical rules. This will be voice specific, though some rules will be shared across languages.") ;;; Mapping of full vowels to reduced vowels, this should be part ;;; of the phoneset definitions (defvar postlex_vowel_reduce_table '((mrpa ((uh @) (i @) (a @) (e @) (u @) (o @) (oo @))) (radio ((ah ax el en em) (ih ax) ; (er axr ax) ; (iy ih) ; (ey ax) (aa ax) (ae ax) (eh ax)))) "postlex_vowel_reduce_table Mapping of vowels to their reduced form. This in an assoc list of phoneset name to an assoc list of full vowel to reduced form.") (defvar postlex_vowel_reduce_cart_tree nil "postlex_vowel_reduce_cart_tree CART tree for vowel reduction.") (defvar postlex_vowel_reduce_cart_tree_hand '((stress is 0) ((p.syl_break < 2) ((syl_break < 2) ((1)) ((0))) ((0))) ((0))) "postlex_vowel_reduce_cart_tree_hand A CART tree for vowel reduction. This is hand-written.") (defvar postlex_vowel_reduce_cart_data ' ((R:SylStructure.parent.gpos is cc) (((0 0.993548) (1 0.00645161) 0)) ((p.R:SylStructure.parent.gpos is md) (((0 0.903226) (1 0.0967742) 0)) ((p.R:SylStructure.parent.gpos is det) ((n.R:SylStructure.parent.gpos is content) ((last_accent < 2.5) ((next_accent < 2.5) ((next_accent < 1.2) ((n.syl_break is 4) (((0 0.967213) (1 0.0327869) 0)) ((syl_break is 4) (((0 0.952381) (1 0.047619) 0)) ((n.syl_break is 4) (((0 0.953488) (1 0.0465116) 0)) ((position_type is single) (((0 0.947368) (1 0.0526316) 0)) ((accented is 0) ((n.accented is 0) (((0 0.857143) (1 0.142857) 0)) (((0 0.415385) (1 0.584615) 1))) (((0 0.974359) (1 0.025641) 0))))))) (((0 0.968254) (1 0.031746) 0))) (((0 0.969697) (1 0.030303) 0))) (((0 0.976744) (1 0.0232558) 0))) (((0 0.990291) (1 0.00970874) 0))) ((next_accent < 108.5) ((p.R:SylStructure.parent.gpos is pps) (((0 0.828947) (1 0.171053) 0)) ((R:SylStructure.parent.gpos is det) ((accented is 0) (((0 0.0599572) (1 0.940043) 1)) (((0 0.949367) (1 0.0506329) 0))) ((p.R:SylStructure.parent.gpos is cc) (((0 0.880952) (1 0.119048) 0)) ((p.R:SylStructure.parent.gpos is wp) (((0 0.875) (1 0.125) 0)) ((p.R:SylStructure.parent.gpos is in) ((n.syl_break is 4) (((0 0.961538) (1 0.0384615) 0)) ((next_accent < 2.5) ((syl_break is 4) (((0 0.95122) (1 0.0487805) 0)) ((next_accent < 1.2) ((accented is 0) ((n.stress is 0) (((0 0.788462) (1 0.211538) 0)) ((R:SylStructure.parent.R:Word.p.gpos is content) (((0 0.863636) (1 0.136364) 0)) ((position_type is single) (((0 0.729167) (1 0.270833) 0)) (((0 0.4) (1 0.6) 1))))) (((0 0.983871) (1 0.016129) 0))) (((0 0.96) (1 0.04) 0)))) (((0 0.963636) (1 0.0363636) 0)))) ((position_type is single) ((syl_break is 4) (((0 0.993865) (1 0.00613497) 0)) ((p.R:SylStructure.parent.gpos is to) (((0 0.984375) (1 0.015625) 0)) ((syl_break is 1) ((accented is 0) ((n.R:SylStructure.parent.gpos is in) (((0 0.869565) (1 0.130435) 0)) ((R:SylStructure.parent.gpos is content) (((0 0.861789) (1 0.138211) 0)) ((p.R:SylStructure.parent.gpos is content) ((p.syl_break is 4) (((0 0.858065) (1 0.141935) 0)) ((R:SylStructure.parent.gpos is in) ((p.syl_break is 1) ((n.R:SylStructure.parent.gpos is det) (((0 0.659574) (1 0.340426) 0)) ((p.stress is 0) (((0 0.422222) (1 0.577778) 1)) (((0 0.582278) (1 0.417722) 0)))) ((n.accented is 0) ((n.R:SylStructure.parent.gpos is content) (((0 0.65) (1 0.35) 0)) ((p.stress is 0) (((0 0.464286) (1 0.535714) 1)) (((0 0.538462) (1 0.461538) 0)))) (((0 0.803279) (1 0.196721) 0)))) ((n.R:SylStructure.parent.gpos is det) (((0 0.952381) (1 0.047619) 0)) ((n.syl_break is 4) (((0 0.833333) (1 0.166667) 0)) ((p.stress is 0) ((p.syl_break is 1) ((n.syl_break is 1) (((0 0.740741) (1 0.259259) 0)) ((R:SylStructure.parent.gpos is aux) (((0 0.478261) (1 0.521739) 1)) (((0 0.769231) (1 0.230769) 0)))) (((0 0.755556) (1 0.244444) 0))) (((0 0.797619) (1 0.202381) 0))))))) (((0 0.870968) (1 0.129032) 0))))) (((0 0.983806) (1 0.0161943) 0))) (((0 0.977778) (1 0.0222222) 0))))) ((next_accent < 21.6) ((p.stress is 0) ((R:SylStructure.parent.R:Word.p.gpos is md) (((0 0.961538) (1 0.0384615) 0)) ((position_type is mid) (((0 0.977612) (1 0.0223881) 0)) ((n.R:SylStructure.parent.gpos is det) (((0 0.916667) (1 0.0833333) 0)) ((R:SylStructure.parent.R:Word.n.gpos is 0) (((0 0.915493) (1 0.084507) 0)) ((R:SylStructure.parent.R:Word.n.gpos is pps) (((0 0.884615) (1 0.115385) 0)) ((n.stress is 0) ((n.syl_break is 4) (((0 0.986755) (1 0.013245) 0)) ((p.syl_break is 4) (((0 0.977011) (1 0.0229885) 0)) ((n.syl_break is 4) (((0 0.965517) (1 0.0344828) 0)) ((last_accent < 1.2) ((last_accent < 0.1) (((0 0.910448) (1 0.0895522) 0)) ((next_accent < 1.2) ((R:SylStructure.parent.R:Word.n.gpos is in) (((0 0.82) (1 0.18) 0)) ((n.syl_break is 0) ((R:SylStructure.parent.R:Word.p.gpos is content) (((0 0.819672) (1 0.180328) 0)) (((0 0.444444) (1 0.555556) 1))) (((0 0.785714) (1 0.214286) 0)))) (((0 0.836364) (1 0.163636) 0)))) (((0 0.962025) (1 0.0379747) 0)))))) ((stress is 0) ((n.syl_break is 4) (((0 0.21875) (1 0.78125) 1)) ((R:SylStructure.parent.R:Word.p.gpos is aux) (((0 0.259259) (1 0.740741) 1)) ((p.syl_break is 1) (((0 0.243094) (1 0.756906) 1)) ((R:SylStructure.parent.R:Word.p.gpos is det) (((0 0.290323) (1 0.709677) 1)) ((R:SylStructure.parent.R:Word.p.gpos is in) (((0 0.3) (1 0.7) 1)) ((syl_break is 1) (((0 0.289157) (1 0.710843) 1)) ((p.syl_break is 4) (((0 0.352941) (1 0.647059) 1)) ((n.syl_break is 0) (((0 0.311475) (1 0.688525) 1)) ((syl_break is 4) (((0 0.4) (1 0.6) 1)) (((0 0.581395) (1 0.418605) 0))))))))))) (((0 1) (1 0) 0))))))))) ((stress is 0) ((R:SylStructure.parent.R:Word.n.gpos is 0) (((0 0.121212) (1 0.878788) 1)) ((next_accent < 2.4) ((R:SylStructure.parent.gpos is content) ((position_type is mid) (((0 0.176895) (1 0.823105) 1)) ((p.syl_break is 1) (((0 0.229167) (1 0.770833) 1)) ((syl_break is 4) (((0 0.242775) (1 0.757225) 1)) ((p.syl_break is 0) ((n.R:SylStructure.parent.gpos is in) (((0 0.253521) (1 0.746479) 1)) ((R:SylStructure.parent.R:Word.p.gpos is in) (((0 0.262774) (1 0.737226) 1)) ((last_accent < 2.1) ((n.R:SylStructure.parent.gpos is aux) (((0 0.304348) (1 0.695652) 1)) ((next_accent < 1.2) ((n.R:SylStructure.parent.gpos is cc) (((0 0.291667) (1 0.708333) 1)) ((syl_break is 1) ((n.syl_break is 4) (((0 0.344828) (1 0.655172) 1)) ((R:SylStructure.parent.R:Word.p.gpos is det) (((0 0.364706) (1 0.635294) 1)) ((n.syl_break is 4) (((0 0.384615) (1 0.615385) 1)) ((last_accent < 1.2) ((p.accented is 0) (((0 0.584906) (1 0.415094) 0)) ((n.accented is 0) ((R:SylStructure.parent.R:Word.p.gpos is content) (((0 0.41) (1 0.59) 1)) (((0 0.6) (1 0.4) 0))) (((0 0.333333) (1 0.666667) 1)))) (((0 0.380952) (1 0.619048) 1)))))) ((p.accented is 0) (((0 0.183673) (1 0.816327) 1)) ((n.R:SylStructure.parent.gpos is content) ((n.stress is 0) (((0 0.295455) (1 0.704545) 1)) ((R:SylStructure.parent.R:Word.p.gpos is content) ((n.syl_break is 1) (((0 0.5) (1 0.5) 0)) (((0 0.40625) (1 0.59375) 1))) (((0 0.333333) (1 0.666667) 1)))) (((0 0.2) (1 0.8) 1)))))) (((0 0.3) (1 0.7) 1)))) (((0 0.302326) (1 0.697674) 1))))) (((0 0.25) (1 0.75) 1)))))) (((0 0.173913) (1 0.826087) 1))) (((0 0.166667) (1 0.833333) 1)))) (((0 1) (1 0) 0)))) (((0 0.2) (1 0.8) 1))))))))) (((0 0.15) (1 0.85) 1))))))) (defvar postlex_mrpa_r_cart_tree '((name is r) ((R:Segment.n.ph_vc is -) ((delete)) ((nil))) ((nil))) "postlex_mrpa_r_cart_tree For remove final R when not between vowels.") ;; Changed this to actually work... (Rob 09/12/04) ;; Changed this to delete the syllable when schwa is unneccesary (awb 19/07/04) (define (postlex_apos_s_check utt) "(postlex_apos_s_check UTT) Deal with possesive s for English (American and British). Delete schwa of 's if previous is not an alveolar or palatal fricative or affricative, and change voiced to unvoiced s if previous is not voiced." (mapcar (lambda (syl) ; word is 's (if (string-equal "'s" (item.feat syl "R:SylStructure.parent.name")) (begin ;; de-voice if last phone of previous word is unvoiced (if (string-equal "-" (item.feat syl "p.R:SylStructure.daughtern.ph_cvox")) (item.set_name (item.relation.daughtern syl 'SylStructure) "s")) ;; change it from "z" to "s" ; if the previous seg is a aveolar or palatal, ; fricative or affricate don't delete schwa otherwise delete it (if (and (member_string (item.feat syl "p.R:SylStructure.daughtern.ph_ctype") '(f a)) (member_string (item.feat syl "p.R:SylStructure.daughtern.ph_cplace") '(a p))) (begin t) (begin ;; delete the schwa (item.delete (item.relation.daughter1 syl 'SylStructure)) ;; attach orphaned s/z to previous word (item.relation.append_daughter (item.prev syl) 'SylStructure (item.relation.daughtern syl 'SylStructure)) ;; delete the now empty syllable (item.delete syl)))))) ;; never happens to if 's is first in an utterance (cdr (utt.relation.items utt 'Syllable))) utt) ;; Changed this to work the other way round, too. Volker 10/08/06 (define (postlex_the_vs_thee utt) "(postlex_the_vs_thee utt) Unnreduce the schwa in \"the\" when a vowel follows. Reduce the vowel in \"the\" when no vowel follows (this requires a lexicon entry for \"the\" with feature \"reduced\", otherwise there will be no reduction)." (let ((fullform (cadr (car (caar (cdr (cdar (lex.lookup_all 'thee))))))) (reducedform (cadr(car(caar(cddr(lex.lookup 'the '(reduced))))))) seg) (mapcar (lambda (word) (if (string-equal "the" (downcase (item.feat word "name"))) (begin (set! seg (item.relation (item.daughtern (item.relation.daughtern word 'SylStructure)) 'Segment)) (if (string-equal "+" (item.feat (item.next seg) 'ph_vc)) (item.set_feat seg 'name fullform) (item.set_feat seg 'name reducedform))))) (utt.relation.items utt 'Word))) utt) (define (postlex_the_vs_thee_changeflag utt) "(postlex_the_vs_thee_changeflag utt) Unnreduce the schwa in \"the\" when a vowel follows. Reduce the vowel in \"the\" when no vowel follows (this requires a lexicon entry for \"the\" with feature \"reduced\", otherwise there will be no reduction)." (let ((fullform (cadr (car (caar (cdr (cdar (lex.lookup_all 'thee))))))) (reducedform (cadr(car(caar(cddr(lex.lookup 'the '(reduced))))))) seg) (mapcar (lambda (word) (if (string-equal "the" (downcase (item.feat word "name"))) (begin (set! seg (item.relation (item.daughtern (item.relation.daughtern word 'SylStructure)) 'Segment)) (if (string-equal "+" (item.feat (item.next seg) 'ph_vc)) (item.set_feat seg 'reducable 0) (item.set_feat seg 'reducable 1))))) (utt.relation.items utt 'Word))) utt) ;; For Multisyn voices only. Volker 14/08/06 (define (postlex_a utt) "(postlex_a utt) If POS of \"a\" is \"nn\" and segment feature \"reducable\", set it to 0. This is a bugfix, but still requires the target cost function to add a penalty if a candidate is reducable but the target is not. expro_target_cost does that." (let(seg) (mapcar (lambda(word) ;; (format t "%s\t%s\n" (item.feat word 'name)(item.feat word 'pos)) (if(and(string-equal "a" (downcase (item.feat word "name"))) (string-equal "nn" (item.feat word "pos"))) (begin (set! seg (item.relation (item.daughtern (item.relation.daughtern word 'SylStructure)) 'Segment)) ;; (format t "should not be reducable\n") (if (eq 1 (parse-number (item.feat seg 'reducable))) (item.set_feat seg 'reducable 0)))) ) (utt.relation.items utt 'Word))) utt) (define (postlex_unilex_vowel_reduction utt) "(postlex_unilex_vowel_reduction utt) Perform vowel reduction based on unilex specification of what can be reduced." (let () (mapcar (lambda (seg) (if (and (eq? (parse-number (item.feat seg "reducable")) 1) (not (> (parse-number (item.feat seg "R:SylStructure.parent.stress")) 0))) (if (not (and (seg_word_final seg) (string-equal (item.feat (item.next seg) 'ph_vc) "+"))) (item.set_feat seg "name" (item.feat seg "reducedform"))))) (utt.relation.items utt 'Segment))) utt) (define (seg_word_final seg) "(seg_word_final seg) Is this segment word final?" (let ((this_seg_word (item.parent (item.relation.parent seg 'SylStructure))) (silence (car (cadr (car (PhoneSet.description '(silences)))))) next_seg_word) (if (item.next seg) (set! next_seg_word (item.parent (item.relation.parent (item.next seg) 'SylStructure)))) (if (or (equal? this_seg_word next_seg_word) (string-equal (item.feat seg "name") silence)) nil t))) ;; imported from postlex_intervoc_r.scm Volker 14/08/06 (define (postlex_intervoc_r utt) "(postlex_intervoc_r UTT) Remove any word-final /r/ which is phrase-final or not going to be inter-vocalic i.e. the following words does not start with a vowel. NOTE: in older versions of unilex-rpx.out for Festival, there is no word-final /r/. " (let (word next_word last_phone following_phone) (set! word (utt.relation.first utt 'Word)) (while word (set! next_word (item.next word)) (set! last_phone (item.daughtern (item.daughtern(item.relation word 'SylStructure)))) (if next_word (begin (set! following_phone (item.daughter1 (item.daughter1 (item.relation next_word 'SylStructure)))) ; last_phone and following_phone should always be defined at this point, ; but since the upgrade to Fedora and characters no longer being in ISO ; but in UTF8, the pound sterling is no longer treated correctly. ; Probably (Token utt) should be fixed. (if (and following_phone last_phone) (begin (format t "%s\t%s %s %s %s\n" (item.name word) (item.name last_phone) (item.name following_phone) (item.feat following_phone 'ph_vc) (item.feat word 'pbreak)) (if(and(equal? "r" (item.name last_phone)) (or(not(equal? "NB" (item.feat word 'pbreak))) (equal? "-" (item.feat following_phone 'ph_vc)))) (begin (format t "\t\t\t/r/ in \"%s %s\" deleted\n" (item.name word)(item.name next_word)) (item.delete last_phone)))))) (if(and last_phone (equal? "r" (item.name last_phone))) (begin (format t "\t\t\tutterance-final /r/ deleted\n") (item.delete last_phone))) ) (set! word (item.next word)))) utt) (define (postlex_stop_deletion utt) "(postlex_stop_deletion utt) Delete any stop or affricative (phone which has a closure) immediately followed by another stop or affricative. Also save the identity of the deleted phone for the context cost functions. Consider: backtrack /b a k t r a k/ -> /b a t r a k/ (actually Jenny reduces : /b a k_cl k t_cl t r a k/ -> /b a k_cl t r a k/) If we then look for a diphone /a t/ we want to favour candidates coming from the same context i.e. which are actually a reduced /a k t/. In the data base, the 1st /a/ gets the feature right_context=k and the /t/ gets the fearture left_context=k. " (let(seg next_seg prev_seg) (set! seg (utt.relation.first utt 'Segment)) (while seg (set! prev_seg (item.prev seg)) (if prev_seg (begin ;(format t "%s %s %s\n" (item.name seg) ; (item.feat seg 'ph_ctype) ; (item.feat seg 'p.ph_ctype)) (if(and(or(equal? "s" (item.feat seg 'ph_ctype)) (equal? "a" (item.feat seg 'ph_ctype))) (or(equal? "s" (item.feat seg 'p.ph_ctype)) (equal? "a" (item.feat seg 'p.ph_ctype))) ; When there are 3 stops in a row, and after the 1st has been ; deleted, this prevents the 2nd to be deleted as well: (equal? 0 (item.feat prev_seg 'left_context))) (begin (set! prev_prev_seg (item.prev prev_seg)) (format t "postlex_stop_deletion: %s in %s\n" (item.name prev_seg) (item.name(item.parent(item.relation.parent prev_seg 'SylStructure)))) (if prev_prev_seg (begin ;(format t "setting left_context of %s and right context of %s to %s\n" ; (item.name seg) ; (item.name prev_prev_seg) ; (item.name prev_seg)) (item.set_feat seg 'left_context (item.name prev_seg)) (item.set_feat prev_prev_seg 'right_context (item.name prev_seg)))) (if(and(item.next seg) (equal? (item.name seg) (item.name prev_seg))) (begin ;(format t "setting left_context of %s to %s\n" ; (item.name (item.next seg) ; (item.name prev_seg)) (item.set_feat (item.next seg) 'left_context (item.name prev_seg)))) (item.delete prev_seg))))) (set! seg (item.next seg)))) utt) (provide 'postlex) festival/lib/tokenpos.scm0000644000175000017500000002373310643443252014126 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Functions used in identifying token types. ;;; (defvar token_most_common '( sym numeric month to day in the of on and writes a years from for jst at million by is was gmt page he that than more since as when with but after about or his i has it date no died number bst who miles university some people an only w year have ago were are pages up days months hours minutes through out had which least hi last now ft this all one its there between cents until over will before past they nearly times tim message so lbs just if age we during she billion then other be time new her first states not you members under would many says degrees two next fax week while bush been around including back campaign american within publisher flight points even early later world countries every edt can president most could their what them former began women killed another also received long americans pounds do dear said km made into did dead war tel still old x took total men like f am less c well late down weeks end chapter among place house away him election death almost students state soviet where version summer man s nation because washington top though m id est these spent seats gnu estimated those lost ian high each copies children acres tons son per my found won off seconds power nations federal born presidential much city begin p name different whose three home hello) "token_most_common A list of (English) words which were found to be most common in an text database and are used as discriminators in token analysis.") (define (token_pos_guess sc) "(tok_pos sc) Returns a general pos for sc's name. numeric All digits number float or comma'd numeric sym Contains at least one non alphanumeric month has month name (or abbrev) day has day name (or abbrev) rootname else downcased alphabetic. Note this can be used to find token_pos but isn't used directly as its not disciminatory enough." (let ((name (downcase (item.name sc)))) (cond ((string-matches name "[0-9]+") 'numeric) ((or (string-matches name "[0-9]+\\.[0-9]+") (string-matches name "[0-9][0-9]?[0-9]?,\\([0-9][0-9][0-9],\\)*[0-9][0-9][0-9]")) 'number) ((string-matches name ".*[^A-Za-z0-9].*") 'sym) ((member_string name '(jan january feb february mar march apr april may jun june jul july aug august sep sept september oct october nov november dec december)) 'month) ((member_string name '(sun sunday mon monday tue tues tuesday wed wednesday thu thurs thursday fri friday sat saturday)) 'day) ((member_string name token_most_common) name) (t '_other_)))) (define (token_no_starting_quote token) "(token_no_starting_quote TOKEN) Check to see if a single quote (or backquote) appears as prepunctuation in this token or any previous one in this utterance. This is used to disambiguate ending single quote as possessive or end quote." (cond ((null token) t) ((string-matches (item.feat token "prepunctuation") "[`']") nil) (t (token_no_starting_quote (item.relation.prev token "Token"))))) (define (token_zerostart sc) "(zerostart sc) Returns, 1 if first char of sc's name is 0, 0 otherwise." (if (string-matches (item.name sc) "^0.*") "1" "0")) (define (tok_roman_to_numstring roman) "(tok_roman_to_numstring ROMAN) Takes a string of roman numerals and converts it to a number and then returns the printed string of that. Only deals with numbers up to 50." (let ((val 0) (chars (symbolexplode roman))) (while chars (cond ((equal? (car chars) 'X) (set! val (+ 10 val))) ((equal? (car chars) 'V) (set! val (+ 5 val))) ((equal? (car chars) 'I) (cond ((equal? (car (cdr chars)) 'V) (set! val (+ 4 val)) (set! chars (cdr chars))) ((equal? (car (cdr chars)) 'X) (set! val (+ 9 val)) (set! chars (cdr chars))) (t (set! val (+ 1 val)))))) (set! chars (cdr chars))) (format nil "%d" val))) (define (num_digits sc) "(num_digits SC) Returns number of digits (actually chars) is SC's name." (string-length (format nil "%s" (item.name sc)))) (define (month_range sc) "(month_range SC) 1 if SC's name is > 0 and < 32, 0 otherwise." (let ((val (parse-number (item.name sc)))) (if (and (> val 0) (< val 32)) "1" "0"))) (define (remove_leading_zeros name) "(remove_leading_zeros name) Remove leading zeros from given string." (let ((nname name)) (while (string-matches nname "^0..*") (set! nname (string-after nname "0"))) nname)) (define (token_money_expand type) "(token_money_expand type) Convert shortened form of money identifier to words if of a known type." (cond ((string-equal type "HK") (list "Hong" "Kong")) ((string-equal type "C") (list "Canadian")) ((string-equal type "A") (list "Australian")) ((< (length type) 4) (mapcar (lambda (letter) (list (list 'name letter) (list 'pos token.letter_pos))) (symbolexplode type))) (t (list type)))) (define (find_month_from_number token string-number) "(find_month_from_number token string-number) Find the textual representation of the month from the given string number" (let ((nnum (parse-number string-number))) (cond ((equal? 1 nnum) (list "January")) ((equal? 2 nnum) (list "February")) ((equal? 3 nnum) (list "March")) ((equal? 4 nnum) (list "April")) ((equal? 5 nnum) (list "May")) ((equal? 6 nnum) (list "June")) ((equal? 7 nnum) (list "July")) ((equal? 8 nnum) (list "August")) ((equal? 9 nnum) (list "September")) ((equal? 10 nnum) (list "October")) ((equal? 11 nnum) (list "November")) ((equal? 12 nnum) (list "December")) (t (cons "month" (builtin_english_token_to_words token string-number)))))) (define (tok_allcaps sc) "(tok_allcaps sc) Returns 1 if sc's name is all capitals, 0 otherwise" (if (string-matches (item.name sc) "[A-Z]+") "1" "0")) (define (tok_section_name sc) "(tok_section_name sc) Returns 1 if sc's name is in list of things that are section/chapter like." (if (member_string (downcase (item.name sc)) '(chapter section part article phrase verse scene act book volume chap sect art vol war fortran saturn trek)) "1" "0")) (define (tok_string_as_letters name) "(tok_string_as_letters NAME) Return list of letters marked as letter part of speech made by exploding NAME." (mapcar (lambda (letter) (list (list 'name letter) (list 'pos token.letter_pos))) (symbolexplode name))) (define (tok_rex sc) "(tok_rex sc) Returns 1 if King like title is within 3 tokens before or 2 after." (let ((kings '(king queen pope duke tsar emperor shah ceasar duchess tsarina empress baron baroness count countess))) (if (or (member_string (downcase (item.feat sc "R:Token.pp.name")) kings) (member_string (downcase (item.feat sc "R:Token.pp.p.name")) kings) (member_string (downcase (item.feat sc "R:Token.n.name")) kings)) "1" "0"))) (define (tok_rex_names sc) "(tok_rex sc) Returns 1 if this is a King-like name." (if (and (member_string (downcase (item.name sc)) '(louis henry charles philip george edward pius william richard ptolemy john paul peter nicholas alexander frederick james alfonso ivan napolean leo gregory catherine alexandria pierre elizabeth mary)) (or (string-equal "" (item.feat sc "punc")) (string-equal "0" (item.feat sc "punc")))) "1" "0")) (provide 'tokenpos) festival/lib/festtest.scm0000644000175000017500000000707207262605650014130 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Some basic functions used in tests for Festival ;;; (define (test_words text) "(test_words TEXT) prints TEXT, Synthesizes TEXT and outputs the words in it." (format t "Word test: %s\n " text) (set! utt1 (utt.synth (eval (list 'Utterance 'Text text)))) (mapcar (lambda (word) (format t "%s " (car word))) (utt.features utt1 'Word '(name))) (format t "\n") t) (define (test_segments text) "(test_segments TEXT) prints TEXT, Synthesizes TEXT and outputs the segments in it." (format t "Segment test: %s\n " text) (set! utt1 (utt.synth (eval (list 'Utterance 'Text text)))) (mapcar (lambda (word) (format t "%s " (car word))) (utt.features utt1 'Segment '(name))) (format t "\n") ) (define (test_phrases text) "(test_phrases TEXT) prints TEXT, Synthesizes TEXT and outputs the words and phrase breaks." (format t "Phrase test: %s \n " text) (set! utt1 (utt.synth (eval (list 'Utterance 'Text text)))) (mapcar (lambda (phrase) (mapcar (lambda (w) (format t "%s " (car (car w)))) (cdr phrase)) (format t "%s\n " (car (car phrase)))) (utt.relation_tree utt1 'Phrase)) (format t "\n") t) (provide 'festtest) festival/lib/cslush.scm0000644000175000017500000000760107262605650013566 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Functions specific to using Festival in cslush part of the OGI toolkit ;;; The OGI toolkit is a complete dialog building system with speech ;;; recognition and synthesis (Festival) it is available for free for ;;; research purposes from ;;; http://www.cse.ogi.edu/CSLU/toolkit/toolkit.html ;;; ;;; Note this cslush interface requires you to compile festival ;;; with tcl (7.6) ;;; ;;; The functions replace the C++ level functions Jacques H. de Villiers ;;; from CSLU wrote for the previous version ;;; (if (not (member 'tcl *modules*)) (error "cslush: can't load cslush, TCL not supported in this installation of Festival.")) (define (cslush.getwave utt) "(cslush.getwave UTT) Extract wave memory info, pass this to wave import in CSLUsh." (format nil "%s %s %s" (utt.wave.info utt 'data_addr) (utt.wave.info utt 'num_samples) (utt.wave.info utt 'sample_rate))) (define (cslush.getphone utt) "(cslush.getphone UTT) Return segment names a single string of phones, for use to pass to TCL." (let ((phones "")) (mapcar (lambda (s) (if (string-equal phones "") (set! phones (format nil "%s" (utt.streamitem.feat utt s 'name))) (set! phones (format nil "%s %s" phones (utt.streamitem.feat utt s 'name))))) (utt.stream utt 'Segment)) phones)) (define (cslush TCLCOMMAND) "(cslush TCLCOMMAND) Pass TCLCOMMAND to TCL interpreter, returns what TCL returns as a string." (tcl_eval TCLCOMMAND)) (provide 'cslush) festival/lib/cart_aux.scm0000644000175000017500000001733111603103562014061 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996-2011 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Some functions for manipulating decision trees ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (cart_prune_tree_thresh tree threshold default) "(prune_cart_tree_thresh TREE THRESHOLD DEFAULT) Prune the classification tree TREE so that all tail nodes with a prediction probabality less than THRESHOLD and changed to return DEFAULT instead. This may be used when different mistakes have actually different penalites hence some control of the defaults need to be controlled." (cond ((cdr tree) ;; a question (list (car tree) (cart_prune_tree_thresh (car (cdr tree)) threshold default) (cart_prune_tree_thresh (car (cdr (cdr tree))) threshold default))) ((< (cart_class_probability (car tree)) threshold) (list (list (list threshold default) default))) (t ;; leave asis tree))) (define (cart_class_probability class) "(cart_class_probability CLASS) Returns the probability of the best class in the cart leaf node CLASS. If CLASS simple has a value and now probabilities the probabilities it assume to be 1.0." (let ((val 0.0)) (set! val (assoc (car (last class)) class)) (if val (car (cdr val)) 1.0))) (define (cart_class_prune_merge tree) "(cart_class_prune_merge tree) Prune all sub trees which are pure. That is they all predict the same class. This can happen when some other pruning technique as modified a sub-tree now making it pure." (let ((pure (cart_tree_pure tree))) (cond (pure pure) ((cdr tree);; a question (list (car tree) (cart_class_prune_merge (car (cdr tree))) (cart_class_prune_merge (car (cdr (cdr tree)))))) (t;; a leaf leave asis tree)))) (define (cart_tree_pure tree) "(cart_tree_pure tree) Returns a probability density function if all nodes in this tree predict the same class and nil otherwise" (cond ((cdr tree) (let ((left (cart_tree_pure (car (cdr tree)))) (right (cart_tree_pure (car (cdr (cdr tree)))))) (cond ((not left) nil) ((not right) nil) ((equal? (car (last left)) (car (last right))) left) (t nil)))) (t ;; its a leaf, so of couse its pure tree))) (define (cart_simplify_tree tree map) "(cart_simplify_tree TREE) Simplify a CART tree by reducing probability density functions to simple single clasifications (no probabilities). This removes valuable information from the tree but makes them smaller easier to read by humans and faster to read by machines. Also the classes may be mapped by the assoc list in map. The bright ones amongst you will note this could be better and merge 'is' operators into 'in' operators in some situations especially if you are ignoring actual probability distributions." (cond ((cdr tree) (list (car tree) (cart_simplify_tree (car (cdr tree)) map) (cart_simplify_tree (car (cdr (cdr tree))) map))) (t (let ((class (car (last (car tree))))) (if (assoc class map) (list (cdr (assoc class map))) (list (last (car tree)))))))) (define (cart_simplify_tree2 tree) "(cart_simplify_tree2 TREE) Simplify a CART tree by reducing probability density functions to only non-zero probabilities." (cond ((cdr tree) (list (car tree) (cart_simplify_tree2 (car (cdr tree))) (cart_simplify_tree2 (car (cdr (cdr tree)))))) (t (list (cart_remove_zero_probs (car tree)))))) (define (cart_remove_zero_probs pdf) "(cart_remove_zero_probs pdf) Removes zero probability classes in pdf, last in list is best in class (as from cart leaf node)." (cond ((null (cdr pdf)) pdf) ((equal? 0 (car (cdr (car pdf)))) (cart_remove_zero_probs (cdr pdf))) (t (cons (car pdf) (cart_remove_zero_probs (cdr pdf)))))) (define (cart_interpret_debug i tree) "(cart_interpret_debug i tree) In comparing output between different implementations (flite vs festival) This prints out the details as it interprets the tree." (cond ((cdr tree) ;; question (format t "%s %s %s\n" (car (car tree)) (upcase (cadr (car tree))) (car (cddr (car tree)))) (set! a (item.feat i (car (car tree)))) (format t "%s\n" a) (cond ((string-equal "is" (cadr (car tree))) (if (string-equal a (car (cddr (car tree)))) (begin (format t " YES\n") (cart_interpret_debug i (car (cdr tree)))) (begin (format t " NO\n") (cart_interpret_debug i (car (cddr tree)))))) ((string-equal "<" (cadr (car tree))) (if (< (parse-number a) (parse-number (car (cddr (car tree))))) (begin (format t " YES\n") (cart_interpret_debug i (car (cdr tree)))) (begin (format t " NO\n") (cart_interpret_debug i (car (cddr tree)))))) (t (format t "unknown q type %l\n" (car tree))))) (t ;; leaf (car tree) ))) ;;; ;;; Prediction tree for OLS trees ;;; applies OLS coefficients from appropriate leaf of tree ;;; (define (ols_tree_predict i tree) ;; Surprisingly simple function does the necessary work (let ((p (wagon i tree))) (apply + (cons (cadr (car (car p))) ;; Intercept (mapcar (lambda (fp) ;; get feature value and multiple by coefficent (* (parse-number (item.feat i (car fp))) (cadr fp))) (cdr (car p))))))) (provide 'cart_aux) festival/lib/pauses.scm0000644000175000017500000002377511332436761013575 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Predicting pause insertion (define (Pauses utt) "(Pauses utt) Insert pauses where required." (let ((rval (apply_method 'Pause_Method utt))) (cond (rval rval) ;; new style (t (Classic_Pauses utt)))) (Pause_optional_deleting_B_X utt)) (define (Classic_Pauses utt) "(Pauses UTT) Predict pause insertion." (let ((words (utt.relation.items utt 'Word)) lastword tpname) (if words (begin (insert_initial_pause utt) ;; always have a start pause (set! lastword (car (last words))) (mapcar (lambda (w) (let ((pbreak (item.feat w "pbreak")) (emph (item.feat w "R:Token.parent.EMPH"))) (cond ((or (string-equal "B" pbreak) (string-equal "BB" pbreak)) (insert_pause utt w)) ; ((string-equal emph "1") ; (insert_pause utt w)) ((equal? w lastword) (insert_pause utt w))))) words) ;; The embarrassing bit. Remove any words labelled as punc or fpunc (mapcar (lambda (w) (let ((pos (item.feat w "pos"))) (if (or (string-equal "punc" pos) (string-equal "fpunc" pos)) (let ((pbreak (item.feat w "pbreak")) (wp (item.relation w 'Phrase))) (if (and (string-matches pbreak "BB?") (item.relation.prev w 'Word)) (item.set_feat (item.relation.prev w 'Word) "pbreak" pbreak)) (item.relation.remove w 'Word) ;; can't refer to w as we've just deleted it (item.relation.remove wp 'Phrase))))) words) ;; 12/01/2006 V.Strom: Even more embarrasing: Delete all silences ;; that are followed by a silence. These silence sequences ;; emerge if 'punc of phrase-final words consists of more than one ;; character, e.g. period+quote. That in turn causes problems in ;; build_utts: the 2nd silence ends up with no features but its name, ;; because there is no corresponding 2nd silence in the phone ;; segmentation to align with. ;; This schould be fixed in the functions below, but it is easier for ;; me to clean up at the end: (set! sil (car (car (cdr (car (PhoneSet.description '(silences))))))) (set! seg (item.next(utt.relation.first utt 'Segment))) (while seg (if(and(equal? sil (item.name seg)) (equal? sil (item.name (item.prev seg)))) (item.delete (item.prev seg))) (set! seg (item.next seg))))) utt)) (define (insert_pause utt word) "(insert_pause UTT WORDITEM) Insert a silence segment after the last segment in WORDITEM in UTT." (let ((lastseg (find_last_seg word)) (silence (car (car (cdr (car (PhoneSet.description '(silences)))))))) (if lastseg (item.relation.insert lastseg 'Segment (list silence) 'after)))) (define (insert_initial_pause utt) "(insert_initial_pause UTT) Always have an initial silence if the utterance is non-empty. Insert a silence segment after the last segment in WORDITEM in UTT." (let ((firstseg (car (utt.relation.items utt 'Segment))) (silence (car (car (cdr (car (PhoneSet.description '(silences)))))))) (if firstseg (item.relation.insert firstseg 'Segment (list silence) 'before)))) (define (insert_final_pause utt) "(insert_final_pause UTT) Always have a final silence if the utterance is non-empty." (let ((lastseg (utt.relation.last utt 'Segment)) (silence (car (car (cdr (car (PhoneSet.description '(silences)))))))) (set! silence (format nil "%l" silence)) ; to make the symbol a string ;(format t "silence is %l\n" silence) ;(format t "lastseg is %l\n" (item.name lastseg)) (if lastseg (if (not(equal? (item.name lastseg) silence)) (begin (format t "iserted final pause %s\n" silence) (item.relation.insert lastseg 'Segment (list silence) 'after)))))) (define (find_last_seg word) ;;; Find the segment that is immediately at this end of this word ;;; If this word is punctuation it might not have any segments ;;; so we have to check back until we find a word with a segment in it (cond ((null word) nil) ;; there are no segs (don't think this can happen) (t (let ((lsyl (item.relation.daughtern word 'SylStructure))) (if lsyl (item.relation.daughtern lsyl 'SylStructure) (find_last_seg (item.relation.prev word 'Word))))))) (define (Unisyn_Pauses utt) "(Unisyn_Pauses UTT) Predict pause insertion in a Unisyn utterance structure." (let ((words (utt.relation.items utt 'Word)) lastword tpname) (if words (begin (us_insert_initial_pause utt) ;; always have a start pause (set! lastword (car (last words))) (mapcar (lambda (w) (let ((pbreak (item.feat w "pbreak")) (emph (item.feat w "R:Token.parent.EMPH"))) (cond ((or (string-equal "B" pbreak) (string-equal "BB" pbreak)) (us_insert_pause utt w)) ; ((string-equal emph "1") ; (us_insert_pause utt w)) ((equal? w lastword) (us_insert_pause utt w))))) words) ;; The embarrassing bit. Remove any words labelled as punc or fpunc (mapcar (lambda (w) (let ((pos (item.feat w "pos"))) (if (or (string-equal "punc" pos) (string-equal "fpunc" pos)) (let ((pbreak (item.feat w "pbreak")) (wp (item.relation w 'Phrase))) (if (and (string-matches pbreak "BB?") (item.relation.prev w 'Word)) (item.set_feat (item.relation.prev w 'Word) "pbreak" pbreak)) (item.relation.remove w 'Word) ;; can't refer to w as we've just deleted it (item.relation.remove wp 'Phrase))))) words))) utt)) (define (us_insert_pause utt word) "(us_insert_pause UTT WORDITEM) Insert a silence segment after the last segment in WORDITEM in UTT." (let ((lastseg (us_find_last_seg word)) (silence "pau")) (if lastseg (item.relation.insert lastseg 'Segment (list silence) 'after)))) (define (us_insert_initial_pause utt) "(us_insert_initial_pause UTT) Always have an initial silence if the utterance is non-empty. Insert a silence segment after the last segment in WORDITEM in UTT." (let ((firstseg (utt.relation.first utt 'Segment)) (silence "pau")) (if firstseg (item.relation.insert firstseg 'Segment (list silence) 'before)))) (define (us_find_last_seg word) ;;; Find the segment that is immediately at this end of this word ;;; If this word is punctuation it might not have any segments ;;; so we have to check back until we find a word with a segment in it (cond ((null word) nil) ;; there are no segs (don't think this can happen) (t (if (item.daughtern_to (item.relation word 'WordStructure) 'Syllable) (item.daughtern_to (item.relation (item.daughtern_to (item.relation word 'WordStructure) 'Syllable) 'SylStructure) 'Segment) (us_find_last_seg (item.relation.prev word 'Word)))))) (define (Pause_optional_deleting_B_X utt) "(Pause_optional_deleting_B_X utt) Delete all phone symbols starting with 'B_' from the segemt relation (a B_150 e.g. is a 150ms pause) if symbol 'Pause_delete_B_X is defined. " ; The B_X never occur in the phone segmentation but are predicted by ; some pause methods, in particular the default I used to produce the ; .utt files for the 2009 test sentences for the Blizzard challange. ; Some participants complained about them and I had to fix it quickly. (if (symbol-bound? 'Pause_delete_B_X) (let(seg ) (set! seg (item.next(utt.relation.first utt 'Segment))) (while seg (set! next_seg (item.next seg)) ;(format t "segment %l\n" (item.name seg)) (if(string-matches (item.name seg) "B_[0-9]*") (item.delete seg)) (set! seg next_seg))))) (provide 'pauses) festival/lib/scfg.scm0000644000175000017500000000626710045556464013217 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Some functions for manipulating a SCFG parse tree (require_module 'parser) (define (scfg_simplify tree) "(scfg_brackets_only tree) Output only the bracketing and the bottom level pos and words." (cond ((not tree) nil) ((car (cdr (assoc 'pos (car (cdr (car tree)))))) ;; terminal node (list (car (cdr (assoc 'pos (car (cdr (car tree)))))) (car (car tree)))) (t (cons (car (car tree)) (mapcar scfg_simplify (cdr tree)))))) (define (scfg_simplify_relation_tree trees) (mapcar scfg_simplify trees)) (defvar scfg_eos_tree eou_tree "scfg_eos_tree In MultiProbParse this CART tree is used to define end of sentence within an utterance. It is applied to the token relation. By default it is set to eou_tree.") (provide 'scfg) festival/lib/mettree.scm0000644000175000017500000000734307262605650013735 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1998 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Some (experimental) data for investigating metrical trees ;;; ;;; Set up generation of metrical tree, this includes getting ;;; a syntactic parse ;;; ;;; Use as ;;; (set! utt1 (metsynth (Utterance Text "For afternoon tea"))) ;;; (utt.relation_tree utt1 'MetricalTree) (require 'scfg) (set! scfg_grammar (load (path-append libdir "scfg_wsj_wp20.gram") t)) (define (mettext utt) (Initialize utt) (Text utt) (Token_POS utt) (Token utt) (POS utt) (print "here1") (Phrasify utt) (print "here2") (ProbParse utt) (print "here3") (auto_metrical_tree utt) ) (define (metsynth utt) (mettext utt) (Wave_Synth utt) ) ;;; Assumed everything is using Roger diphones ;;(lex.create "cmu_mettree") ;;;(lex.set.phoneset "radio_phones") ;;(lex.set.phoneset "radio_phones") (define (setup_cmu_mettree_lex) "(setup_cmu_mettreelex) Lexicon derived from the CMU lexicon (cmudict-0.1), around 100,000 entries, in the radio phoneset (sort of darpa-like)." (if (not (member_string "cmu_mettree" (lex.list))) (begin (print "making cmu lexicon") (lex.create "cmu_mettree") (lex.set.compile.file (path-append lexdir "cmu_mettree_lex.out")) (lex.set.phoneset "radio") (require 'lts__us) ;; US English letter to sound rules (lex.set.lts.method 'lts_rules) (lex.set.lts.ruleset 'nrl_us)))) (provide 'mettree) festival/lib/java.scm0000644000175000017500000000500307262605650013200 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1998 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Functions specific to supporting a Java client ;;; ;; none required yet (provide 'java) festival/lib/clunits.scm0000644000175000017500000002442510077563027013750 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Carnegie Mellon University and ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1998-2001 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH, CARNEGIE MELLON UNIVERSITY AND THE ;; ;;; CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH REGARD TO ;; ;;; THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY ;; ;;; AND FITNESS, IN NO EVENT SHALL THE UNIVERSITY OF EDINBURGH, CARNEGIE ;; ;;; MELLON UNIVERSITY NOR THE CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, ;; ;;; INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER ;; ;;; RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION ;; ;;; OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF ;; ;;; OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Cluster Unit selection support (Black and Taylor Eurospeech '97) ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Run-time support, selection and synthesis and some debugging functions ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (require_module 'clunits) (defvar cluster_synth_pre_hooks nil) (defvar cluster_synth_post_hooks nil) (defvar clunits_time time) ;; some old voices might use this (defSynthType Cluster (apply_hooks cluster_synth_pre_hooks utt) (Clunits_Select utt) (Clunits_Get_Units utt) (Clunits_Join_Units utt) (apply_hooks cluster_synth_post_hooks utt) utt ) (define (Clunits_Join_Units utt) "(Clunits_Join_Units utt) Join the preselected and gotten units into a waveform." (let ((join_method (get_param 'join_method clunits_params 'simple))) ;; Choice of function to put them together (cond ((string-equal join_method 'windowed) (Clunits_Windowed_Wave utt) (clunits::fix_segs_durs utt)) ((string-equal join_method 'smoothedjoin) (Clunits_SmoothedJoin_Wave utt) (clunits::fix_segs_durs utt)) ((string-equal join_method 'none) t) ((string-equal join_method 'modified_lpc) (defvar UniSyn_module_hooks nil) (Param.def "unisyn.window_name" "hanning") (Param.def "unisyn.window_factor" 1.0) (Parameter.def 'us_sigpr 'lpc) (mapcar (lambda (u s) (item.set_feat s "source_end" (item.feat u "end"))) (utt.relation.items utt 'Unit) (utt.relation.items utt 'Segment)) (us_unit_concat utt) (if (not (member 'f0 (utt.relationnames utt))) (targets_to_f0 utt)) (if (utt.relation.last utt 'Segment) (set! pm_end (+ (item.feat (utt.relation.last utt 'Segment) "end") 0.02)) (set! pm_end 0.02)) (us_f0_to_pitchmarks utt 'f0 'TargetCoef pm_end) (us_mapping utt 'segment_single) (us_generate_wave utt (Parameter.get 'us_sigpr) 'analysis_period)) ((string-equal join_method 'smoothed_lpc) ; (format t "smoothed_lpc\n") (defvar UniSyn_module_hooks nil) (Param.def "unisyn.window_name" "hanning") (Param.def "unisyn.window_factor" 1.0) (Parameter.def 'us_sigpr 'lpc) (mapcar (lambda (u s) (item.set_feat s "source_end" (item.feat u "end")) (item.set_feat s "unit_duration" (- (item.feat u "seg_end") (item.feat u "seg_start"))) ) (utt.relation.items utt 'Unit) (utt.relation.items utt 'Segment)) (us_unit_concat utt) (mapcar (lambda (u s) (item.set_feat s "num_frames" (item.feat u "num_frames"))) (utt.relation.items utt 'Unit) (utt.relation.items utt 'Segment)) (if (not (member 'f0 (utt.relationnames utt))) (targets_to_f0 utt)) (if (utt.relation.last utt 'Segment) (set! pm_end (+ (item.feat (utt.relation.last utt 'Segment) "end") 0.02)) (set! pm_end 0.02)) (us_f0_to_pitchmarks utt 'f0 'TargetCoef pm_end) (cl_mapping utt clunits_params) (us_generate_wave utt (Parameter.get 'us_sigpr) 'analysis_period)) (t (Clunits_Simple_Wave utt))) utt ) ) (define (clunits::units_selected utt filename) "(clunits::units_selected utt filename) Output selected unitsfile indexes for each unit in the given utterance. Results saved in given file name, or stdout if filename is \"-\"." (let ((fd (if (string-equal filename "-") t (fopen filename "w"))) (end 0) (sample_rate (cadr (assoc 'sample_rate (wave.info (utt.wave utt)))))) (format fd "#\n") (mapcar (lambda (s) (let ((dur (/ (- (item.feat s "samp_end") (item.feat s "samp_start")) sample_rate)) (start (/ (item.feat s "samp_start") sample_rate))) (set! end (+ end dur)) (format fd "%f 125 %s ; %s %10s %f %f %f\n" end (string-before (item.name s) "_") (item.name s) (item.feat s "fileid") (item.feat s "unit_start") (item.feat s "unit_middle") (item.feat s "unit_end")) )) (utt.relation.items utt 'Unit)) (if (not (string-equal filename "-")) (fclose fd)) t)) (define (clunits::units_segs utt filename) "(clunits::units_segs utt filename) Svaes the unit selections (alone) for display." (let ((fd (if (string-equal filename "-") t (fopen filename "w"))) (end 0) (sample_rate (cadr (assoc 'sample_rate (wave.info (utt.wave utt)))))) (format fd "#\n") (mapcar (lambda (s) (let ((dur (/ (- (item.feat s "samp_end") (item.feat s "samp_start")) sample_rate)) (start (/ (item.feat s "samp_start") sample_rate))) (set! end (+ end dur)) (format fd "%f 125 %s \n" end (string-before (item.name s) "_") ; (item.name s) ) )) (utt.relation.items utt 'Unit)) (if (not (string-equal filename "-")) (fclose fd)) t)) (define (clunits::fix_segs_durs utt) "(clunits::fix_segs_durs utt) Takes the actual unit times and places then back on the segs." (let ((end 0) (sample_rate (cadr (assoc 'sample_rate (wave.info (utt.wave utt)))))) (mapcar (lambda (u s) (let ((dur (/ (- (item.feat u "samp_end") (item.feat u "samp_start")) sample_rate)) (seg_start (/ (- (item.feat u "samp_seg_start") (item.feat u "samp_start")) sample_rate))) (if (item.prev s) (item.set_feat (item.prev s) "end" (+ (item.feat s "p.end") seg_start))) (set! end (+ end dur)) (item.set_feat s "end" end))) (utt.relation.items utt 'Unit) (utt.relation.items utt 'Segment) ) utt)) (define (clunits::display utt) "(clunits::display utt) Display utterance with emulabel. Note this saves files in scratch/wav/ and scratch/lab/." (let ((id "cl01")) (utt.save.wave utt (format nil "scratch/wav/%s.wav" id)) (utt.save.segs utt (format nil "scratch/lab/%s.lab" id)) (system "cd scratch; emulabel ../etc/emu_lab cl01 &") t)) ; (define (clunits::debug_resynth_units utt) ; "(clunits::debug_resynth_units utt) ; Check each of the units in utt against the related label ; files and re-synth with any given new boundaries. Note this is ; will only work if the segment still overlaps with its original and ; also note that with a rebuild of the clunits db a complete different ; set of units may be selected for this utterance." ; (let () ; (mapcar ; (lambda (unit) ; (clunits::check_unit_boundaries unit)) ; (utt.relation.items utt 'Unit)) ; ;; This can't be done like this ... ; (Clunits_Get_Units utt) ;; get unit signal/track stuff ; (Clunits_Join_Units utt) ;; make a complete waveform ; (apply_hooks cluster_synth_post_hooks utt) ; utt) ; ) (define (clunits::join_parameters utt) "(clunits::join_parameters utt) Join selected paremeters (rather than the signal), used in F0 and Articulatory selection." (let ((params nil) (num_channels 0) (num_frames 0 )) (mapcar (lambda (unit) (set! num_frames (+ num_frames (track.num_frames (item.feat unit "coefs")))) (set! num_channels (track.num_channels (item.feat unit "coefs"))) (format t "coounting %d %d\n" num_frames num_channels) ) (utt.relation.items utt 'Unit)) (set! params (track.resize nil 0 num_channels)) (mapcar (lambda (unit) (set! frames 0) (format t "inserting \n") (format t "%l %l %l %l %l\n" params (track.num_frames params) (item.feat unit "coefs") 0 (track.num_frames (item.feat unit "coefs"))) (track.insert params (track.num_frames params) (item.feat unit "coefs") 0 (track.num_frames (item.feat unit "coefs"))) ) (utt.relation.items utt 'Unit)) (utt.relation.create utt "AllCoefs") (set! coefs_item (utt.relation.append utt "AllCoefs")) (item.set_feat coefs_item "name" "AllCoefs") (item.set_feat coefs_item "AllCoefs" params) utt )) (provide 'clunits) festival/lib/clunits_build.scm0000644000175000017500000004012712243416252015116 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Carnegie Mellon University and ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1998-2005 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH, CARNEGIE MELLON UNIVERSITY AND THE ;; ;;; CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH REGARD TO ;; ;;; THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY ;; ;;; AND FITNESS, IN NO EVENT SHALL THE UNIVERSITY OF EDINBURGH, CARNEGIE ;; ;;; MELLON UNIVERSITY NOR THE CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, ;; ;;; INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER ;; ;;; RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION ;; ;;; OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF ;; ;;; OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Cluster Unit selection support (Black and Taylor Eurospeech '97) ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; clunits build support ;;; ;;; There are five stages to this ;;; Load in all utterances ;;; Load in their coefficients ;;; Collect together the units of the same type ;;; build distance tables from them ;;; dump features for them ;;; (require_module 'clunits) ;; C++ modules support (require 'clunits) ;; run time scheme support ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (do_all) (let () (format t "Loading utterances and sorting types\n") (set! utterances (acost:db_utts_load clunits_params)) (set! unittypes (acost:find_same_types utterances clunits_params)) (acost:name_units unittypes) (format t "Dumping features for clustering\n") (acost:dump_features unittypes utterances clunits_params) (format t "Loading coefficients\n") (acost:utts_load_coeffs utterances) ;; If you are short of diskspace try this (acost:disttabs_and_clusters unittypes clunits_params) ;; or if you have lots of diskspace try ; (format t "Building distance tables\n") ; (acost:build_disttabs unittypes clunits_params) ; ;; Build the cluster trees (requires disttabs and features) ; (format t "Building cluster trees\n") ; (acost:find_clusters (mapcar car unittypes) clunits_params) ;; Tidy up and put things together (acost:collect_trees (mapcar car unittypes) clunits_params) (format t "Saving unit catalogue\n") (acost:save_catalogue utterances clunits_params) ) ) (define (do_init) (set! utterances (acost:db_utts_load clunits_params)) (set! unittypes (acost:find_same_types utterances clunits_params)) (acost:name_units unittypes) t) (define (acost:disttabs_and_clusters unittypes clunits_params) "(acost:disttabs_and_custers unittypes) Cause it uses so much diskspace, build each table individually and them the cluster, removing the table before moving on to the next." (mapcar (lambda (uu) (acost:build_disttabs (list uu) clunits_params) (acost:find_clusters (list (car uu)) clunits_params) (delete-file (format nil "%s/%s/%s%s" (get_param 'db_dir clunits_params "./") (get_param 'disttabs_dir clunits_params "disttabs/") (car uu) (get_param 'disttabs_ext clunits_params ".disttab"))) ) unittypes) t) (define (acost:db_utts_load params) "(acost:db_utts_load params) Load in all utterances identified in database." (let ((files (car (cdr (assoc 'files params))))) (set! acost:all_utts (mapcar (lambda (fname) (set! utt_seg (Utterance Text fname)) (utt.load utt_seg (string-append (get_param 'db_dir params "./") (get_param 'utts_dir params "festival/utts/") fname (get_param 'utts_ext params ".utt"))) utt_seg) files)))) (define (acost:utts_load_coeffs utterances) "(acost:utts_load_coeffs utterances) Loading the acoustic coefficients of for each utterance." (mapcar (lambda (utt) (acost:utt.load_coeffs utt clunits_params)) utterances) t) (define (acost:find_same_types utterances params) "(acost:find_same_types utterances) Find all the stream items of the same type and collect them into lists of that type." (let ((clunit_name_feat (get_param 'clunit_name_feat params "name")) (clunit_relation (get_param 'clunit_relation params "Segment"))) (set! acost:unittypes nil) (mapcar (lambda (u) (mapcar (lambda (s) (let ((cname (item.feat s clunit_name_feat))) (if (not (string-equal "ignore" cname)) (begin (item.set_feat s "clunit_name" (item.feat s clunit_name_feat)) (let ((p (assoc (item.feat s "clunit_name") acost:unittypes))) (if p (set-cdr! p (cons s (cdr p))) (set! acost:unittypes (cons (list (item.feat s "clunit_name") s) acost:unittypes)))))))) (utt.relation.items u clunit_relation))) utterances) (acost:prune_unittypes acost:unittypes params))) (define (acost:prune_unittypes unittypes params) "(acost:prune_unittypes unittypes) If unit types are complex (contain an _) then remove all unittypes sets with less than unittype_prune_threshold (typically 3)." (if (string-matches (car (car unittypes)) ".*_.*") (let ((ut nil) (pt (get_param 'unittype_prune_threshold params 0))) (while unittypes (if (or (eq? pt 0) (> (length (cdr (car unittypes))) pt)) (set! ut (cons (car unittypes) ut))) (set! unittypes (cdr unittypes))) (reverse ut)) unittypes)) (define (acost:name_units unittypes) "(acost:name_units unittypes) Names each unit with a unique id and number the occurrences of each type." (let ((idnum 0) (tynum 0)) (mapcar (lambda (s) (set! tynum 0) (mapcar (lambda (si) (item.set_feat si "unitid" idnum) (set! idnum (+ 1 idnum)) (item.set_feat si "occurid" tynum) (set! tynum (+ 1 tynum))) (cdr s)) (format t "units \"%s\" %d\n" (car s) tynum)) unittypes) (format t "total units %d\n" idnum) idnum)) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Generating feature files (define (acost:dump_features unittypes utterances params) "(acost:dump_features unittypes utterances params) Do multiple passes over the utterances for each unittype and dump the desired features. This would be easier if utterances weren't require for feature functions." (mapcar (lambda (utype) (acost:dump_features_utype (car utype) (cdr utype) utterances params)) unittypes) t) (define (acost:dump_features_utype utype uitems utterances params) "(acost:dump_features_utype utype utterances params) Dump features for all items of type utype." (let ((fd (fopen (string-append (get_param 'db_dir params "./") (get_param 'feats_dir params "festival/feats/") utype (get_param 'feats_ext params ".feats")) "w")) (feats (car (cdr (assoc 'feats params))))) (format t "Dumping features for %s\n" utype) (mapcar (lambda (s) (mapcar (lambda (f) (set! fval (unwind-protect (item.feat s f) "0")) (if (or (string-equal "" fval) (string-equal " " fval)) (format fd "%l " fval) (format fd "%s " fval))) feats) (format fd "\n")) uitems) (fclose fd))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Tree building functions (defvar wagon-balance-size 0) (define (acost:find_clusters unittypes clunits_params) "Use wagon to find the best clusters." (mapcar (lambda (unittype) (build_tree unittype clunits_params)) unittypes) t) (define (build_tree unittype clunits_params) "Build tree with Wagon for this unittype." (let ((command (format nil "%s -desc %s -data '%s' -balance %s -distmatrix '%s' -stop %s -output '%s' %s" (get_param 'wagon_progname clunits_params "wagon") (if (probe_file (string-append (get_param 'db_dir clunits_params "./") (get_param 'wagon_field_desc clunits_params "wagon") "." unittype)) ;; So there can be unittype specific desc files (string-append (get_param 'db_dir clunits_params "./") (get_param 'wagon_field_desc clunits_params "wagon") "." unittype) (string-append (get_param 'db_dir clunits_params "./") (get_param 'wagon_field_desc clunits_params "wagon"))) (string-append (get_param 'db_dir clunits_params "./") (get_param 'feats_dir clunits_params "festival/feats/") unittype (get_param 'feats_ext clunits_params ".feats")) (get_param 'wagon_balance_size clunits_params 0) (string-append (get_param 'db_dir clunits_params "./") (get_param 'disttabs_dir clunits_params "festival/disttabs/") unittype (get_param 'disttabs_ext clunits_params ".disttab")) (get_param 'wagon_cluster_size clunits_params 10) (string-append (get_param 'db_dir clunits_params "./") (get_param 'trees_dir clunits_params "festival/trees/") unittype (get_param 'trees_ext clunits_params ".tree")) (get_param 'wagon_other_params clunits_params "") ))) (format t "%s\n" command) (system command))) (defvar clunits_tree_minimum_leafs 0) (define (acost:collect_trees unittypes params) "Collect the trees into one file as an assoc list" (let ((fd (fopen (string-append (get_param 'db_dir params "./") (get_param 'trees_dir params "festival/trees/") (get_param 'index_name params "all.") (get_param 'trees_ext params ".tree")) "wb")) (tree_pref (string-append (get_param 'db_dir params "./") (get_param 'trees_dir params "festival/trees/"))) (cluster_prune_limit (get_param 'cluster_prune_limit params 0)) (cluster_merge (get_param 'cluster_merge params 0))) (format fd ";; Autogenerated list of selection trees\n") (mapcar (lambda (fp) (format fd ";; %l %l\n" (car fp) (car (cdr fp)))) params) (format fd "(set! clunits_selection_trees '(\n") (mapcar (lambda (unit) (set! tree (car (load (string-append tree_pref unit ".tree") t))) (if (> cluster_prune_limit 0) (set! tree (cluster_tree_prune tree cluster_prune_limit))) (if (> cluster_merge 0) (set! tree (tree_merge_leafs tree cluster_merge))) (if (boundp 'temp_tree_convert) (set! tree (temp_tree_convert))) (if (> (tree_num_units tree) clunits_tree_minimum_leafs) (pprintf (list unit tree) fd))) unittypes) (format fd "))\n") (fclose fd))) (define (cluster_tree_prune_in_line prune_limit) "(cluster_tree_prune_in_line) Prune number of units in each cluster in each tree *by* prune_limit, if negative, or *to* prune_limit, if positive." (set! sucs_select_trees (mapcar (lambda (t) (cluster_tree_prune t prune_limit)) sucs_select_trees))) (define (tree_merge_leafs tree depth) "(tree_merge_leafs tree depth) Merge the leafs of the tree at goven depth. This allows the trees to be pruned then the single leafs joined together into larger clusters (so the viterbi part has something to do)." (let ((num_leafs (tree_num_leafs tree))) (cond ((< num_leafs 2) tree) ;; already at the foot ((< num_leafs depth) (tree_collect_leafs tree)) (t (list (car tree) (tree_merge_leafs (car (cdr tree)) depth) (tree_merge_leafs (car (cdr (cdr tree))) depth)))))) (define (tree_num_leafs tree) "(tree_num_leafs tree) Number of leafs of given tree." (cond ((cdr tree) (+ (tree_num_leafs (car (cdr tree))) (tree_num_leafs (car (cdr (cdr tree)))))) (t 1))) (define (tree_num_units tree) "(tree_num_units tree) Number of leafs of given tree." (cond ((cdr tree) (+ (tree_num_units (car (cdr tree))) (tree_num_units (car (cdr (cdr tree)))))) (t (length (caar tree)) ))) (define (tree_collect_leafs tree) "(tree_collect_leafs tree) Combine all units in the leafs." (cond ((cdr tree) (let ((a (tree_collect_leafs (car (cdr tree)))) (b (tree_collect_leafs (car (cdr (cdr tree)))))) (list (list (append (caar a) (caar b)) 10.0)))) (t tree))) (define (cluster_tree_prune tree prune_limit) "(cluster_tree_prune TREE PRUNE_LIMIT) Reduce the number of elements in the (CART) tree leaves to PRUNE_LIMIT removing the ones further from the cluster centre. Maybe later this should have guards on minimum number of units that must remain in the tree and a per unit type limit." (cond ((cdr tree) ;; a question (list (car tree) (cluster_tree_prune (car (cdr tree)) prune_limit) (cluster_tree_prune (car (cdr (cdr tree))) prune_limit))) (t ;; tree leave (list (list (remove_n_worst (car (car tree)) (if (< prune_limit 0) (* -1 prune_limit) (- (length (car (car tree))) prune_limit))) (car (cdr (car tree)))))))) (define (remove_n_worst lll togo) "(remove_n_worst lll togo) Remove togo worst items from lll." (cond ((< togo 0) lll) ((equal? 0 togo) lll) (t (remove_n_worst (remove (worst_unit (cdr lll) (car lll)) lll) (- togo 1))))) (define (worst_unit lll worst_so_far) "(worst_unit lll worst_so_far) Returns unit with worst score in list." (cond ((null lll) worst_so_far) ((< (car (cdr worst_so_far)) (car (cdr (car lll)))) (worst_unit (cdr lll) (car lll))) (t (worst_unit (cdr lll) worst_so_far)))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Save the unit catalogue for use in the run-time index (define (acost:save_catalogue utterances clunits_params) "(acost:save_catalogue utterances clunits_params) Save the catalogue with named units with times." (let ((fd (fopen (string-append (get_param 'db_dir clunits_params "./") (get_param 'catalogue_dir clunits_params "trees/") (get_param 'index_name clunits_params "catalogue.") ".catalogue") "wb")) (num_units 0) ) (format fd "EST_File index\n") (format fd "DataType ascii\n") (format fd "NumEntries %d\n" (apply + (mapcar (lambda (u) (length (utt.relation.items u 'Segment))) utterances))) (format fd "IndexName %s\n" (get_param 'index_name clunits_params "cluser")) (format fd "EST_Header_End\n") (mapcar (lambda (u) (mapcar (lambda (s) (format fd "%s_%s %s %f %f %f\n" (item.feat s "clunit_name") (item.feat s 'occurid) (utt.feat u 'fileid) (item.feat s 'segment_start) (item.feat s 'segment_mid) (item.feat s 'segment_end))) (utt.relation.items u 'Segment))) utterances) (fclose fd))) (provide 'clunits_build.scm) festival/lib/siteinit.scm0000644000175000017500000000675712272720611014120 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Site specific initialisation file ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; If festival's internal audio playing support doesn't work on your ;; machine you can make Festival use your own program to play waveform ;; files. Uncomment the following and change "play" to the name of ;; your local program that can play files ;(Parameter.set 'Audio_Required_Format 'riff) ;(Parameter.set 'Audio_Method 'Audio_Command) ;; Apple OSX (if you can file afplay) ;(Parameter.set 'Audio_Command "afplay $FILE") ;; SOX (play) often a good alternative on cygwin and linux ;(Parameter.set 'Audio_Command "play -q $FILE") ;; Windows 7 (when sox's play doesn't work -- but this might not exit) ;(Parameter.set 'Audio_Command "c:/Windows/System32/WindowsPowerShell/v1.0/powershell -c '(New-Object Media.Soundplayer C:/cygwin'$FILE').PlaySync(); Exit;'") ;; If you want a voice different from the system installed default ;; uncomment the following line and change the name to the voice you ;; want ;(set! voice_default 'voice_cmu_us_awb_arctic_hts) (provide 'siteinit) festival/lib/engmorph.scm0000644000175000017500000001353010523416062014071 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: December 1997 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; THIS IS EXPERIMENTAL AND DOES *NOT* WORK ;;; ;;; Koskenniemi-style context rewrite rules for English Morphographemics ;;; Basically splits words into their (potential) morphemes. ;;; ;;; Based (roughly) on the rules in "Computational Morphology" ;;; Ritchie et al. MIT Press 1992. ;;; ;;; This is not a Scheme file and can't be loaded and evaluated ;;; It is designed for use with the wfst tools in the speech tools ;;; e.g. wfst_build -type kk -o engmorph.wfst -detmin engmorph.scm ;;; (KKrules engmorph (Alphabets ;; Input Alphabet (a b c d e f g h i j k l m n o p q r s t u v w x y z #) ;; Output Alphabet (a b c d e f g h i j k l m n o p q r s t u v w x y z + #) ) (Sets (LET a b c d e f g h i j k l m n o p q r s t u v w x y z) ) (Rules ;; The basic rules ( a => nil --- nil) ( b => nil --- nil) ( c => nil --- nil) ( d => nil --- nil) ( e => nil --- nil) ( f => nil --- nil) ( g => nil --- nil) ( h => nil --- nil) ( i => nil --- nil) ( j => nil --- nil) ( k => nil --- nil) ( l => nil --- nil) ( m => nil --- nil) ( n => nil --- nil) ( o => nil --- nil) ( p => nil --- nil) ( q => nil --- nil) ( r => nil --- nil) ( s => nil --- nil) ( t => nil --- nil) ( u => nil --- nil) ( v => nil --- nil) ( w => nil --- nil) ( x => nil --- nil) ( y => nil --- nil) ( z => nil --- nil) ( # => nil --- nil) ; ( _epsilon_/+ => (or LET _epsilon_/e ) --- (LET)) ( _epsilon_/+ => (or LET _epsilon_/e) --- nil) ;; The rules that do interesting things ;; Epenthesis ;; churches -> church+s ;; boxes -> box+s (e/+ <=> (or (s h) (or s x z) (i/y) (c h)) --- (s)) ;; Gemination (b/+ <=> ( (or b c d f g h j k l m n p q r s t v w z) (or a e i o u y) b ) --- ((or a e i o u))) (d/+ <=> ((or b c d f g h j k l m n p q r s t v w z) (or a e i o u y) d ) --- ((or a e i o u))) (f/+ <=> ((or b c d f g h j k l m n p q r s t v w z) (or a e i o u y) f ) --- ((or a e i o u))) (g/+ <=> ((or b c d f g h j k l m n p q r s t v w z) (or a e i o u y) g ) --- ((or a e i o u))) (m/+ <=> ((or b c d f g h j k l m n p q r s t v w z) (or a e i o u y) m ) --- ((or a e i o u))) (p/+ <=> ((or b c d f g h j k l m n p q r s t v w z) (or a e i o u y) p ) --- ((or a e i o u))) (s/+ <=> ((or b c d f g h j k l m n p q r s t v w z) (or a e i o u y) s ) --- ((or a e i o u))) (t/+ <=> ((or b c d f g h j k l m n p q r s t v w z) (or a e i o u y) t ) --- ((or a e i o u))) (z/+ <=> ((or b c d f g h j k l m n p q r s t v w z) (or a e i o u y) z ) --- ((or a e i o u))) (n/+ <=> ((or b c d f g h j k l m n p q r s t v w z) (or a e i o u y) n ) --- ((or a e i o u))) (l/+ <=> ((or b c d f g h j k l m n p q r s t v w z) (or a e i o u y) l ) --- ((or a e i o u))) (r/+ <=> ((or b c d f g h j k l m n p q r s t v w z) (or a e i o u y) r ) --- ((or a e i o u))) ;; tries->try+s ( i/y <=> ((or b c d f g h j k l m n p q r s t v w x z)) --- ((or ( e/+ s ) ( _epsilon_/+ (or a d e f h i l m n o p s w y))))) ;; Elision ;; moved -> move+ed (_epsilon_/e <=> ((or a e i o u ) (or b c d f g j k l m n p q r s t v x z)) --- ( _epsilon_/+ (or a e i o u ))) ) ) festival/lib/engmorphsyn.scm0000644000175000017500000001360110523416062014622 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: December 1997 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; THIS IS EXPERIMENTAL AND DOES *NOT* WORK ;;; ;;; ;;; An English morpho-syntax finite-state grammar ;;; This is used for morphological decomposition of unknown words ;;; specifically (only) words that are not found in the lexicon. ;;; This idea is that when an unknown word is found an attempt is made ;;; to see if it contains any well known morphological inflections or ;;; derivations, if so a better use of LTS can be made on the root, of ;;; none are found this ;;; ;;; ;;; Based on "Analysis of Unknown Words through Morphological ;;; Decomposition", Black, van de Plassche, Willians, European ACL 91. ;;; with the anyword matcher from a question by Lauri Karttunen after ;;; the talk. ;;; ;;; The suffixes and finite-state morph-syntax grammar is based ;;; (very roughly) on the rules in "Computational Morphology" ;;; Ritchie et al. MIT Press 1992. ;;; ;;; Can be compiled with ;;; wfst_build -type rg -o engmorphsyn.wfst -detmin engmorphsyn.scm ;;; ;;; The result can be combined with the morphographemic rules ;;; with ;;; wfst_build -type compose engmorph.wfst engmorphsyn.wfst -detmin -o engstemmer.wfst ;;; ;;; echo "# b o x e/+ s #" | wfst_run -wfst engstemmer.wfst -recog ;;; state 0 #/# -> 1 ;;; state 1 b/b -> 3 ;;; state 3 o/o -> 17 ;;; state 17 x/x -> 14 ;;; state 14 e/+ -> 36 ;;; state 36 s/s -> 34 ;;; state 34 #/# -> 16 ;;; OK. ;;; echo "# b o x e s #" | wfst_run -wfst engstemmer.wfst -recog ;;; state 0 #/# -> 1 ;;; state 1 b/b -> 3 ;;; state 3 o/o -> 17 ;;; state 17 x/x -> 14 ;;; state 14 e/e -> 22 ;;; state 22 s/s -> -1 (RegularGrammar engsuffixmorphosyntax ;; Sets ( (V a e i o u y) (C b c d f g h j k l m n p q r s t v w x y z) ) ;; Rules ( ;; A word *must* have a suffix to be recognized (Word -> # Syls Suffix ) (Word -> # Syls End ) ;; This matches any string of characters that contains at least one vowel (Syls -> Syl Syls ) (Syls -> Syl ) (Syl -> Cs V Cs ) (Cs -> C Cs ) (Cs -> ) (Suffix -> VerbSuffix ) (Suffix -> NounSuffix ) (Suffix -> AdjSuffix ) (VerbSuffix -> VerbFinal End ) (VerbSuffix -> VerbtoNoun NounSuffix ) (VerbSuffix -> VerbtoNoun End ) (VerbSuffix -> VerbtoAdj AdjSuffix ) (VerbSuffix -> VerbtoAdj End ) (NounSuffix -> NounFinal End ) (NounSuffix -> NountoNoun NounSuffix ) (NounSuffix -> NountoNoun End ) (NounSuffix -> NountoAdj AdjSuffix ) (NounSuffix -> NountoAdj End ) (NounSuffix -> NountoVerb VerbSuffix ) (NounSuffix -> NountoVerb End ) (AdjSuffix -> AdjFinal End ) (AdjSuffix -> AdjtoAdj AdjSuffix) (AdjSuffix -> AdjtoAdj End) (AdjSuffix -> AdjtoAdv End) ;; isn't any Adv to anything (End -> # ) ;; word boundary symbol *always* present (VerbFinal -> + e d) (VerbFinal -> + i n g) (VerbFinal -> + s) (VerbtoNoun -> + e r) (VerbtoNoun -> + e s s) (VerbtoNoun -> + a t i o n) (VerbtoNoun -> + i n g) (VerbtoNoun -> + m e n t) (VerbtoAdj -> + a b l e) (NounFinal -> + s) (NountoNoun -> + i s m) (NountoNoun -> + i s t) (NountoNoun -> + s h i p) (NountoAdj -> + l i k e) (NountoAdj -> + l e s s) (NountoAdj -> + i s h) (NountoAdj -> + o u s) (NountoVerb -> + i f y) (NountoVerb -> + i s e) (NountoVerb -> + i z e) (AdjFinal -> + e r) (AdjFinal -> + e s t) (AdjtoAdj -> + i s h) (AdjtoAdv -> + l y) (AdjtoNoun -> + n e s s) (AdjtoVerb -> + i s e) (AdjtoVerb -> + i z e) ) ) festival/lib/lts.scm0000644000175000017500000001735011061232501013050 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1998 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Functions specific to supporting a trained LTS rules ;;; (define (lts_rules_predict word feats) (let ((dcword (downcase word)) (syls) (phones)) (if (string-matches dcword "[a-z]*") (begin (set! phones (cdr (reverse (cdr (reverse (lts_predict dcword)))))) (set! phones (add_lex_stress word feats phones)) (set! syls (lex.syllabify.phstress phones)) ;; (set! syls (add_lex_stress word syls)) ) (set! syls nil)) (format t "word %l phones %l\n" word syls) (list word nil syls))) ;(define (add_lex_stress word syls) ; (cond ; ((> (length syls) 1) ; (set-car! (cdr (nth (- (length syls) 2) syls)) 1)) ; ((word-is-content word english_guess_pos) ; (set-car! (cdr (car syls)) 1))) ; syls) (define (word-is-content word guess_pos) (cond ((null guess_pos) t) ((member_string word (cdr (car guess_pos))) nil) (t (word-is-content word (cdr guess_pos))))) (defvar lts_pos nil) (define (lts_predict word rules) "(lts_predict word rules) Return list of phones related to word using CART trees." (let ((utt (make_let_utt (enworden (wordexplode word))))) (predict_phones utt rules) (cdr (reverse (cdr (reverse ;; remove #'s (mapcar (lambda (p) (intern (item.name p))) (utt.relation.items utt 'PHONE)))))) ) ) (define (wordexplode lets) (if (consp lets) lets (symbolexplode lets))) (define (make_let_utt letters) "(make_let_utt letters) Build an utterances from th4ese letters." (let ((utt (Utterance Text ""))) (utt.relation.create utt 'LTS) (utt.relation.create utt 'LETTER) (utt.relation.create utt 'PHONE) ;; Create letter stream (mapcar (lambda (l) (let ((lsi (utt.relation.append utt 'LETTER))) (item.set_feat lsi "pos" lts_pos) (item.set_name lsi l))) letters) utt)) (define (predict_phones utt rules) "(predict_phones utt) Predict phones using CART." (add_new_phone utt (utt.relation.first utt 'LETTER) '#) (mapcar (lambda (lsi) (let ((tree (car (cdr (assoc_string (item.name lsi) rules))))) (if (not tree) (format t "failed to find tree for %s\n" (item.name lsi)) (let ((p (wagon_predict lsi tree))) ; (format t "predict %s %s\n" (item.name lsi) p) (cond ((string-matches p ".*-.*-.*-.*") ; a quad one (add_new_phone utt lsi (string-before p "-")) (add_new_phone utt lsi (string-before (string-after p "-") "-")) (add_new_phone utt lsi (string-before (string-after (string-after p "-") "-") "-")) (add_new_phone utt lsi (string-after (string-after (string-after p "-") "-") "-"))) ((string-matches p ".*-.*-.*") ; a triple one (add_new_phone utt lsi (string-before p "-")) (add_new_phone utt lsi (string-before (string-after p "-") "-")) (add_new_phone utt lsi (string-after (string-after p "-") "-"))) ((string-matches p ".*-.*");; a double one (add_new_phone utt lsi (string-before p "-")) (add_new_phone utt lsi (string-after p "-"))) (t (add_new_phone utt lsi p))))))) (reverse (cdr (reverse (cdr (utt.relation.items utt 'LETTER)))))) (add_new_phone utt (utt.relation.last utt 'LETTER) '#) utt) (define (add_new_phone utt lsi p) "(add_new_phone utt lsi p) Add new phone linking to letter, ignoreing it if its _epsilon_." (if (not (equal? p '_epsilon_)) (let ((psi (utt.relation.append utt 'PHONE))) (item.set_name psi p) (item.relation.append_daughter (utt.relation.append utt 'LTS lsi) 'LTS psi) ))) (define (enworden lets) (cons '# (reverse (cons '# (reverse lets))))) ;;; Lexical stress assignment ;;; (define (add_lex_stress word pos phones tree) "(add_lex_stress word syls) Predict lexical stress by decision tree." (let ((utt (Utterance Text "")) (si) (nphones)) (utt.relation.create utt 'Letter) (set! si (utt.relation.append utt 'Letter)) (item.set_feat si 'pos pos) (item.set_feat si 'numsyls (count_syls phones)) (item.set_feat si 'sylpos 1) (set! nphones (add_lex_stress_syl phones si tree)) ; (format t "%l\n" phones) ; (format t "%l\n" nphones) nphones)) (define (count_syls phones) (cond ((null phones) 0) ((string-matches (car phones) "[aeiou@].*") (+ 1 (count_syls (cdr phones)))) (t (count_syls (cdr phones))))) (define (add_lex_stress_syl phones si tree) "(add_lex_stress_syl phones si tree) Add lexical stressing." (cond ((null phones) nil) ((string-matches (car phones) "[aeiou@].*") (item.set_feat si 'phone (car phones)) (item.set_feat si 'name (car phones)) (item.set_feat si 'num2end (- (+ 1 (item.feat si 'numsyls)) (item.feat si 'sylpos))) (set! stress (wagon_predict si tree)) (item.set_feat si 'sylpos (+ 1 (item.feat si 'sylpos))) (cons (if (not (string-equal stress "0")) (string-append (car phones) stress) (car phones)) (add_lex_stress_syl (cdr phones) si tree))) (t (cons (car phones) (add_lex_stress_syl (cdr phones) si tree))))) ;;; Morphological analysis ;(define (wfst_stemmer) ; (wfst.load 'stemmer "/home/awb/projects/morpho/engstemmer.wfst") ; (wfst.load 'stemmerL "/home/awb/projects/morpho/engstemmerL.wfst") ; t) ;(define (stem word) ; (wfst.transduce 'stemmer (enworden (symbolexplode word)))) ;(define (stemL word) ; (wfst.transduce 'stemmerL (enworden (symbolexplode word)))) (provide 'lts) festival/lib/lts_build.scm0000644000175000017500000005566412307655151014257 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1998 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Functions for building LTS rules sets from lexicons ;;; ;;; (defvar pl-table nil) (define (allaligns phones letters) "(cummulate phones lets) Aligns all possible ways for these strings." (cond ((null letters) ;; (wrongly) assume there are never less letters than phones (if phones (format t "wrong end: %s\n" word)) nil) ((null phones) nil) (t (if (< (length phones) (length letters)) (begin (cummulate '_epsilon_ (car letters)) (allaligns phones (cdr letters)))) (cummulate (car phones) (car letters)) (allaligns (cdr phones) (cdr letters))))) (define (valid-pair phone letter) "(valid-pair phone letter) If predefined to be valid." (let ((entry1 (assoc_string letter pl-table))) (if entry1 (assoc_string phone (cdr entry1)) nil))) (define (valid-pair-e phone nphone letter) "(valid-pair-e phone letter) Special cases for when epsilon may be inserted before letter." (let ((ll (assoc_string letter pl-table)) (pp (intern (string-append phone "-" nphone)))) (assoc_string pp (cdr ll)))) (define (find-aligns phones letters) "(find-aligns phones letters) Find all feasible alignments." (let ((r nil)) (cond ((and (null (cdr phones)) (null (cdr letters)) (equal? (car phones) (car letters)) (equal? '# (car phones))) (list (list (cons '# '#)))) ;; valid end match (t (if (valid-pair '_epsilon_ (car letters)) (set! r (mapcar (lambda (p) (cons (cons '_epsilon_ (car letters)) p)) (find-aligns phones (cdr letters))))) (if (valid-pair (car phones) (car letters)) (set! r (append r (mapcar (lambda (p) (cons (cons (car phones) (car letters)) p)) (find-aligns (cdr phones) (cdr letters)))))) ;; Hmm, change this to always check doubles (if (valid-pair-e (car phones) (car (cdr phones)) (car letters)) (set! r (append r (mapcar (lambda (p) (cons (cons (intern (format nil "%s-%s" (car phones) (car (cdr phones)))) (car letters)) p)) (find-aligns (cdr (cdr phones)) (cdr letters)))))) r)))) (define (findallaligns phones letters) (let ((a (find-aligns phones letters))) (if (null a) (begin (set! failedaligns (+ 1 failedaligns)) (format t "failed: %l %l\n" letters phones))) a)) (define (cummulate phone letter) "(cummulate phone letter) record the alignment of this phone and letter." (if (or (equal? phone letter) (and (not (equal? phone '#)) (not (equal? letter '#)))) (let ((entry1 (assoc_string letter pl-table)) score) (if (equal? phone '_epsilon_) (set! score 0.1) (set! score 1)) (if entry1 (let ((entry2 (assoc_string phone (cdr entry1)))) (if entry2 (set-cdr! entry2 (+ score (cdr entry2))) (set-cdr! entry1 (cons (cons phone 1) (cdr entry1))))) (set! pl-table (cons (cons letter (list (cons phone score))) pl-table))) t))) (define (score-pair phone letter) "(score-pair phone letter) Give score for this particular phone letter pair." (let ((entry1 (assoc_string letter pl-table))) (if entry1 (let ((entry2 (assoc_string phone (cdr entry1)))) (if entry2 (cdr entry2) 0)) 0))) (define (cummulate-aligns aligns) (mapcar (lambda (a) (mapcar (lambda (p) (cummulate (car p) (cdr p))) a)) aligns) t) (define (cummulate-pairs trainfile) "(cummulate-pairs trainfile) Build cummulatation table from allowable alignments in trainfile." (set! failedaligns 0) (set! allaligns 0) (if (not pl-table) (set! pl-table (mapcar (lambda (l) (cons (car l) (mapcar (lambda (c) (cons c 0)) (cdr l)))) allowables))) (let ((fd (fopen trainfile "r")) (c 0) (d 0) (entry)) (while (not (equal? (set! entry (readfp fd)) (eof-val))) (if (equal? c 1000) (begin (format t "ENTRY: %d %l\n" (set! d (+ 1000 d)) entry) (set! c 0))) (set! word (car entry)) (cummulate-aligns (findallaligns (enworden (car (cdr (cdr entry)))) (enworden (wordexplode (car entry))))) (set! allaligns (+ 1 allaligns)) (format t "aligned %d\n" allaligns) (set! c (+ 1 c))) (fclose fd) (format t "failedaligns %d/%d\n" failedaligns allaligns) )) (define (find_best_alignment phones letters) "(find_best_alignment phones letters) Find the alignement containg the most frequent alignment pairs." ;; hackily do this as a global (set! fba_best_score 0) (set! fba_best nil) (find-best-align phones letters nil 0) fba_best ) (define (find-best-align phones letters path score) "(find-best-align phones letters) Find all feasible alignments." (cond ((null letters) (if (> score fba_best_score) (begin (set! fba_best_score score) (set! fba_best (reverse path)))) nil) (t (if (valid-pair '_epsilon_ (car letters)) (find-best-align phones (cdr letters) (cons (cons '_epsilon_ (car letters)) path) (+ score (score-pair '_epsilon_ (car letters))))) (if (valid-pair (car phones) (car letters)) (find-best-align (cdr phones) (cdr letters) (cons (cons (car phones) (car letters))path) (+ score (score-pair (car phones) (car letters))))) (if (valid-pair-e (car phones) (car (cdr phones)) (car letters)) (find-best-align (cdr (cdr phones)) (cdr letters) (cons (cons (intern (format nil "%s-%s" (car phones) (car (cdr phones)))) (car letters)) path) (+ score (score-pair (intern (format nil "%s-%s" (car phones) (car (cdr phones)))) (car letters)))))))) (define (align_and_score phones letters path score) "(align_and_score phones lets) Aligns all possible ways for these strings." (cond ((null letters) (if (> score fba_best_score) (begin (set! fba_best_score score) (set! fba_best (reverse path)))) nil) (t (if (< (length phones) (length letters)) (align_and_score phones (cdr letters) (cons '_epsilon_ path) (+ score (score-pair '_epsilon_ (car letters))))) (align_and_score (cdr phones) (cdr letters) (cons (car phones) path) (+ score (score-pair (car phones) (car letters))))))) (define (aligndata file ofile) (let ((fd (fopen file "r")) (ofd (fopen ofile "w")) (c 1) (entry)) (while (not (equal? (set! entry (readfp fd)) (eof-val))) (set! lets (enworden (wordexplode (car entry)))) (set! bp (find_best_alignment (enworden (car (cdr (cdr entry)))) lets)) (if (not bp) (format t "align failed: %l\n" entry) (save_info (car (cdr entry)) bp ofd)) (set! c (+ 1 c))) (fclose fd) (fclose ofd))) (define (enworden lets) (cons '# (reverse (cons '# (reverse lets))))) (define (wordexplode lets) (if (consp lets) lets (symbolexplode lets))) (define (save_info pos bp ofd) "(save_info pos bp ofd) Cut out one expensive step and 50M of diskspace and just save it in a simpler format." (format ofd "( ( ") (mapcar (lambda (l) (if (not (string-equal "#" (cdr l))) (format ofd "%l " (cdr l)))) bp) (format ofd ") %l" pos) (mapcar (lambda (l) (if (not (string-equal "#" (car l))) (format ofd " %s" (car l)))) bp) (format ofd " )\n")) (define (normalise-table pl-table) "(normalise-table pl-table) Change scores into probabilities." (mapcar (lambda (s) (let ((sum (apply + (mapcar cdr (cdr s))))) (mapcar (lambda (p) (if (equal? sum 0) (set-cdr! p 0) (set-cdr! p (/ (cdr p) sum)))) (cdr s)))) pl-table) t) (define (save-table pre) (normalise-table pl-table) (set! fd (fopen (string-append pre "pl-tablesp.scm") "w")) (format fd "(set! pl-table '\n") (pprintf pl-table fd) (format fd ")\n") (fclose fd) t) (define (build-feat-file alignfile featfile) "(build-feat-file alignfile featfile) Build a feature file from the given align file. The feature file contain predicted phone, and letter with 3 preceding and 3 succeeding letters." (let ((fd (fopen alignfile "r")) (ofd (fopen featfile "w")) (entry) (pn) (sylpos 1)) (while (not (equal? (set! entry (readfp fd)) (eof-val))) ;; (format t "read: %l\n" entry) (set! lets (append '(0 0 0 0 #) (wordexplode (car entry)) '(# 0 0 0 0))) (set! phones (cdr (cdr entry))) (set! pn 5) (mapcar (lambda (p) (format ofd "%s %s %s %s %s %s %s %s %s %s %s\n" p (nth (- pn 4) lets) (nth (- pn 3) lets) (nth (- pn 2) lets) (nth (- pn 1) lets) (nth pn lets) (nth (+ pn 1) lets) (nth (+ pn 2) lets) (nth (+ pn 3) lets) (nth (+ pn 4) lets) (cond ((not (consp (car (cdr entry)))) (car (cdr entry))) ((not (consp (caar (cdr entry)))) (caar (cdr entry))) (t nil)) ;; sylpos ;; numsyls ;; num2end ) (set! pn (+ 1 pn))) phones)) (fclose fd) (fclose ofd)) ) (define (merge_models name filename allowables) "(merge_models name filename) Merge the models into a single list of cart trees as a variable named by name, in filename." (require 'cart_aux) (let (trees fd) (set! trees nil) (set! lets (mapcar car allowables)) (while lets (if (probe_file (format nil "lts.%s.tree" (car lets))) (begin (format t "%s\n" (car lets)) (set! tree (car (load (format nil "lts.%s.tree" (car lets)) t))) (set! tree (cart_simplify_tree2 tree nil)) (set! trees (cons (list (car lets) tree) trees)))) (set! lets (cdr lets))) (set! trees (reverse trees)) (set! fd (fopen filename "w")) (format fd ";; LTS rules \n") (format fd "(set! %s '(\n" name) (mapcar (lambda (tree) (pprintf tree fd)) trees) (format fd "))\n") (fclose fd)) ) (define (lts_testset file cartmodels) "(lts_testset file cartmodels) Test an aligned lexicon file against a set of cart trees. Prints out The number of letters correct (for each letter), total number of letters correct and the total number of words correct. cartmodels is the structure as saved by merge_models." (let ((fd (fopen file "r")) (entry) (wordcount 0) (correctwords 0) (phonecount 0) (correctphones 0)) (while (not (equal? (set! entry (readfp fd)) (eof-val))) (let ((letters (enworden (wordexplode (car entry)))) (phones (enworden (cdr (cdr entry)))) (pphones)) (set! wordcount (+ 1 wordcount)) (set! pphones (gen_cartlts letters (car (cdr entry)) cartmodels)) ; (set! pphones ; (or ; unwind-protect ; (gen_vilts letters (car (cdr entry)) ; cartmodels wfstname) ; nil)) (if (equal? (ph-normalize pphones) (ph-normalize phones)) (set! correctwords (+ 1 correctwords)) (or nil (format t "failed %l %l %l %l\n" (car entry) (car (cdr entry)) phones pphones))) (count_correct_letters ;; exclude #, cause they're always right (cdr letters) (cdr phones) (cdr pphones)) (set! phonecount (+ (length (cdr (cdr letters))) phonecount)) )) (fclose fd) (mapcar (lambda (linfo) (format t "%s %d correct %d (%2.2f)\n" (car linfo) (car (cdr linfo)) (car (cdr (cdr linfo))) (/ (* (car (cdr (cdr linfo))) 100) (car (cdr linfo)))) (set! correctphones (+ correctphones (car (cdr (cdr linfo)))))) correct_letter_table) (format t "phones %d correct %d (%2.2f)\n" phonecount correctphones (/ (* correctphones 100) phonecount)) (format t "words %d correct %d (%2.2f)\n" wordcount correctwords (/ (* correctwords 100) wordcount)) (format t "tree model has %d nodes\n" (apply + (mapcar (lambda (a) (cart_tree_node_count (car (cdr a)))) cartmodels))) )) (define (cart_tree_node_count tree) "(tree_node_count tree) Count the number nodes (questions and leafs) in the given CART tree." (cond ((cdr tree) (+ 1 (cart_tree_node_count (car (cdr tree))) (cart_tree_node_count (car (cdr (cdr tree)))))) (t 1))) (defvar correct_letter_table (mapcar (lambda (l) (list l 0 0)) '(a b c d e f g h i j k l m n o p q r s t u v w x y z)) "correct_letter_table List used to cummulate the number of correct (and incorrect) letter to phone predictions. This list will be extended if there are more letters in your alphabet, though it doesn't take a fairly western european view of the alphabet, but you can change this yourself is necessary.") (define (count_correct_letters lets phs pphs) "(count_correct_letters lets phs pphs) Count which letters have the correct phone prediction. Cummulate this is a per letter table." (cond ((or (null phs) (null pphs) (null lets)) (format t "misaligned entry\n") nil) ((and (null (cdr lets)) (null (cdr phs)) (null (cdr pphs))) nil) ;; omit final # (t (let ((letinfo (assoc_string (car lets) correct_letter_table))) (if (not letinfo) (set! correct_letter_table (append correct_letter_table (list (set! letinfo (list (car lets) 0 0)))))) (set-car! (cdr letinfo) (+ 1 (car (cdr letinfo)))) ;; total (if (equal? (car phs) (car pphs)) ;; correct (set-car! (cdr (cdr letinfo)) (+ 1 (car (cdr (cdr letinfo)))))) (count_correct_letters (cdr lets) (cdr phs) (cdr pphs)))))) (define (ph-normalize ph) (cond ((null ph) nil) ((string-equal "_epsilon_" (car ph)) (ph-normalize (cdr ph))) ((string-matches (car ph) ".*-.*") (cons (string-before (car ph) "-") (cons (string-after (car ph) "-") (ph-normalize (cdr ph))))) (t (cons (car ph) (ph-normalize (cdr ph)))))) (define (make_let_utt_p letters pos) "(make_let_utt letters) Build an utterances from th4ese letters." (let ((utt (Utterance Text ""))) (utt.relation.create utt 'LTS) (utt.relation.create utt 'LETTER) (utt.relation.create utt 'PHONE) ;; Create letter stream (mapcar (lambda (l) (let ((lsi (utt.relation.append utt 'LETTER))) (item.set_name lsi l) (item.set_feat lsi "pos" pos))) letters) utt)) (define (gen_vilts letters pos cartmodels ngram) "(get_vilts letters pos cartmodels ngram) Use cart plus ngrams in viterbi search." (require 'lts) (let ((utt (make_let_utt_p letters pos))) (set! gen_vit_params (list (list 'Relation "LETTER") (list 'return_feat "phone") (list 'p_word "#") (list 'pp_word "0") (list 'ngramname ngram) ; (list 'wfstname ngram) (list 'cand_function 'lts_cand_function))) (Gen_Viterbi utt) (mapcar (lambda (lsi) (intern (item.feat lsi "phone"))) (utt.relation.items utt 'LETTER)))) (define (gen_cartlts letters pos cartmodels) "(get_cartlts letters cartmodels) Generate the full list of predicted phones, including epsilon and unexpanded multi-phones." (require 'lts) (let ((utt (make_let_utt_p letters pos))) (enworden (mapcar (lambda (lsi) (let ((tree (car (cdr (assoc_string (item.name lsi) cartmodels)))) (p)) (if (not tree) (begin (format t "failed to find tree for %s\n" (item.name lsi)) nil) (begin (set! p (wagon_predict lsi tree)) (item.set_feat lsi "val" p) p)))) (reverse (cdr (reverse (cdr (utt.relation.items utt 'LETTER))))))))) (define (reduce_lexicon entryfile exceptionfile lts_function) "(reduce_lexicon entryfile exceptionfile lts_function) Look up each word in entryfile using the current lexicon, if the entry doesn't match save it in the exception file. This is a way of reducing the lexicon based on a letter to sound model (and lexical stress model, if appropriate)." (let ((fd (fopen entryfile "r")) (ofd (fopen exceptionfile "w")) (entry) (wordcount 0) (correctwords 0)) (while (not (equal? (set! entry (readfp fd)) (eof-val))) (if (and (consp entry) (> (length entry) 1)) (let ((lts (lts_function (car entry) (car (cdr entry)))) (encount (lex.entrycount (car entry)))) (set! wordcount (+ 1 wordcount)) (if (and (equal? (nth 2 entry) (nth 2 lts)) (< encount 2)) (set! correctwords (+ 1 correctwords)) (format ofd "%l\n" entry)) ))) (fclose fd) (fclose ofd) (format t "words %d correct %d (%2.2f)\n" wordcount correctwords (/ (* correctwords 100) wordcount)) )) (define (dump-flat-entries infile outfile ltype) (let ((ifd (fopen infile "r")) (ofd (fopen outfile "w")) clength entry) ; (set! entry (readfp ifd)) ; (if (or (consp entry) (not (string-equal entry "MNCL"))) ; (begin ; (format t "Expected MNCL at start of file: not a compiled lexicon\n") ; (exit))) (while (not (equal? (set! entry (readfp ifd)) (eof-val))) (cond ((not (consp entry)) t) ;; not an entry ((string-equal ltype "utf8") (set! clength (length (utf8explode (car entry))))) (t (set! clength (length (car entry))))) (cond ((not (consp entry)) t) ;; not an entry ((and ;(string-matches (car entry) "...*") ;(< clength 14) (not (string-matches (car entry) ".*'.*")) ;; no quotes (car (cddr entry))) ;; non-nil pronounciation (begin (cond ((string-equal ltype "utf8") (format ofd "( %l %l (" (utf8explode (car entry)) (cadr entry))) ((string-equal ltype "asis") (format ofd "( \"%s\" %l (" (car entry) (cadr entry))) (t (format ofd "( \"%s\" %l (" (downcase (car entry)) (cadr entry)))) (if (consp (car (car (cddr entry)))) (begin ;; it is syllabified) (mapcar (lambda (syl) (mapcar (lambda (seg) (cond ((string-matches seg "[aeiouAEIOU@].*") (format ofd "%s " (string-append seg (cadr syl)))) (t (format ofd "%s " seg)))) (car syl))) (car (cddr entry)))) (begin ;; it is already flat (mapcar (lambda (p) (format ofd "%s " p)) (car (cddr entry))) )) (format ofd "))\n"))) (t nil))) (fclose ifd) (fclose ofd))) (define (dump-lets-phones infile) "(dump-lets-phones infile) Dump all the letters to alllets.out and phones to allphones.out for processing. This expects an external script to sort and uniquify them. This is done in scheme so we can get utf8/non-utf8 to be easy." (let ((ifd (fopen infile "r")) (lfd (fopen "alllets.out" "w")) (apfd (fopen "allphones.out" "w")) (pfd (fopen "let2phones.out" "w")) entry) (while (not (equal? (set! entry (readfp ifd)) (eof-val))) (mapcar (lambda (l) (format lfd "%s\n" l) (format pfd "%s " l) (mapcar (lambda (p) (format pfd "%s " p)) (car (cddr entry))) (format pfd "\n")) (wordexplode (car entry))) (mapcar (lambda (p) (format apfd "%s " p)) (car (cddr entry))) (format apfd "\n") ) (fclose ifd) (fclose lfd) (fclose pfd) (fclose apfd) t)) (define (dump-flat-entries-all infile outfile) "(dump-flat-entries-all infile outfile) Do this for *all* entries not just ones with more than three chars." (let ((ifd (fopen infile "r")) (ofd (fopen outfile "w")) entry) (readfp ifd) ;; skip "MNCL" (while (not (equal? (set! entry (readfp ifd)) (eof-val))) (if (consp entry) (begin (format ofd "( \"%s\" %s (" (downcase (car entry)) (cadr entry)) (mapcar (lambda (syl) (mapcar (lambda (seg) (cond ; ((string-equal seg "ax") ; (format ofd "%s " seg)) ((string-matches seg "[aeiouAEIOU@].*") (format ofd "%s " (string-append seg (cadr syl)))) (t (format ofd "%s " seg)))) (car syl))) (car (cddr entry))) (format ofd "))\n")))) (fclose ifd) (fclose ofd))) (provide 'lts_build) festival/lib/Sable.v0_2.dtd0000644000175000017500000001445707262605650014060 0ustar awbawb %ISOlat1; festival/lib/sable-latin.ent0000644000175000017500000001766607262616703014501 0ustar awbawb festival/lib/festival.el0000644000175000017500000002553710235433645013725 0ustar awbawb;;; ;;; File: festival.el ;;; Emacs Lisp ;;; ;;; Alan W Black CSTR (awb@cstr.ed.ac.uk) June 1996 ;;; ;;; Provide an emacs mode for interfacing to the festival speech ;;; synthesizer system ;;; ;;; I've looked at many examples from the emacs Lisp directory ;;; copying relevant bits from here and there, so this can only ;;; reasonably inherit the GNU licence (GPL) ;;; ;;; Setup: ;;; In your .emacs add the following 2 lines to get a Say menu: ;;; ;;; (autoload 'say-minor-mode "festival" "Menu for using Festival." t) ;;; (say-minor-mode t) ;;; (setq auto-mode-alist ;;; (append '(("\\.festivalrc$" . scheme-mode)) auto-mode-alist)) ;;; ;;; The following gives you pretty colors in emacs-19 if you are into ;;; such things ;;; ;;; Some colors for scheme mode ;;; (hilit-set-mode-patterns ;;; '(scheme-mode) ;;; '( ;;; (";.*" nil comment) ;;; (hilit-string-find ?\\ string) ;;; ("^\\s *(def\\s +" "\\()\\|nil\\)" defun) ;;; ("^\\s *(defvar\\s +\\S +" nil decl) ;;; ("^\\s *(set\\s +\\S +" nil decl) ;;; ("^\\s *(defconst\\s +\\S +" nil define) ;;; ("^\\s *(\\(provide\\|require\\).*$" nil include) ;;; ("(\\(let\\*?\\|cond\\|if\\|or\\|and\\|map\\(car\\|concat\\)\\|prog[n1*]?\\|while\\|lambda\\|function\\|Parameter\\|set\\([qf]\\|car\\|cdr\\)?\\|nconc\\|eval-when-compile\\|condition-case\\|unwind-protect\\|catch\\|throw\\|error\\)[ \t\n]" 1 keyword))) ;;; ;;; ;;;-------------------------------------------------------------------- ;;; Copyright (C) Alan W Black 1996 ;;; This code is distributed in the hope that it will be useful, ;;; but WITHOUT ANY WARRANTY. No author or distributor accepts ;;; responsibility to anyone for the consequences of using this code ;;; or for whether it serves any particular purpose or works at all, ;;; unless explicitly stated in a written agreement. ;;; ;;; Everyone is granted permission to copy, modify and redistribute ;;; this code, but only under the conditions described in the GNU ;;; Emacs General Public License. A copy of this license is ;;; distrubuted with GNU Emacs so you can know your rights and ;;; responsibilities. It should be in a file named COPYING. Among ;;; other things, the copyright notice and this notice must be ;;; preserved on all copies. ;;;-------------------------------------------------------------------- ;;; (defvar festival-program-name "festival") (defvar festival-process nil) (defvar festival-tmp-file (format "/tmp/festival-emacs-tmp-%s" (user-real-login-name)) "Filename to save input for Festivial.") (defun festival-fast () (interactive) (festival-send-command '(Parameter.set 'Duration.Stretch 0.8))) (defun festival-slow () (interactive) (festival-send-command '(Parameter.set 'Duration.Stretch 1.2))) (defun festival-ndur () (interactive) (festival-send-command '(Parameter.set 'Duration.Stretch 1.0))) (defun festival-intro () (interactive) (festival-send-command '(intro))) (defun festival-gsw () (interactive) (festival-send-command '(voice_gsw_diphone))) (defun festival-rab () (interactive) (festival-send-command '(voice_rab_diphone))) (defun festival-ked () (interactive) (festival-send-command '(voice_ked_diphone))) (defun festival-kal () (interactive) (festival-send-command '(voice_kal_diphone))) (defun festival-don () (interactive) (festival-send-command '(voice_don_diphone))) (defun festival-welsh () (interactive) (festival-send-command '(voice_welsh_hl))) (defun festival-spanish () (interactive) (festival-send-command '(voice_spanish_el))) (defun festival-say-string (string) "Send string to festival and have it said" (interactive "sSay: ") (festival-start-process) (process-send-string festival-process (concat "(SayText " (format "%S" string) ") "))) (defun festival-send-command (cmd) "Send command to festival" (interactive "px") (festival-start-process) (process-send-string festival-process (format "%S " cmd))) (defun festival-process-status () (interactive) (if festival-process (message (format "Festival process status: %s" (process-status festival-process))) (message (format "Festival process status: NONE")))) (defun festival-start-process () "Check status of process and start it if necessary" (interactive ) (let ((process-connection-type t)) (if (and festival-process (eq (process-status festival-process) 'run)) 't ;;(festival-kill-festival t) (message "Starting new synthesizer process...") (sit-for 0) (setq festival-process (start-process "festival" (get-buffer-create "*festival*") festival-program-name))) )) (defun festival-kill-process () "Kill festival sub-process" (interactive) (if festival-process (kill-process festival-process)) (setq festival-process nil) (message "Festival process killed")) (defun festival-send-string (string) "Send given string to fesitval process." (interactive) (festival-start-process) (process-send-string festival-process string)) (defun festival-say-region (reg-start reg-end) "Send given region to festival for saying. This saves the region as a file in /tmp and then tells festival to say that file. The major mode is *not* passed as text mode name to Festival." (interactive "r") (write-region reg-start reg-end festival-tmp-file) (festival-send-command (list 'tts festival-tmp-file nil))) (defun festival-say-buffer () "Send given region to festival for saying. This saves the region as a file in /tmp and then tells festival to say that file. The major-mode is passed as a text mode to Festival." (interactive) (write-region (point-min) (point-max) festival-tmp-file) ;; Because there may by sgml-like sub-files mentioned ;; ensure festival tracks the buffer's default-directory (festival-send-command (list 'cd (expand-file-name default-directory))) (if (equal "-mode" (substring (format "%S" major-mode) -5 nil)) (if (equal "sgml" (substring (format "%S" major-mode) 0 -5)) (festival-send-command (list 'tts festival-tmp-file "sable")) (festival-send-command (list 'tts festival-tmp-file (substring (format "%S" major-mode) 0 -5)))) (festival-send-command (list 'tts festival-tmp-file nil)))) ;; ;; say-minor-mode provides a menu offering various speech synthesis commands ;; (defvar say-minor-mode nil) (defun say-minor-mode (arg) "Toggle say minor mode. With arg, turn say-minor-mode on iff arg is positive." (interactive "P") (setq say-minor-mode (if (if (null arg) (not say-minor-mode) (> (prefix-numeric-value arg) 0)) t)) (force-mode-line-update)) (setq say-params-menu (make-sparse-keymap "Pitch/Duration")) (fset 'say-params-menu (symbol-value 'say-params-menu)) (define-key say-params-menu [say-fast] '("Fast" . festival-fast)) (define-key say-params-menu [say-slow] '("Slow" . festival-slow)) (define-key say-params-menu [say-ndur] '("Normal Dur" . festival-ndur)) (setq say-lang-menu (make-sparse-keymap "Select language")) (fset 'say-lang-menu (symbol-value 'say-lang-menu)) (define-key say-lang-menu [say-lang-spain1] '("Spanish el" . festival-spanish)) (define-key say-lang-menu [say-lang-welsh1] '("Welsh hl" . festival-welsh)) (define-key say-lang-menu [say-lang-eng5] '("English gsw" . festival-gsw)) (define-key say-lang-menu [say-lang-eng4] '("English don" . festival-don)) (define-key say-lang-menu [say-lang-eng3] '("English rab" . festival-rab)) (define-key say-lang-menu [say-lang-eng2] '("English ked" . festival-ked)) (define-key say-lang-menu [say-lang-eng1] '("English kal" . festival-kal)) ;(define-key say-params-menu [say-set-dur-stretch] ; '("Set Duration Stretch" . festival-set-dur-stretch)) ;(define-key say-params-menu [say-high] '("High" . festival-high)) ;(define-key say-params-menu [say-low] '("Low" . festival-low)) ;(define-key say-params-menu [say-npit] '("Normal Pitch" . festival-npit)) ;(define-key say-params-menu [say-set-pitch-stretch] ; '("Set Pitch Stretch" . festival-set-pitch-stretch)) (setq say-minor-mode-map (make-sparse-keymap)) (setq say-menu (make-sparse-keymap "SAY")) (define-key say-minor-mode-map [menu-bar SAY] (cons "Say" say-menu)) (define-key say-minor-mode-map [menu-bar SAY festival-intro] '("Festival Intro" . festival-intro)) (define-key say-minor-mode-map [menu-bar SAY festival-process-status] '("Festival status" . festival-process-status)) (define-key say-minor-mode-map [menu-bar SAY festival-kill-process] '("Kill Festival" . festival-kill-process)) (define-key say-minor-mode-map [menu-bar SAY festival-start-process] '("(Re)start Festival" . festival-start-process)) ;;(define-key say-menu [separator-process] '("--")) ;;(define-key say-menu [params] '("Pitch/Durations" . say-params-menu)) (define-key say-menu [separator-buffers] '("--")) (define-key say-menu [festival-send-command] '("Festival eval command" . festival-send-command)) (define-key say-menu [say-lang-menu] '("Select language" . say-lang-menu)) (define-key say-menu [festival-say-buffer] '("Say buffer" . festival-say-buffer)) (define-key say-menu [festival-say-region] '("Say region" . festival-say-region)) (setq minor-mode-map-alist (cons (cons 'say-minor-mode say-minor-mode-map) minor-mode-map-alist)) (or (assq 'say-minor-mode minor-mode-alist) (setq minor-mode-alist (cons '(say-minor-mode "") minor-mode-alist))) ;;; ;;; A FESTIVAL inferior mode (copied from prolog.el) ;;; (defvar inferior-festival-mode-map nil) (defun inferior-festival-mode () "Major mode for interacting with an inferior FESTIVAL process. The following commands are available: \\{inferior-festival-mode-map} Entry to this mode calls the value of `festival-mode-hook' with no arguments, if that value is non-nil. Likewise with the value of `comint-mode-hook'. `festival-mode-hook' is called after `comint-mode-hook'. You can send text to the inferior FESTIVAL from other buffers using the commands `send-region', `send-string' Return at end of buffer sends line as input. Return not at end copies rest of line to end and sends it. \\[comint-kill-input] and \\[backward-kill-word] are kill commands, imitating normal Unix input editing. \\[comint-interrupt-subjob] interrupts the shell or its current subjob if any. \\[comint-stop-subjob] stops. \\[comint-quit-subjob] sends quit signal." (interactive) (require 'comint) (comint-mode) (setq major-mode 'inferior-festival-mode mode-name "Inferior FESTIVAL" comint-prompt-regexp "^festival> ") (if inferior-festival-mode-map nil (setq inferior-festival-mode-map (copy-keymap comint-mode-map)) (festival-mode-commands inferior-festival-mode-map)) (use-local-map inferior-festivalr-mode-map) (run-hooks 'festival-mode-hook)) ;;;###autoload (defun run-festival () "Run an inferior FESTIVAL process, input and output via buffer *festival*." (interactive) (require 'comint) (switch-to-buffer (make-comint "festival" festival-program-name)) (inferior-festival-mode)) (provide 'festival) festival/lib/scfg_wsj_wp20.gram0000644000175000017500000003750010045556464015110 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;-*-mode:scheme-*- ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; A Stochastic context free grammar for the wp20 tag set with 19 ;;; nonterminals ;;; ;;; This was trained from 10,000 sentences (00-04) of the UPenn WSJ tree ;;; bank using the inside-outside algorithm seeded with the bracketing from ;;; the treebank. The implementation is the scfg_ suite in the ;; speech tools and is based on the paper "Inside-Outside ;;; Reestimation from partially bracketed corpora", F Pereira and ;;; Y. Schabes. pp 128-135, 30th ACL, Newark, Delaware 1992. ;;; ;;; This grammar with 19 nonterminals was trained for 174 passes ;;; using a fifth of training data each time. It was tested against ;;; independent data both bracketed and unbracketed. After training, ;;; all rules with a probability less than 1.0e-6 were pruned. ;;; ;;; On an unseen test set of 686 sentences (from wsj/05/) this gets ;;; 92.2397% bracketing accuracy and 29.5918% sentences fully correct ;;; ;;; previous best 15_20 grammar ;;; 90.2377% bracketing accuracy and 24.7813% sentences fully correct ;;; ;;; Training this grammar took a long time. This is best grammar ;;; by testing grammars varying the number of non-terminals from 11-25 ;;; as the number of NTs increases the time for training also increases ;;; This 19_20 grammar took 20 days on a Sun Ultra 1 140, but I also ;;; had to search 11-18 to confirm this is best, which was done with a ;;; collection of Ultra 140s 170s and Pentium Pros (Linux and FreeBSD) ;;; (0.00593452 NT00 NT00 NT00) (0.0319023 NT00 NT00 NT13) (0.00105452 NT00 NT00 NT18) (0.00061816 NT00 NT02 NT10) (0.000399698 NT00 NT02 NT12) (0.0383818 NT00 NT05 NT00) (0.00011458 NT00 NT06 NT03) (0.00164298 NT00 NT06 NT17) (0.00153884 NT00 NT07 NT07) (0.00118244 NT00 NT07 NT12) (0.00171642 NT00 NT07 NT13) (0.00031308 NT00 NT07 NT17) (0.0949408 NT00 NT09 NT18) (0.000932166 NT00 NT10 NT03) (0.000150288 NT00 NT10 NT17) (0.0152371 NT00 NT12 NT18) (0.73409 NT00 NT14 NT13) (0.0403652 NT00 NT14 NT18) (0.000195643 NT00 NT16 NT07) (0.0134222 NT00 NT18 NT13) (0.015624 NT00 NT18 NT18) (0.00251118 NT01 NT01 NT07) (0.00354571 NT01 NT01 NT11) (0.22337 NT01 NT01 NT16) (0.0467048 NT01 NT02 NT05) (0.000518329 NT01 NT04 NT01) (0.000100574 NT01 NT06 NT05) (0.0480904 NT01 NT07 NT05) (0.000358197 NT01 NT11 NT11) (0.00278007 NT01 NT16 NT05) (0.000179198 NT01 NT16 NT15) (0.00140099 NT01 n) (0.00228587 NT01 v) (0.524988 NT01 dt) (0.00128028 NT01 in) (0.0660845 NT01 j) (0.0131026 NT01 cd) (0.00584238 NT01 r) (0.0548382 NT01 prp) (0.000445004 NT01 wdt) (0.00135794 NT01 wp) (0.000195991 NT01 wrb) (0.000264526 NT02 NT01 NT01) (0.00243627 NT02 NT01 NT02) (0.613543 NT02 NT01 NT07) (0.00180865 NT02 NT01 NT11) (0.0042804 NT02 NT01 NT16) (0.0392418 NT02 NT02 NT07) (0.026104 NT02 NT02 NT12) (0.000916683 NT02 NT02 NT16) (0.00158862 NT02 NT04 NT01) (0.000206161 NT02 NT04 NT02) (0.00343189 NT02 NT04 NT16) (0.000417113 NT02 NT07 NT05) (0.0988457 NT02 NT07 NT07) (0.000931386 NT02 NT07 NT11) (0.00073236 NT02 NT07 NT12) (0.000153421 NT02 NT10 NT13) (0.00163484 NT02 NT11 NT02) (0.0379562 NT02 NT11 NT07) (0.0149 NT02 NT11 NT11) (0.00105811 NT02 NT11 NT12) (0.000175184 NT02 NT16 NT02) (0.0403395 NT02 NT16 NT07) (0.00297703 NT02 NT16 NT12) (0.0875026 NT02 n) (0.00496719 NT02 v) (0.000409658 NT02 dt) (0.00239978 NT02 j) (0.010203 NT02 r) (0.000194628 NT02 pdt) (0.000377009 NT03 NT04 NT02) (0.11551 NT03 NT08 NT13) (0.347629 NT03 NT09 NT13) (0.484911 NT03 NT10 NT13) (0.00188291 NT03 NT11 NT12) (0.0495461 NT03 NT17 NT13) (0.00918797 NT04 NT03 NT05) (0.000303954 NT04 NT04 NT02) (0.00284848 NT04 NT04 NT04) (0.00710115 NT04 NT04 NT12) (0.000597744 NT04 NT04 NT15) (0.000377075 NT04 NT04 NT16) (0.00130088 NT04 NT09 NT05) (0.00175428 NT04 NT10 NT13) (0.000127716 NT04 NT15 NT04) (0.00013648 NT04 NT15 NT06) (0.00045093 NT04 NT15 NT07) (0.000626479 NT04 NT15 NT16) (0.000563588 NT04 NT16 NT15) (0.0232089 NT04 NT17 NT05) (0.000138094 NT04 NT17 NT15) (0.00094009 NT04 n) (0.671108 NT04 v) (0.0150619 NT04 punc) (0.00056566 NT04 dt) (0.144629 NT04 r) (0.00270621 NT04 prp) (0.0449587 NT04 to) (0.0543755 NT04 md) (0.00839747 NT04 wdt) (0.00813689 NT04 wp) (0.000560496 NT05 NT07 NT05) (0.000901219 NT05 NT15 NT07) (0.180172 NT05 punc) (0.533041 NT05 cc) (0.285244 NT05 pos) (0.00164003 NT06 NT00 NT13) (0.00222915 NT06 NT01 NT06) (0.275903 NT06 NT01 NT07) (0.00191616 NT06 NT01 NT11) (0.00316549 NT06 NT01 NT12) (0.000730143 NT06 NT01 NT14) (0.000559842 NT06 NT02 NT06) (0.0236744 NT06 NT02 NT07) (0.00284929 NT06 NT02 NT09) (0.155052 NT06 NT02 NT12) (0.00387995 NT06 NT02 NT14) (0.0161403 NT06 NT02 NT18) (0.000110944 NT06 NT04 NT01) (0.00237845 NT06 NT04 NT02) (0.00625142 NT06 NT04 NT06) (0.00118802 NT06 NT04 NT08) (0.000132901 NT06 NT04 NT10) (0.000192545 NT06 NT04 NT11) (0.000199118 NT06 NT06 NT01) (0.0081704 NT06 NT06 NT12) (0.00198439 NT06 NT06 NT14) (0.000889455 NT06 NT06 NT18) (0.00142038 NT06 NT07 NT05) (0.0820095 NT06 NT07 NT07) (0.000112894 NT06 NT07 NT09) (0.0220243 NT06 NT07 NT12) (0.000133911 NT06 NT07 NT14) (0.00100807 NT06 NT07 NT17) (0.000191764 NT06 NT08 NT13) (0.000340112 NT06 NT10 NT08) (0.000126776 NT06 NT10 NT09) (0.0136266 NT06 NT10 NT12) (0.00867414 NT06 NT10 NT13) (0.00341334 NT06 NT10 NT18) (0.00154851 NT06 NT11 NT12) (0.00104947 NT06 NT12 NT12) (0.000219189 NT06 NT14 NT05) (0.00313879 NT06 NT14 NT13) (0.000745073 NT06 NT15 NT02) (0.000433144 NT06 NT15 NT06) (0.000159867 NT06 NT15 NT16) (0.00124313 NT06 NT16 NT02) (0.00918606 NT06 NT16 NT07) (0.00373496 NT06 NT16 NT12) (0.014053 NT06 NT18 NT13) (0.0155714 NT06 n) (0.00123379 NT06 punc) (0.0152764 NT06 dt) (0.00123486 NT06 j) (0.00359625 NT06 r) (0.212966 NT06 prp) (0.00199168 NT06 cc) (0.0383471 NT06 wdt) (0.0182587 NT06 wp) (0.00204833 NT06 wrb) (0.0109929 NT06 ex) (0.0011995 NT07 NT05 NT16) (0.119588 NT07 NT07 NT07) (0.000353596 NT07 NT07 NT11) (0.000177793 NT07 NT07 NT12) (0.00101956 NT07 NT11 NT11) (0.000357614 NT07 NT15 NT01) (0.00084812 NT07 NT15 NT06) (0.0182872 NT07 NT16 NT07) (0.00018607 NT07 NT16 NT11) (0.856315 NT07 n) (0.000736333 NT07 v) (0.000645479 NT08 NT00 NT09) (0.000990156 NT08 NT01 NT02) (0.0410251 NT08 NT01 NT07) (0.0013863 NT08 NT01 NT09) (0.000242552 NT08 NT01 NT12) (0.00174478 NT08 NT01 NT14) (0.000596656 NT08 NT01 NT16) (0.00130945 NT08 NT02 NT07) (0.166303 NT08 NT02 NT09) (0.0143253 NT08 NT02 NT12) (0.0113813 NT08 NT02 NT14) (0.000597887 NT08 NT02 NT16) (0.0133053 NT08 NT03 NT09) (0.0109076 NT08 NT03 NT17) (0.000211313 NT08 NT04 NT01) (0.0105796 NT08 NT04 NT02) (0.00440181 NT08 NT04 NT04) (0.00203737 NT08 NT04 NT06) (0.213275 NT08 NT04 NT08) (0.0781169 NT08 NT04 NT09) (0.0190657 NT08 NT04 NT10) (0.00319326 NT08 NT04 NT12) (0.000693766 NT08 NT04 NT15) (0.00112226 NT08 NT04 NT16) (0.00117025 NT08 NT06 NT02) (0.00807496 NT08 NT06 NT08) (0.0183971 NT08 NT06 NT09) (0.00127343 NT08 NT06 NT14) (0.0322725 NT08 NT06 NT17) (0.00396897 NT08 NT07 NT07) (0.0154729 NT08 NT07 NT09) (0.000708139 NT08 NT07 NT10) (0.00186499 NT08 NT07 NT11) (0.000701346 NT08 NT07 NT14) (0.0116278 NT08 NT08 NT09) (0.0965117 NT08 NT10 NT09) (0.000142086 NT08 NT10 NT12) (0.000210725 NT08 NT10 NT14) (0.00336223 NT08 NT11 NT07) (0.00183799 NT08 NT11 NT09) (0.00109249 NT08 NT11 NT11) (0.000880671 NT08 NT11 NT12) (0.0032493 NT08 NT12 NT08) (0.0372072 NT08 NT12 NT09) (0.00113127 NT08 NT12 NT12) (0.00892231 NT08 NT15 NT02) (0.00383754 NT08 NT15 NT06) (0.000528365 NT08 NT15 NT07) (0.0060705 NT08 NT15 NT08) (0.00853698 NT08 NT15 NT10) (0.0349777 NT08 NT15 NT14) (0.000202857 NT08 NT16 NT06) (0.00709689 NT08 NT16 NT07) (0.000240097 NT08 NT16 NT08) (0.0401819 NT08 NT16 NT09) (0.00124754 NT08 NT16 NT14) (0.00862498 NT08 n) (0.0115193 NT08 v) (0.000974267 NT08 in) (0.0169837 NT08 j) (0.00626434 NT08 r) (0.00437851 NT08 prp) (0.0062359 NT09 NT01 NT07) (0.000165196 NT09 NT01 NT14) (0.00151872 NT09 NT02 NT04) (0.000660061 NT09 NT02 NT15) (0.000434321 NT09 NT02 NT16) (0.00805872 NT09 NT03 NT09) (0.000180982 NT09 NT04 NT08) (0.050609 NT09 NT04 NT09) (0.000307442 NT09 NT04 NT15) (0.00281491 NT09 NT04 NT17) (0.000295911 NT09 NT06 NT15) (0.00133828 NT09 NT07 NT11) (0.0235741 NT09 NT12 NT09) (0.00121997 NT09 NT12 NT12) (0.00391762 NT09 NT15 NT01) (0.173027 NT09 NT15 NT02) (0.000462089 NT09 NT15 NT06) (0.0276663 NT09 NT15 NT07) (0.210483 NT09 NT15 NT08) (0.000177004 NT09 NT15 NT09) (0.243402 NT09 NT15 NT10) (0.0174403 NT09 NT15 NT11) (0.00646962 NT09 NT15 NT12) (0.155174 NT09 NT15 NT14) (0.00930502 NT09 NT15 NT17) (0.000311399 NT09 NT16 NT02) (0.0052031 NT09 NT16 NT07) (0.00742336 NT09 NT16 NT09) (0.000409254 NT09 in) (0.0019424 NT09 j) (0.0393282 NT09 r) (0.00016039 NT09 prp) (0.00268682 NT10 NT01 NT07) (0.00173594 NT10 NT01 NT09) (0.00550051 NT10 NT01 NT10) (0.00269002 NT10 NT01 NT11) (0.00881491 NT10 NT01 NT12) (0.0158503 NT10 NT02 NT02) (0.00229071 NT10 NT02 NT07) (0.00765082 NT10 NT02 NT09) (0.00102327 NT10 NT02 NT11) (0.474288 NT10 NT02 NT12) (0.0119086 NT10 NT02 NT14) (0.000270767 NT10 NT02 NT15) (0.00425023 NT10 NT02 NT16) (0.0533347 NT10 NT04 NT02) (0.00286524 NT10 NT04 NT06) (0.0687658 NT10 NT04 NT10) (0.0157381 NT10 NT04 NT12) (0.000809508 NT10 NT05 NT12) (0.00188343 NT10 NT06 NT04) (0.000155481 NT10 NT06 NT09) (0.00569591 NT10 NT06 NT14) (0.00233367 NT10 NT06 NT17) (0.000189475 NT10 NT07 NT05) (0.018548 NT10 NT07 NT07) (0.00472354 NT10 NT07 NT09) (0.0121145 NT10 NT07 NT11) (0.0698482 NT10 NT07 NT12) (0.000402661 NT10 NT07 NT16) (0.00183044 NT10 NT07 NT17) (0.00166519 NT10 NT10 NT02) (0.015445 NT10 NT10 NT09) (0.019208 NT10 NT10 NT12) (0.000942866 NT10 NT10 NT18) (0.00149941 NT10 NT11 NT01) (0.00624706 NT10 NT11 NT02) (0.0381755 NT10 NT11 NT11) (0.00754256 NT10 NT11 NT12) (0.00139213 NT10 NT15 NT02) (0.000523505 NT10 NT15 NT06) (0.0015256 NT10 NT15 NT10) (0.00119525 NT10 NT15 NT12) (0.00683524 NT10 NT16 NT02) (0.000398591 NT10 NT16 NT04) (0.0701558 NT10 NT16 NT07) (0.00198721 NT10 NT16 NT11) (0.0075364 NT10 NT16 NT12) (0.0186618 NT10 n) (0.000591828 NT10 uh) (0.157827 NT11 NT11 NT11) (0.0422576 NT11 NT15 NT11) (0.00247895 NT11 NT15 NT16) (0.000257833 NT11 dt) (0.754818 NT11 cd) (0.0421123 NT11 r) (0.00236916 NT12 NT01 NT07) (0.000118511 NT12 NT02 NT16) (0.00638739 NT12 NT04 NT02) (0.0055731 NT12 NT04 NT04) (0.0340903 NT12 NT04 NT12) (0.00102031 NT12 NT04 NT15) (0.00143793 NT12 NT04 NT16) (0.000102621 NT12 NT04 NT17) (0.0032774 NT12 NT06 NT04) (0.000366976 NT12 NT07 NT07) (0.00218153 NT12 NT07 NT11) (0.0117989 NT12 NT11 NT07) (0.00303601 NT12 NT12 NT12) (0.0747798 NT12 NT13 NT03) (0.000232806 NT12 NT15 NT01) (0.341016 NT12 NT15 NT02) (0.0190932 NT12 NT15 NT06) (0.100931 NT12 NT15 NT07) (0.193386 NT12 NT15 NT10) (0.0142796 NT12 NT15 NT11) (0.000915196 NT12 NT16 NT07) (0.000299768 NT12 NT16 NT11) (0.0135637 NT12 NT16 NT12) (0.115493 NT12 n) (0.00344871 NT12 v) (0.0262404 NT12 punc) (0.000493049 NT12 in) (0.00235382 NT12 j) (0.0192274 NT12 r) (0.00199831 NT12 prp) (0.000209376 NT13 NT11 NT15) (0.00188858 NT13 NT13 NT03) (0.540855 NT13 punc) (0.00804226 NT13 cc) (0.000413617 NT14 NT00 NT09) (0.0218326 NT14 NT00 NT14) (0.000451496 NT14 NT00 NT18) (0.00149459 NT14 NT01 NT07) (0.00384046 NT14 NT01 NT17) (0.00138254 NT14 NT02 NT09) (0.0525259 NT14 NT03 NT14) (0.000893974 NT14 NT04 NT02) (0.000175088 NT14 NT04 NT06) (0.000478859 NT14 NT04 NT08) (0.00086439 NT14 NT04 NT09) (0.00529624 NT14 NT04 NT10) (0.000476852 NT14 NT04 NT12) (0.00549502 NT14 NT04 NT14) (0.0281873 NT14 NT05 NT14) (0.76715 NT14 NT06 NT17) (0.00303311 NT14 NT07 NT07) (0.00027137 NT14 NT07 NT09) (0.000748841 NT14 NT07 NT12) (0.0874896 NT14 NT07 NT17) (0.00416962 NT14 NT09 NT14) (0.00175999 NT14 NT10 NT09) (0.000710869 NT14 NT11 NT17) (0.000723932 NT14 NT12 NT07) (0.00440147 NT14 NT12 NT14) (0.000761726 NT14 NT14 NT09) (0.00084762 NT14 NT14 NT17) (0.000323644 NT14 NT15 NT02) (0.00264492 NT14 NT15 NT14) (0.000238841 NT14 NT16 NT07) (0.000126025 NT14 NT16 NT09) (0.000217731 NT14 r) (0.00024161 NT14 wrb) (0.000366989 NT15 NT04 NT04) (0.00127143 NT15 NT04 NT15) (0.00137902 NT15 NT11 NT07) (0.000109067 NT15 NT15 NT04) (0.00380199 NT15 NT15 NT06) (0.000193842 NT15 NT15 NT15) (0.000253898 NT15 NT15 NT16) (0.00556123 NT15 v) (0.0798535 NT15 punc) (0.557206 NT15 in) (0.0519477 NT15 cc) (0.170466 NT15 of) (0.113587 NT15 to) (0.0125211 NT15 wrb) (0.00146961 NT15 pdt) (0.000682686 NT16 NT01 NT16) (0.000353409 NT16 NT02 NT02) (0.0034721 NT16 NT02 NT05) (0.00392739 NT16 NT04 NT04) (0.0225952 NT16 NT04 NT16) (0.00368407 NT16 NT05 NT16) (0.000275916 NT16 NT06 NT05) (0.0263102 NT16 NT07 NT05) (0.00344251 NT16 NT07 NT12) (0.00271063 NT16 NT07 NT16) (0.000950873 NT16 NT10 NT13) (0.0229124 NT16 NT11 NT07) (0.0173136 NT16 NT11 NT11) (0.0094147 NT16 NT11 NT16) (0.00210054 NT16 NT13 NT03) (0.000417271 NT16 NT15 NT01) (0.0100377 NT16 NT15 NT11) (0.000679194 NT16 NT16 NT05) (0.00203961 NT16 NT16 NT11) (0.00352444 NT16 NT16 NT12) (0.0133536 NT16 NT16 NT16) (0.0041124 NT16 n) (0.0518387 NT16 v) (0.0133556 NT16 punc) (0.746857 NT16 j) (0.0325454 NT16 cd) (0.000994964 NT16 r) (0.000325555 NT17 NT03 NT09) (0.000431668 NT17 NT03 NT17) (0.000283523 NT17 NT04 NT01) (0.00308221 NT17 NT04 NT02) (0.000106449 NT17 NT04 NT07) (0.584517 NT17 NT04 NT08) (0.0389749 NT17 NT04 NT09) (0.00927257 NT17 NT04 NT10) (0.000698039 NT17 NT04 NT11) (0.0594712 NT17 NT04 NT14) (0.000381951 NT17 NT04 NT16) (0.248255 NT17 NT04 NT17) (0.000264379 NT17 NT05 NT08) (0.00194384 NT17 NT05 NT10) (0.000308808 NT17 NT05 NT14) (0.000271388 NT17 NT07 NT08) (0.000131093 NT17 NT07 NT10) (0.00011195 NT17 NT07 NT17) (0.000462643 NT17 NT08 NT09) (0.00153331 NT17 NT11 NT07) (0.00214335 NT17 NT11 NT11) (0.000307068 NT17 NT11 NT12) (0.000550528 NT17 NT15 NT10) (0.000125644 NT17 NT16 NT02) (0.000474489 NT17 NT17 NT09) (0.00032483 NT17 NT17 NT18) (0.045027 NT17 v) (0.00425503 NT18 NT07 NT18) (0.978831 NT18 NT13 NT00) (0.00130119 NT18 NT13 NT03) (0.0155958 NT18 NT17 NT13) festival/lib/speech.properties0000644000175000017500000000013007262605650015134 0ustar awbawb# Register speech engines cstr.festival.EngineCentral=cstr.festival.jsapi.EngineCentral festival/lib/Singing.v0_1.dtd0000644000175000017500000000155007603100403014376 0ustar awbawb %ISOlat1; festival/lib/email-mode.scm0000644000175000017500000001014711363106376014273 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; An example tts text mode for reading email messages, this includes ;;; support for extracting the interesting headers from the message ;;; and for dealing with quoted text. Its all very primitive and ;;; will easily be confused but its here just as an example ;;; (define (email_init_func) "(email_init_func) Called on starting email text mode." (voice_rab_diphone) (set! email_previous_t2w_func token_to_words) (set! english_token_to_words email_token_to_words) (set! token_to_words english_token_to_words) (set! email_in_quote nil)) (define (email_exit_func) "(email_exit_func) Called on exit email text mode." (set! english_token_to_words email_previous_t2w_func) (set! token_to_words english_token_to_words)) (define (email_token_to_words token name) "(email_token_to_words utt token name) Email spcific token to word rules." (cond ((string-matches name "<.*@.*>") (append (email_previous_t2w_func token (string-after (string-before name "@") "<")) (cons "at" (email_previous_t2w_func token (string-before (string-after name "@") ">"))))) ((and (string-matches name ">") (string-matches (item.feat token "whitespace") "[ \t\n]*\n *")) (voice_cmu_us_awb_cg) nil ;; return nothing to say ) (t ;; for all other cases (if (string-matches (item.feat token "whitespace") ".*\n[ \n]*") (voice_rab_diphone)) (email_previous_t2w_func token name)))) (set! tts_text_modes (cons (list 'email ;; mode name (list ;; email mode params (list 'init_func email_init_func) (list 'exit_func email_exit_func) '(filter "email_filter"))) tts_text_modes)) (provide 'email-mode) festival/lib/ogimarkup-mode.scm0000644000175000017500000001746211363106376015211 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; An example tts text mode for reading OGI's CSLU toolkit mark up ;;; ;;; Note not all tokens do something in festival but all are removed ;;; from the actual text ;;; (defvar ogimarkup_eou_tree '((n.name matches "<.*") ((1)) ((n.whitespace matches ".*\n.*\n\\(.\\|\n\\)*") ;; A significant break (2 nls) ((1)) ((punc in ("?" ":" "!")) ((1)) ((punc is ".") ;; This is to distinguish abbreviations vs periods ;; These are heuristics ((name matches "\\(.*\\..*\\|[A-Z][A-Za-z]?[A-Za-z]?\\|etc\\)") ;; an abbreviation ((n.whitespace is " ") ((0)) ;; if abbrev single space isn't enough for break ((n.name matches "[A-Z].*") ((1)) ((0)))) ((n.whitespace is " ") ;; if it doesn't look like an abbreviation ((n.name matches "[A-Z].*") ;; single space and non-cap is no break ((1)) ((0))) ((1)))) ((0))))))) (define (ogimarkup_init_func) "Called on starting ogimarkup text mode." (set! ogimarkup_in_tag nil) (set! ogimarkup_tagtokens "") (set! ogimarkup_previous_t2w_func token_to_words) (set! english_token_to_words ogimarkup_token_to_words) (set! token_to_words ogimarkup_token_to_words) (set! ogimarkup_previous_eou_tree eou_tree) (set! eou_tree ogimarkup_eou_tree)) (define (ogimarkup_exit_func) "Called on exit ogimarkup text mode." (Parameter.set 'Duration_Stretch 1.0) (set! token_to_words ogimarkup_previous_t2w_func) (set! english_token_to_words ogimarkup_previous_t2w_func) (set! eou_tree ogimarkup_previous_eou_tree)) (define (ogimarkup_token_to_words token name) "(ogimarkup_token_to_words token name) OGI markup specific token to word rules. Tags may have optional argument e.g. or which means the tag may be over a number of tokens." (let (tag (arg nil) (rval nil)) (cond ((string-matches name "<.*") (set! ogimarkup_tagtokens "") (set! tag (string-after name "<")) (if (string-matches tag ".*>$") (set! tag (string-before tag ">")) (if (string-matches (set! arg (item.feat token "n.name")) ".*>$") (set! arg (string-before arg ">")))) (set! ogimarkup_in_tag tag) (cond ((string-equal tag "slow") (Parameter.set 'Duration_Stretch 1.3)) ((string-equal tag "SLOW") (Parameter.set 'Duration_Stretch 2.0)) ((string-equal tag "normal") (Parameter.set 'Duration_Stretch 1.0)) ((string-matches tag "FAST") (Parameter.set 'Duration_Stretch 0.5)) ((string-matches tag "fast") (Parameter.set 'Duration_Stretch 0.8)) ((string-matches tag"spell") ;; This ain't really right as we'll get an utterance break here (set! rval (symbolexplode arg))) ((string-matches tag "phone") ;; This ain't really right as we'll get an utterance break here (item.set_feat token "token_pos" "digits") ;; canonical phone number (set! rval (ogimarkup_previous_t2w_func token arg))) ((string-matches tag "male") (if (and (member 'OGIresLPC *modules*) (symbol-bound? 'voice_aec_diphone)) (voice_aec_diphone) (voice_kal_diphone))) ((string-matches tag "Male") (if (and (member 'OGIresLPC *modules*) (symbol-bound? 'voice_mwm_diphone)) (voice_mwm_diphone) (voice_cmu_us_rms_cg))) ((string-matches tag "MALE") (if (and (member 'OGIresLPC *modules*) (symbol-bound? 'voice_jph_diphone)) (voice_jph_diphone) (voice_rab_diphone))) ((string-matches tag "FT") t) ;; do nothing until the end of this tag ((string-matches (downcase tag) "female") ;; only one female voice so map female Female FEMALE to it (if (and (member 'OGIresLPC *modules*) (symbol-bound? 'voice_tll_diphone)) (voice_tll_diphone) (voice_cmu_us_slt_arctic_hts)))) (if (string-matches name ".*>$") (set! ogimarkup_in_tag nil)) rval ;; mostly nil ) ((string-matches name ".*>$") (set! ogimarkup_tagtokens (string-append ogimarkup_tagtokens (ogimarkup_get_token_string token t))) ;; delete final > (if (string-equal ogimarkup_in_tag "FT") (ogimarkup_festival_eval ogimarkup_tagtokens)) (set! ogimarkup_in_tag nil) ;; end of tag nil) (ogimarkup_in_tag (set! ogimarkup_tagtokens (string-append ogimarkup_tagtokens (ogimarkup_get_token_string token nil))) nil) ;; still in tag (t ;; for all other cases (ogimarkup_previous_t2w_func token name))))) (set! tts_text_modes (cons (list 'ogimarkup ;; mode name (list ;; ogimarkup mode params (list 'init_func ogimarkup_init_func) (list 'exit_func ogimarkup_exit_func))) tts_text_modes)) (define (ogimarkup_get_token_string token delend) "(ogimarkup_get_token_string TOKEN DELEND) return string for token including whitespace and punctuation. If DELEND is true remove > from the name." (string-append (item.feat token "whitespace") (item.feat token "prepunctuation") (if delend (string-before (item.feat token "name") ">") (item.feat token "name")) (if (string-equal "0" (item.feat token "punc")) "" (item.feat token "punc")))) (define (ogimarkup_festival_eval tagtokens) "(ogimarkup_festival_eval TAGTOKENS Take a string of the tokens within the tag and read an s-expression from it and then evaluate it." (let ((com "") (command nil)) (set! command (read-from-string tagtokens)) (eval command))) (provide 'ogimarkup-mode) festival/lib/sable-mode.scm0000644000175000017500000005006012437426122014265 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1998 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Festival (1.3.X) support for SABLE 0.2 the SGML/XML based mark up ;; ;;; language. ;; ;;; ;; ;;; This is XML version requiring Edinburgh's LTG's rxp XML parser as ;; ;;; distributed with Festival ;; ;;; ;; (require_module 'rxp) ;;(set! auto-text-mode-alist ;; (cons ;; (cons "\\.sable$" 'sable) ;; auto-text-mode-alist)) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; Remember where to find these two XML entities. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (xml_register_id "-//SABLE//DTD SABLE speech mark up//EN" (path-append libdir "Sable.v0_2.dtd") ) (xml_register_id "-//SABLE//ENTITIES Added Latin 1 for SABLE//EN" (path-append libdir "sable-latin.ent") ) ;; (print (xml_registered_ids)) (defvar SABLE_RXDOUBLE "-?\\(\\([0-9]+\\.[0-9]*\\)\\|\\([0-9]+\\)\\|\\(\\.[0-9]+\\)\\)\\([eE][---+]?[0-9]+\\)?") (defvar sable_pitch_base_map '((highest 1.2) (high 1.1) (medium 1.0) (default 1.0) (low 0.9) (lowest 0.8))) (defvar sable_pitch_med_map '((highest 1.2) (high 1.1) (medium 1.0) (default 1.0) (low 0.9) (lowest 0.8))) (defvar sable_pitch_range_map '((largest 1.2) (large 1.1) (medium 1.0) (default 1.0) (small 0.9) (smallest 0.8))) (defvar sable_rate_speed_map '((fastest 1.5) (fast 1.2) (medium 1.0) (default 1.0) (slow 0.8) (slowest 0.6))) (defvar sable_volume_level_map '((loudest 2.0) (loud 1.5) (default 1.0) (medium 1.0) (quiet 0.5))) (define (sable_init_globals) (set! utts nil) (set! sable_omitted_mode nil) (set! sable_word_features_stack nil) (set! sable_pitch_context nil) (set! sable_vol_context nil) (set! sable_vol_type 'no_change) (set! sable_vol_factor 1.0) (set! sable_current_language 'britishenglish) (set! sable_unsupported_language nil) (set! sable_language_stack nil) (set! sable_current_speaker 'voice_kal_diphone) (set! sable_speaker_stack nil) ) (define (sable_token_to_words token name) "(sable_token_to_words utt token name) SABLE mode token specific analysis." (cond ((or sable_omitted_mode sable_unsupported_language) ;; don't say anything (whole utterance) nil) ((string-equal "1" (item.feat token "done_sable_sub")) ;; to catch recursive calls this when splitting up sub expressions (sable_previous_token_to_words token name)) ((and (not (string-equal "0" (item.feat token "sable_sub"))) (string-equal "0" (item.feat token "p.sable_sub"))) (let (words (sub (item.feat token "sable_sub"))) (item.set_feat token "done_sable_sub" "1") (set! words (apply append (mapcar (lambda (w) (set! www (sable_previous_token_to_words token w)) www) (read-from-string sub)))) (item.set_feat token "done_sable_sub" "0") words)) ((string-equal "1" (item.feat token "sable_ignore")) ;; don't say anything (individual word) nil) ((string-equal "1" (item.feat token "sable_ipa")) ;; Each token is an IPA phone (item.set_feat token "phonemes" (sable-map-ipa name)) (list name)) ((string-equal "1" (item.feat token "sable_literal")) ;; Only deal with spell here (let ((subwords) (subword)) (item.set_feat token "pos" token.letter_pos) (mapcar (lambda (letter) ;; might be symbols or digits (set! subword (sable_previous_token_to_words token letter)) (if subwords (set! subwords (append subwords subword)) (set! subwords subword))) (symbolexplode name)) subwords)) ((not (string-equal "0" (item.feat token "token_pos"))) ;; bypass the prediction stage, if English (if (member_string (Parameter.get 'Language) '(britishenglish americanenglish)) (builtin_english_token_to_words token name) (sable_previous_token_to_words token name))) ;; could be others here later (t (sable_previous_token_to_words token name)))) (defvar sable_elements '( ("(SABLE" (ATTLIST UTT) (eval (list sable_current_speaker)) ;; so we know what state we start in (sable_setup_voice_params) nil ) (")SABLE" (ATTLIST UTT) (xxml_synth UTT) ;; Synthesis the remaining tokens nil ) ;; Utterance break elements ("(LANGUAGE" (ATTLIST UTT) ;; Status: probably complete (xxml_synth UTT) (set! sable_language_stack (cons (list sable_current_language sable_unsupported_language) sable_language_stack)) ;; Select a new language (let ((language (upcase (car (xxml_attval "ID" ATTLIST))))) (cond ((or (string-equal language "SPANISH") (string-equal language "ES")) (set! sable_current_language 'spanish) (set! sable_unsupported_language nil) (select_language 'spanish)) ((or (string-equal language "ENGLISH") (string-equal language "EN")) (set! sable_current_language 'britishenglish) (set! sable_unsupported_language nil) (select_language 'britishenglish)) (t ;; skip languages you don't know ;; BUG: if current language isn't English this wont work (apply_hooks tts_hooks (eval (list 'Utterance 'Text (string-append "Some text in " language)))) (set! sable_unsupported_language t))) nil)) (")LANGUAGE" (ATTLIST UTT) (xxml_synth UTT) (set! sable_unsupported_language (car (cdr (car sable_language_stack)))) (set! sable_current_language (car (car sable_language_stack))) (set! sable_language_stack (cdr sable_language_stack)) (if (not sable_omitted_mode) (begin (select_language sable_current_language) (sable_setup_voice_params))) nil) ("(SPEAKER" (ATTLIST UTT) ;; Status: GENDER/AGE ignored, should be done by sable-def-speaker ;; function to define Festival voices to SABLE (xxml_synth UTT) (set! sable_speaker_stack (cons sable_current_speaker sable_speaker_stack)) (cond ((not equal? sable_current_language 'britishenglish) (print "SABLE: choosen unknown voice, current voice unchanged")) ((equal? (car (xxml_attval "NAME" ATTLIST)) 'male1) (set! sable_current_speaker 'voice_kal_diphone) (voice_kal_diphone)) ((equal? (car (xxml_attval "NAME" ATTLIST)) 'male2) (set! sable_current_speaker 'voice_cmu_us_rms_cg) (voice_cmu_us_rms_cg)) ((equal? (car (xxml_attval "NAME" ATTLIST)) 'male3) (set! sable_current_speaker 'voice_ked_diphone) (voice_ked_diphone)) ((equal? (car (xxml_attval "NAME" ATTLIST)) 'male4) (set! sable_current_speaker 'voice_rab_diphone) (voice_rab_diphone)) ((equal? (car (xxml_attval "NAME" ATTLIST)) 'male5) (set! sable_current_speaker 'voice_cmu_us_awb_cg) (voice_cmu_us_awb_cg)) ((equal? (car (xxml_attval "NAME" ATTLIST)) 'female1) (set! sable_current_speaker 'voice_cmu_us_slt_cg) (voice_us1_mbrola)) (t (set! sable_current_speaker (intern (string-append "voice_" (car (xxml_attval "NAME" ATTLIST))))) (eval (list sable_current_speaker)))) (sable_setup_voice_params) nil) (")SPEAKER" (ATTLIST UTT) (xxml_synth UTT) (set! sable_utt UTT) (set! sable_current_speaker (car sable_speaker_stack)) (set! sable_speaker_stack (cdr sable_speaker_stack)) (eval (list sable_current_speaker)) (sable_setup_voice_params) nil) ("BREAK" (ATTLIST UTT) ;; Status: probably complete ;; may cause an utterance break (let ((level (upcase (car (xxml_attval "LEVEL" ATTLIST))))) (cond ((null UTT) nil) ((string-equal "LARGE" level) (xxml_synth UTT) nil) (t (let ((last_token (utt.relation.last UTT'Token))) (if last_token (item.set_feat last_token "pbreak" "B")) UTT))))) ("(DIV" (ATLIST UTT) ;; Status: probably complete (xxml_synth UTT) nil) ("AUDIO" (ATTLIST UTT) ;; Status: MODE (background) ignored, only insertion supported ;; mime type of file also ignored, as its LEVEL (let ((tmpfile (make_tmp_filename))) ;; ignoring mode-background (and will for sometime) ;; ignoring level option (xxml_synth UTT) ;; synthesizing anything ready to be synthesized (get_url (car (xxml_attval "SRC" ATTLIST)) tmpfile) (apply_hooks tts_hooks (eval (list 'Utterance 'Wave tmpfile))) (delete-file tmpfile) nil)) ("(EMPH" (ATTLIST UTT) ;; Status: nesting makes no difference, levels ignored ;; Festival is particularly bad at adding specific emphasis ;; that's what happens when you use statistical methods that ;; don't include any notion of emphasis ;; This is *not* recursive and only one level of EMPH supported (sable_push_word_features) (set! xxml_word_features (cons (list "dur_stretch" 1.6) (cons (list "EMPH" "1") xxml_word_features))) UTT) (")EMPH" (ATTLIST UTT) (set! xxml_word_features (sable_pop_word_features)) UTT) ("(PITCH" (ATTLIST UTT) ;; Status: probably complete ;; At present festival requires an utterance break here (xxml_synth UTT) (set! sable_pitch_context (cons int_lr_params sable_pitch_context)) (let ((base (sable_interpret_param (car (xxml_attval "BASE" ATTLIST)) sable_pitch_base_map (cadr (assoc 'target_f0_mean int_lr_params)) sable_pitch_base_original)) (med (sable_interpret_param (car (xxml_attval "MED" ATTLIST)) sable_pitch_med_map (cadr (assoc 'target_f0_mean int_lr_params)) sable_pitch_med_original)) (range (sable_interpret_param (car (xxml_attval "RANGE" ATTLIST)) sable_pitch_range_map (cadr (assoc 'target_f0_std int_lr_params)) sable_pitch_range_original)) (oldmean (cadr (assoc 'target_f0_mean int_lr_params)))) ;; Festival (if it supports anything) supports mean and std ;; so we treat base as med if med doesn't seem to do anything (if (equal? med oldmean) (set! med base)) (set! int_lr_params (cons (list 'target_f0_mean med) (cons (list 'target_f0_std range) int_lr_params))) nil)) (")PITCH" (ATTLIST UTT) (xxml_synth UTT) (set! int_lr_params (car sable_pitch_context)) (set! sable_pitch_context (cdr sable_pitch_context)) nil) ("(RATE" (ATTLIST UTT) ;; Status: can't deal with absolute word per minute SPEED. (sable_push_word_features) ;; can't deal with words per minute value (let ((rate (sable_interpret_param (car (xxml_attval "SPEED" ATTLIST)) sable_rate_speed_map (sable_find_fval "dur_stretch" xxml_word_features 1.0) sable_rate_speed_original))) (set! xxml_word_features (cons (list "dur_stretch" (/ 1.0 rate)) xxml_word_features)) UTT)) (")RATE" (ATTLIST UTT) (set! xxml_word_features (sable_pop_word_features)) UTT) ("(VOLUME" (ATTLIST UTT) ;; Status: probably complete ;; At present festival requires an utterance break here (xxml_synth UTT) (set! sable_vol_context (cons (list sable_vol_type sable_vol_factor) sable_vol_context)) (let ((level (sable_interpret_param (car (xxml_attval "LEVEL" ATTLIST)) sable_volume_level_map sable_vol_factor 1.0))) (cond ((string-matches (car (xxml_attval "LEVEL" ATTLIST)) ".*%") (set! sable_vol_type 'relative)) ((string-matches (car (xxml_attval "LEVEL" ATTLIST)) SABLE_RXDOUBLE) (set! sable_vol_type 'absolute)) (t (set! sable_vol_type 'relative))) (set! sable_vol_factor level)) nil) (")VOLUME" (ATTLIST UTT) (xxml_synth UTT) (set! sable_vol_type (car (car sable_vol_context))) (set! sable_vol_factor (car (cdr (car sable_vol_context)))) (set! sable_vol_context (cdr sable_vol_context)) nil) ("(ENGINE" (ATTLIST UTT) ;; Status: probably complete (xxml_synth UTT) (if (string-matches (car (xxml_attval "ID" ATTLIST)) "festival.*") (let ((datastr "")) (mapcar (lambda (c) (set! datastr (string-append datastr " " c))) (xxml_attval "DATA" ATTLIST)) (apply_hooks tts_hooks (eval (list 'Utterance 'Text datastr))) (set! sable_omitted_mode t)) ;; ignore contents ;; else ;; its not relevant to me ) nil) (")ENGINE" (ATTLIST UTT) (xxml_synth UTT) (set! sable_omitted_mode nil) nil) ("MARKER" (ATTLIST UTT) ;; Status: does nothing ;; Can't support this without low-level control of audio spooler (format t "SABLE: marker \"%s\"\n" (car (xxml_attval "MARK" ATTLIST))) UTT) ("(PRON" (ATTLIST UTT) ;; Status: IPA currently ignored (sable_push_word_features) ;; can't deal with words per minute value (let ((ipa (xxml_attval "IPA" ATTLIST)) (sub (xxml_attval "SUB" ATTLIST))) (cond (ipa (format t "SABLE: ipa ignored\n") (set! xxml_word_features (cons (list "sable_ignore" "1") xxml_word_features))) (sub (set! xxml_word_features (cons (list "sable_sub" (format nil "%l" sub)) xxml_word_features)) (set! xxml_word_features (cons (list "sable_ignore" "1") xxml_word_features)))) UTT)) (")PRON" (ATTLIST UTT) (set! xxml_word_features (sable_pop_word_features)) UTT) ("(SAYAS" (ATTLIST UTT) ;; Status: only a few of the types are dealt with (sable_push_word_features) (set! sable_utt UTT) ;; can't deal with words per minute value (let ((mode (downcase (car (xxml_attval "MODE" ATTLIST)))) (modetype (car (xxml_attval "MODETYPE" ATTLIST)))) (cond ((string-equal mode "literal") (set! xxml_word_features (cons (list "sable_literal" "1") xxml_word_features))) ((string-equal mode "phone") (set! xxml_word_features (cons (list "token_pos" "digits") xxml_word_features))) ((string-equal mode "ordinal") (set! xxml_word_features (cons (list "token_pos" "ordinal") xxml_word_features))) ((string-equal mode "cardinal") (set! xxml_word_features (cons (list "token_pos" "cardinal") xxml_word_features))) (t ;; blindly trust festival to get it right t)) UTT)) (")SAYAS" (ATTLIST UTT) (set! xxml_word_features (sable_pop_word_features)) UTT) )) (define (sable_init_func) "(sable_init_func) Initialisation for SABLE mode" (sable_init_globals) (voice_kal_diphone) (set! sable_previous_elements xxml_elements) (set! xxml_elements sable_elements) (set! sable_previous_token_to_words english_token_to_words) (set! english_token_to_words sable_token_to_words) (set! token_to_words sable_token_to_words)) (define (sable_exit_func) "(sable_exit_func) Exit function for SABLE mode" (set! xxml_elements sable_previous_elements) (set! token_to_words sable_previous_token_to_words) (set! english_token_to_words sable_previous_token_to_words)) (define (sable_push_word_features) "(sable_push_word_features) Save current word features on stack." (set! sable_word_features_stack (cons xxml_word_features sable_word_features_stack))) (define (sable_adjust_volume utt) "(sable_adjust_volume utt) Amplify or attenutate signale based on value of sable_vol_factor and sable_vol_type (absolute or relative)." (set! utts (cons utt utts)) (cond ((equal? sable_vol_type 'no_change) utt) ((equal? sable_vol_type 'absolute) (utt.wave.rescale utt sable_vol_factor 'absolute)) ((equal? sable_vol_type 'relative) (utt.wave.rescale utt sable_vol_factor)) (t (format stderr "SABLE: volume unknown type \"%s\"\n" sable_vol_type) utt)) utt) (define (sable_pop_word_features) "(sable_pop_word_features) Pop word features from stack." (let ((r (car sable_word_features_stack))) (set! sable_word_features_stack (cdr sable_word_features_stack)) r)) (define (sable_find_fval feat flist def) (cond ((null flist) def) ((string-equal feat (car (car flist))) (car (cdr (car flist)))) (t (sable_find_fval feat (cdr flist) def)))) (define (sable_interpret_param ident map original current) "(sable_interpret_param IDENT MAP ORIGINAL CURRENT) If IDENT is in map return ORIGINAL times value in map, otherwise treat IDENT of the form +/-N% and modify CURRENT accordingly." (let ((mm (assoc ident map))) (cond (mm (* original (car (cdr mm)))) ((string-matches ident SABLE_RXDOUBLE) (parse-number ident)) ((string-matches ident ".*%") (+ current (* current (/ (parse-number (string-before ident "%")) 100.0)))) ;; ((string-matches ident ".*%") ;; (* current (/ (parse-number (string-before ident "%")) 100.0))) ((not ident) current) (t (format stderr "SABLE: modifier \"%s\" not of float, tag or +/-N\n" ident) current)))) (define (sable_setup_voice_params) "(sable_setup_voice_params) Set up original values for various voice parameters." (set! sable_pitch_base_original (cadr (assoc 'target_f0_mean int_lr_params))) (set! sable_pitch_med_original (cadr (assoc 'target_f0_mean int_lr_params))) (set! sable_pitch_range_original (cadr (assoc 'target_f0_std int_lr_params))) (set! sable_rate_speed_original 1.0) (if (and after_synth_hooks (not (consp after_synth_hooks))) (set! after_synth_hooks (cons after_synth_hooks (list sable_adjust_volume))) (set! after_synth_hooks (append after_synth_hooks (list sable_adjust_volume)))) ) ;;; Declare the new mode to Festival (set! tts_text_modes (cons (list 'sable ;; mode name (list (list 'init_func sable_init_func) (list 'exit_func sable_exit_func) '(analysis_type xml) )) tts_text_modes)) (provide 'sable-mode) festival/lib/soleml-mode.scm0000644000175000017500000003027707262605650014507 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1998 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Support for an SGML based mark-up language used in the SOLE ;;; project. This is all still experimental. ;;; ;;; This currently treats one file as one utterance (to make dealing with ;;; the SOLE museaum database easy (set! soleml_word_features_stack nil) (defvar sole_current_node nil) (define (soleml_token_to_words utt token name) "(soleml_token_to_words utt token name) SOLEML mode token specific analysis." (cond (t (soleml_previous_token_to_words utt token name)))) (define (voice_soleml) "(soleml_voice) Speaker specific initialisation for SOLE museum data." (voice_rab_diphone) ;; Utterances only come at end of file (set! eou_tree '((0))) ) (defvar soleml_elements '( ("(SOLEML" (ATTLIST UTT) ;; required to identify type (voice_soleml) ;; so we know what state we start in (set! soleml_utt (Utterance Tokens nil)) (utt.stream.create soleml_utt 'Token) (utt.relation.create soleml_utt 'SOLEML) (set! sole_current_node (utt.relation_append soleml_utt 'SOLEML (cons "sole-ml" ATTLIST))) soleml_utt ) (")SOLEML" (ATTLIST UTT) ;; required to identify end token ;; Don't really want to synthesize this ;; (xxml_synth UTT) ;; Synthesis the remaining tokens (set! soleml_utt UTT) UTT ) ;; Utterance break elements ("(LANGUAGE" (ATTLIST UTT) ;; Select a new language (select_language (car (xxml_attval "NAME" ATTLIST))) UTT) ("(VOICE" (ATTLIST UTT) ;;(xxml_synth UTT) ;; Select a new voice (cond ((equal? (car (xxml_attval "NAME" ATTLIST)) 'male1) (voice_soleml_diphone)) ((equal? (car (xxml_attval "NAME" ATTLIST)) 'male2) (voice_soleml_diphone)) ((equal? (car (xxml_attval "NAME" ATTLIST)) 'male3) (voice_soleml_diphone)) (t (print "SOLEML: selecting unknown voice") (voice_soleml_diphone))) UTT) ;; phrase-boundary // mark on token (??) ;; punct-elem // mark on token ;; sem-elem ;; text-elem // ignore ;; rhet-elem has nucleus and satellite ;; anaphora-elem ;; syn-elem ;; info-struct-elem ;; other-elem ("(PUNCT-ELEM" (ATTLIST UTT) (soleml_push_word_features) (set! xxml_word_features (cons (list "punct-elem" "1") (soleml_conv_attlist ATTLIST))) UTT) (")PUNCT-ELEM" (ATTLIST UTT) (set! xxml_word_features (soleml_pop_word_features)) UTT) ("(PHRASE-BOUNDARY" (ATTLIST UTT) (if (string-equal "4" (car (xxml_attval "STRENGTH" ATTLIST))) (begin ;; (xxml_synth UTT) UTT) (let ((last_token (car (last (utt.stream UTT 'Token))))) (if last_token (item.set_feat last_token "pbreak" "B")) UTT))) ;; For each recursive element simply build a new node ("(RHET-ELEM" (ATTLIST UTT) (let ((sdesc (list 'rhet-elem (soleml_conv_attlist ATTLIST)))) (set! sole_current_node (node.append_daughter sole_current_node sdesc)) UTT)) (")RHET-ELEM" (ATTLIST UTT) (set! sole_current_node (node.parent sole_current_node)) UTT) ("(RHET-EMPH" (ATTLIST UTT) (let ((sdesc (list 'rhet-emph (soleml_conv_attlist ATTLIST)))) (set! sole_current_node (node.append_daughter sole_current_node sdesc)) UTT)) (")RHET-EMPH" (ATTLIST UTT) (set! sole_current_node (node.parent sole_current_node)) UTT) ("(ANAPHORA-ELEM" (ATTLIST UTT) (let ((sdesc (list 'anaphora-elem (soleml_conv_attlist ATTLIST)))) (set! sole_current_node (node.append_daughter sole_current_node sdesc)) UTT)) (")ANAPHORA-ELEM" (ATTLIST UTT) (set! sole_current_node (node.parent sole_current_node)) UTT) ("(SYN-ELEM" (ATTLIST UTT) (let ((sdesc (list 'syn-elem (soleml_conv_attlist ATTLIST)))) (set! sole_current_node (node.append_daughter sole_current_node sdesc)) UTT)) (")SYN-ELEM" (ATTLIST UTT) (set! sole_current_node (node.parent sole_current_node)) UTT) ("(CONNECTIVE" (ATTLIST UTT) (let ((sdesc (list 'connective (soleml_conv_attlist ATTLIST)))) (set! sole_current_node (node.append_daughter sole_current_node sdesc)) UTT)) (")CONNECTIVE" (ATTLIST UTT) (set! sole_current_node (node.parent sole_current_node)) UTT) ("(TEXT-ELEM" (ATTLIST UTT) (let ((sdesc (list 'text-elem (soleml_conv_attlist ATTLIST)))) (set! sole_current_node (node.append_daughter sole_current_node sdesc)) UTT)) (")TEXT-ELEM" (ATTLIST UTT) (set! sole_current_node (node.parent sole_current_node)) UTT) ("(SEM-ELEM" (ATTLIST UTT) (let ((sdesc (list 'sem-elem (soleml_conv_attlist ATTLIST)))) (set! sole_current_node (node.append_daughter sole_current_node sdesc)) UTT)) (")SEM-ELEM" (ATTLIST UTT) (set! sole_current_node (node.parent sole_current_node)) UTT) ("(INFO-STRUCT-ELEM" (ATTLIST UTT) (let ((sdesc (list 'info-struct-elem (soleml_conv_attlist ATTLIST)))) (set! sole_current_node (node.append_daughter sole_current_node sdesc)) UTT)) (")INFO-STRUCT-ELEM" (ATTLIST UTT) (set! sole_current_node (node.parent sole_current_node)) UTT) ("(OTHER-ELEM" (ATTLIST UTT) (let ((sdesc (list 'other-elem (soleml_conv_attlist ATTLIST)))) (set! sole_current_node (node.append_daughter sole_current_node sdesc)) UTT)) (")OTHER-ELEM" (ATTLIST UTT) (set! sole_current_node (node.parent sole_current_node)) UTT) ("(NUCLEUS" (ATTLIST UTT) (let ((sdesc (list 'nucleus (soleml_conv_attlist ATTLIST)))) (set! sole_current_node (node.append_daughter sole_current_node sdesc)) UTT)) (")NUCLEUS" (ATTLIST UTT) (set! sole_current_node (node.parent sole_current_node)) UTT) ("(SATELLITE" (ATTLIST UTT) (let ((sdesc (list 'satellite (soleml_conv_attlist ATTLIST)))) (set! sole_current_node (node.append_daughter sole_current_node sdesc)) UTT)) (")SATELLITE" (ATTLIST UTT) (set! sole_current_node (node.parent sole_current_node)) UTT) ;; Other control functions (probably not used in SOLE) ("(CALL" (ATTLIST UTT) ;; (xxml_synth UTT) (if (string-matches (car (xxml_attval "ENGID" ATTLIST)) "festival.*") (let ((comstr "")) (mapcar (lambda (c) (set! comstr (string-append comstr " " c))) (xxml_attval "COMMAND" ATTLIST)) (eval (read-from-string comstr)))) UTT) ("(DEFINE" (ATTLIST UTT) ;; (xxml_synth UTT) (if (not (string-equal "NATIVE" (car (xxml_attval "SCHEME" ATTLIST)))) (format t "DEFINE: unsupported SCHEME %s, definition ignored\n" (car (xxml_attval "SCHEME" ATTLIST))) (lex.add.entry (list (car (xxml_attval "WORDS" ATTLIST)) ;; head form nil ;; pos (lex.syllabify.phstress (xxml_attval "PRONS" ATTLIST))))) UTT) ("(SOUND" (ATTLIST UTT) ;; (xxml_synth UTT) (if (not soleml_omitted_mode) (apply_hooks tts_hooks (eval (list 'Utterance 'Wave (car (xxml_attval "SRC" ATTLIST)))))) UTT) ("(EMPH" (ATTLIST UTT) ;; Festival is particularly bad at adding specific emphasis ;; that's what happens when you use statistical methods that ;; don't include any notion of emphasis ;; This is *not* recursive (soleml_push_word_features) (set! xxml_word_features (cons (list "EMPH" "1") xxml_word_features)) UTT) (")EMPH" (ATTLIST UTT) (set! xxml_word_features (soleml_pop_word_features)) UTT) ("(WORD" (ATTLIST UTT) ;; a word in-line (let ((name (xxml_attval "NAME" ATTLIST)) (pos (xxml_attval "POS" ATTLIST)) (accent (xxml_attval "ACCENT" ATTLIST)) (tone (xxml_attval "TONE" ATTLIST)) (phonemes (xxml_attval "PHONEMES" ATTLIST)) token) (utt.item.insert UTT 'Token) ;; add new Token (set! token (utt.stream.tail UTT 'Token)) (item.set_name token (car name)) (if pos (item.set_feat token "pos" (car pos))) (if accent (item.set_feat token "accent" (car accent))) (if tone (item.set_feat token "tone" (car tone))) (if phonemes (item.set_feat token "phonemes" (format nil "%l" phonemes))) UTT)) )) (define (soleml_init_func) "(soleml_init_func) Initialisation for SOLEML mode" (voice_soleml) (set! soleml_previous_elements xxml_elements) (set! xxml_elements soleml_elements) (set! xxml_token_hooks soleml_token_function) (set! soleml_previous_token_to_words english_token_to_words) (set! english_token_to_words soleml_token_to_words) (set! token_to_words soleml_token_to_words)) (define (soleml_exit_func) "(soleml_exit_func) Exit function for SOLEML mode" (set! xxml_elements soleml_previous_elements) (set! token_to_words soleml_previous_token_to_words) (set! english_token_to_words soleml_previous_token_to_words)) (define (soleml_token_function si) "(soleml_token_function si) This is called for each token found." (node.append_daughter sole_current_node si)) (define (soleml_push_word_features) "(soleml_push_word_features) Save current word features on stack." (set! soleml_word_features_stack (cons xxml_word_features soleml_word_features_stack))) (define (soleml_pop_word_features) "(soleml_pop_word_features) Pop word features from stack." (let ((r (car soleml_word_features_stack))) (set! soleml_word_features_stack (cdr soleml_word_features_stack)) r)) (define (soleml_conv_attlist alist) "(soleml_conv_attlist alist) Flatten alist arguments." (cond ((null alist) nil) ((null (car (cdr (car alist)))) (soleml_conv_attlist (cdr alist))) ((equal? (length (car (cdr (car alist)))) 1) (cons (list (car (car alist)) (car (car (cdr (car alist))))) (soleml_conv_attlist (cdr alist)))) (t (cons (list (car (car alist)) (format nil "%l" (car (cdr (car alist))))) (soleml_conv_attlist (cdr alist)))))) (set! tts_text_modes (cons (list 'soleml ;; mode name (list ;; email mode params (list 'init_func soleml_init_func) (list 'exit_func soleml_exit_func) '(analysis_type xxml) (list 'filter (format nil "%s -D %s " sgml_parse_progname libdir)))) tts_text_modes)) (provide 'soleml-mode) festival/lib/singing-mode.scm0000644000175000017500000005605012320321444014632 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Festival Singing Mode ;;; ;;; Written by Dominic Mazzoni ;;; Carnegie Mellon University ;;; 11-752 - "Speech: Phonetics, Prosody, Perception and Synthesis" ;;; Spring 2001 ;;; ;;; Extended by Milan Zamazal , 2006: ;;; - Slur support. ;;; - Czech support. ;;; - Some cleanup. ;;; - Print debugging information only when singing-debug is true. ;;; ;;; This code is public domain; anyone may use it freely. ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (require_module 'rxp) (xml_register_id "-//SINGING//DTD SINGING mark up//EN" (path-append xml_dtd_dir "Singing.v0_1.dtd") ) (xml_register_id "-//SINGING//ENTITIES Added Latin 1 for SINGING//EN" (path-append xml_dtd_dir "sable-latin.ent") ) ;; Set this to t to enable debugging messages: (defvar singing-debug nil) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; XML parsing functions ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; singing_xml_targets ;; ;; This variable defines the actions that are to be taken when ;; parsing each of our XML tags: SINGING, PITCH, DURATION, and REST. ;; ;; When we get the pitch and duration of each token, we store them ;; in features of the token. Later our intonation and duration ;; functions access these features. ;; (defvar singing_xml_elements '( ("(SINGING" (ATTLIST UTT) (set! singing_pitch_att_list nil) (set! singing_dur_att_list nil) (set! singing_global_time 0.0) (set! singing_bpm (get-bpm ATTLIST)) (set! singing_bps (/ singing_bpm 50.0)) ;; change this back to 60 nil) (")SINGING" (ATTLIST UTT) (xxml_synth UTT) ;; Synthesize the remaining tokens nil) ("(PITCH" (ATTLIST UTT) (set! singing_pitch_att_list ATTLIST) UTT) (")PITCH" (ATTLIST UTT) (let ((freq (get-freqs singing_pitch_att_list))) (if singing-debug (begin (print "freqs") (print freq))) (singing-append-feature! UTT 'freq freq)) UTT) ("(DURATION" (ATTLIST UTT) (set! singing_dur_att_list ATTLIST) UTT) (")DURATION" (ATTLIST UTT) (let ((dur (get-durs singing_dur_att_list))) (if singing-debug (begin (print "durs") (print dur))) (singing-append-feature! UTT 'dur dur)) UTT) ("(REST" (ATTLIST UTT) (let ((dur (get-durs ATTLIST))) (if singing-debug (begin (print "rest durs") (print dur))) (singing-append-feature! UTT 'rest (caar dur))) UTT) )) ;; ;; get-bpm ;; ;; Given the attribute list of a SINGING tag, returns the beats ;; per minute of the song from the BPM parameter. ;; (define (get-bpm atts) (parse-number (car (car (cdr (assoc 'BPM atts)))))) ;; ;; get-durs ;; ;; Given the attribute list of a DURATION tag, returns a list of ;; durations in seconds for the syllables of the word enclosed by ;; this tag. ;; ;; It first looks for a BEATS parameter, and converts these to ;; seconds using BPM, which was set in the SINGING tag. If this ;; is not present, it looks for the SECONDS parameter. ;; (define (get-durs atts) (let ((seconds (car (car (cdr (assoc 'SECONDS atts))))) (beats (car (car (cdr (assoc 'BEATS atts)))))) (if (equal? beats 'X) (mapcar (lambda (lst) (mapcar parse-number lst)) (string->list seconds)) (mapcar (lambda (lst) (mapcar (lambda (x) (/ (parse-number x) singing_bps)) lst)) (string->list beats))))) ;; ;; get-freqs ;; ;; Given the attribute list of a PITCH tag, returns a list of ;; frequencies in Hertz for the syllables of the word enclosed by ;; this tag. ;; ;; It first looks for a NOTE parameter, which can contain a MIDI ;; note of the form "C4", "D#3", or "Ab6", and if this is not ;; present it looks for the FREQ parameter. ;; (define (get-freqs atts) (let ((freqs (car (car (cdr (assoc 'FREQ atts))))) (notes (car (car (cdr (assoc 'NOTE atts)))))) (if (equal? notes 'X) (mapcar (lambda (lst) (mapcar parse-number lst)) (string->list freqs)) (mapcar (lambda (lst) (mapcar note->freq lst)) (string->list notes))))) ;; ;; note->freq ;; ;; Converts a string representing a MIDI note such as "C4" and ;; turns it into a frequency. We use the convention that ;; A5=440 (some call this note A3). ;; (define (note->freq note) (if singing-debug (format t "note is %l\n" note)) (set! note (format nil "%s" note)) (if singing-debug (print_string note)) (let (l octave notename midinote thefreq) (set! l (string-length note)) (set! octave (substring note (- l 1) 1)) (set! notename (substring note 0 (- l 1))) (set! midinote (+ (* 12 (parse-number octave)) (notename->midioffset notename))) (set! thefreq (midinote->freq midinote)) (if singing-debug (format t "note %s freq %f\n" note thefreq)) thefreq)) ;; ;; midinote->freq ;; ;; Converts a MIDI note number (1 - 127) into a frequency. We use ;; the convention that 69 = "A5" =440 Hz. ;; (define (midinote->freq midinote) (* 440.0 (pow 2.0 (/ (- midinote 69) 12)))) ;; ;; notename->midioffset ;; ;; Utility function that looks up the name of a note like "F#" and ;; returns its offset from C. ;; (define (notename->midioffset notename) (parse-number (car (cdr (assoc_string notename note_names))))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; Pitch modification functions ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; singing_f0_targets ;; ;; This function replaces the normal intonation function used in ;; festival. For each syllable, it extracts the frequency that ;; was calculated from the XML tags and stored in the token this ;; syllable comes from, and sets this frequency as both the start ;; and end f0 target. Really straightforward! ;; (defvar singing-last-f0 nil) (define (singing_f0_targets utt syl) "(singing_f0_targets utt syl)" (let ((start (item.feat syl 'syllable_start)) (end (item.feat syl 'syllable_end)) (freqs (mapcar parse-number (syl->freq syl))) (durs (syl->durations syl))) (let ((total-durs (apply + durs)) (total-time (- end start)) (time start) (prev-segment (item.prev (item.relation (item.daughter1 (item.relation syl 'SylStructure)) 'Segment))) (last-f0 singing-last-f0)) (if freqs (begin (set! singing-last-f0 (car (last freqs))) (append (if (and last-f0 prev-segment (item.prev prev-segment) (string-equal (item.feat prev-segment 'name) (car (car (cdr (car (PhoneSet.description '(silences)))))))) (let ((s (item.feat prev-segment "p.end")) (e (item.feat prev-segment "end"))) (list (list (+ s (* (- e s) 0.8)) last-f0) (list (+ s (* (- e s) 0.9)) (car freqs))))) (apply append (mapcar (lambda (d f) (let ((range (* (/ d total-durs) total-time)) (old-time time)) (set! time (+ time range)) (let ((range-fraction (* 0.1 range))) (list (list (+ old-time range-fraction) f) (list (- time range-fraction) f))))) durs freqs)))))))) ;; ;; syl->freq ;; ;; Given a syllable, looks up the frequency in its token. The token ;; stores a list of all of the frequencies associated with its ;; syllables, so this syllable grabs the frequency out of the list ;; corresponding to its index within the word. (This assumes that ;; a frequency was given for each syllable, and that a token ;; corresponds directly to a word. Singing-mode is not guaranteed ;; to work at all if either of these things are not true.) ;; (define (syl->freq syl) (let ((index (item.feat syl "R:Syllable.pos_in_word")) (freqs (singing-feat syl "R:SylStructure.parent.R:Token.parent.freq"))) (nth index freqs))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; Duration modification functions ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; singing_duration_method ;; ;; Calculates the duration of each phone in the utterance, in three ;; passes. Consult the three functions it calls, below, to see what ;; each one does. ;; (define (singing_duration_method utt) (mapcar singing_adjcons_syllable (utt.relation.items utt 'Syllable)) (singing_do_initial utt (car (utt.relation.items utt 'Token))) (mapcar singing_do_syllable (utt.relation.items utt 'Syllable)) (mapcar singing_fix_segment (utt.relation.items utt 'Segment)) utt) ;; ;; singing_adjcons_syllable ;; ;; First pass. Looks at the first phone of each syllable and ;; adjusts the starting time of this syllable such that the ;; perceived start time of the first phone is at the beginning ;; of the originally intended start time of the syllable. ;; ;; If this is not done, telling it to say the word "ta" at time ;; 2.0 actually doesn't "sound" like it says the "t" sound until ;; about 2.1 seconds. ;; ;; This function has a little bit of duplicated code from ;; singing_do_syllable, below - it could be modularized a little ;; better. ;; (define (singing_adjcons_syllable syl) (let ((totlen (apply + (mapcar (lambda (s) (get_avg_duration (item.feat s "name"))) (item.leafs (item.relation syl 'SylStructure))))) (syldur (apply + (syl->durations syl))) ;; figure out the offset of the first phone (phone1 (item.daughter1 (item.relation syl 'SylStructure))) (prevsyl (item.prev (item.relation syl 'Syllable)))) (let ((offset (get_duration_offset (item.feat phone1 "name")))) (if singing-debug (format t "offset: %f\n" offset) ) (if (< syldur totlen) (set! offset (* offset (/ syldur totlen)))) (if singing-debug (format t "Want to adjust syl by %f\n" offset)) (if prevsyl (begin (item.set_feat prevsyl 'subtractoffset offset) (item.set_feat syl 'addoffset offset)))))) ;; ;; singing_do_syllable ;; ;; Second pass. For each syllable, adds up the amount of time ;; that would normally be spent in consonants and vowels, based ;; on the average durations of these phones. Then, if the ;; intended length of this syllable is longer than this total, ;; stretch only the vowels; otherwise shrink all phones ;; proportionally. This function actually sets the "end" time ;; of each phone using a global "singing_global_time" variable. ;; ;; We also handle rests at this point, which are tagged onto the ;; end of the previous token. ;; (defvar singing-max-short-vowel-length 0.11) (define (singing_do_initial utt token) (if (equal? (item.name token) "") (let ((restlen (car (item.feat token 'rest)))) (if singing-debug (format t "restlen %l\n" restlen)) (if (> restlen 0) (let ((silence (car (car (cdr (assoc 'silences (PhoneSet.description))))))) (set! singing_global_time restlen) (item.relation.insert (utt.relation.first utt 'Segment) 'Segment (list silence (list (list "end" singing_global_time))) 'before)))))) (define (singing_do_syllable syl) (let ((conslen 0.0) (vowlen 0.0) (segments (item.leafs (item.relation syl 'SylStructure)))) ;; if there are no vowels, turn a middle consonant into a vowel; ;; hopefully this works well for languages where syllables may be ;; created by some consonants too (let ((segments* segments) (vowel-found nil)) (while (and segments* (not vowel-found)) (if (equal? "+" (item.feat (car segments*) "ph_vc")) (set! vowel-found t) (set! segments* (cdr segments*)))) (if (not vowel-found) (item.set_feat (nth (nint (/ (- (length segments) 1) 2)) segments) "singing-vc" "+"))) ;; sum up the length of all of the vowels and consonants in ;; this syllable (mapcar (lambda (s) (let ((slen (get_avg_duration (item.feat s "name")))) (if (or (equal? "+" (item.feat s "ph_vc")) (equal? "+" (item.feat s "singing-vc"))) (set! vowlen (+ vowlen slen)) (set! conslen (+ conslen slen))))) segments) (let ((totlen (+ conslen vowlen)) (syldur (apply + (syl->durations syl))) (addoffset (item.feat syl 'addoffset)) (subtractoffset (item.feat syl 'subtractoffset)) offset) (set! offset (- subtractoffset addoffset)) (if singing-debug (format t "Vowlen: %f conslen: %f totlen: %f\n" vowlen conslen totlen)) (if (< offset (/ syldur 2.0)) (begin (set! syldur (- syldur offset)) (if singing-debug (format t "Offset: %f\n" offset)))) (if singing-debug (format t "Syldur: %f\n" syldur)) (if (> totlen syldur) ;; if the total length of the average durations in the syllable is ;; greater than the total desired duration of the syllable, stretch ;; the time proportionally for each phone (let ((stretch (/ syldur totlen))) (mapcar (lambda (s) (let ((slen (* stretch (get_avg_duration (item.feat s "name"))))) (set! singing_global_time (+ slen singing_global_time)) (item.set_feat s 'end singing_global_time))) (item.leafs (item.relation syl 'SylStructure)))) ;; otherwise, stretch the vowels and not the consonants (let ((voweltime (- syldur conslen))) (let ((vowelstretch (/ voweltime vowlen)) (phones (mapcar car (car (cdar (PhoneSet.description '(phones))))))) (mapcar (lambda (s) (let ((slen (get_avg_duration (item.feat s "name")))) (if (or (equal? "+" (item.feat s "ph_vc")) (equal? "+" (item.feat s "singing-vc"))) (begin (set! slen (* vowelstretch slen)) ;; If the sound is long enough, better results ;; may be achieved by using longer versions of ;; the vowels. (if (> slen singing-max-short-vowel-length) (let ((sname (string-append (item.feat s "name") ":"))) (if (member_string sname phones) (item.set_feat s "name" sname)))))) (set! singing_global_time (+ slen singing_global_time)) (item.set_feat s 'end singing_global_time))) segments)))))) (let ((restlen (car (syl->rest syl)))) (if singing-debug (format t "restlen %l\n" restlen)) (if (> restlen 0) (let ((lastseg (item.daughtern (item.relation syl 'SylStructure))) (silence (car (car (cdr (assoc 'silences (PhoneSet.description)))))) (singing_global_time* singing_global_time)) (let ((seg (item.relation lastseg 'Segment)) (extra-pause-length 0.00001)) (set! singing_global_time (+ restlen singing_global_time)) (item.insert seg (list silence (list (list "end" singing_global_time))) 'after) ;; insert a very short extra pause to avoid after-effects, especially ;; after vowels (if (and seg (equal? (item.feat seg "ph_vc") "+") (< extra-pause-length restlen)) (item.insert seg (list silence (list (list "end" (+ singing_global_time* extra-pause-length)))) 'after))))))) ;; ;; singing_fix_segment ;; ;; Third pass. Finds any segments (phones) that we didn't catch earlier ;; (say if they didn't belong to a syllable, like silence) and sets them ;; to zero duration ;; (define (singing_fix_segment seg) (if (equal? 0.0 (item.feat seg 'end)) (if (equal? nil (item.prev seg)) (item.set_feat seg 'end 0.0) (item.set_feat seg 'end (item.feat (item.prev seg) 'end))) (if singing-debug (format t "segment: %s end: %f\n" (item.name seg) (item.feat seg 'end))))) ;; returns the duration of a syllable (stored in its token) (define (syl->durations syl) (let ((index (item.feat syl "R:Syllable.pos_in_word")) (durs (singing-feat syl "R:SylStructure.parent.R:Token.parent.dur"))) (mapcar parse-number (nth index durs)))) ;; returns the duration of the rest following a syllable (define (syl->rest syl) (let ((index (item.feat syl "R:Syllable.pos_in_word")) (durs (singing-feat syl "R:SylStructure.parent.R:Token.parent.dur")) (pauselen (singing-feat syl "R:SylStructure.parent.R:Token.parent.rest"))) (if (equal? index (- (length durs) 1)) (list (or pauselen 0.0)) (list 0.0)))) ;; get the average duration of a phone (define (get_avg_duration phone) (let ((pd (assoc_string phone phoneme_durations))) (if pd (car (cdr pd)) 0.08))) ;; get the duration offset of a phone (see the description above) (define (get_duration_offset phone) (parse-number (car (cdr (assoc_string phone phoneme_offsets*))))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; Other utility functions ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (char-quote string) (if (member string '("*" "+" "?" "[" "]" ".")) (string-append "[" string "]") string)) (define (split-string string separator) (if (string-matches string (string-append ".+" (char-quote separator) ".+")) (cons (string-before string separator) (split-string (string-after string separator) separator)) ;; We have to convert the weird XML attribute value type to string (list (string-append string "")))) (define (string->list string) (mapcar (lambda (s) (split-string s "+")) (split-string string ","))) (define (singing-append-feature! utt feature value) (let ((tokens (utt.relation.items utt 'Token))) (if tokens ;; we have to wrap value into a list to work around a Festival bug (item.set_feat (car (last tokens)) feature (list value)) (begin (utt.relation.append utt 'Token '("" ((name "") (whitespace "") (prepunctuation "") (punc "")))) (item.set_feat (car (last (utt.relation.items utt 'Token))) feature (list value)))))) (define (singing-feat item feature) (let ((value (item.feat item feature))) (if (equal? value 0) nil (car value)))) (define (current-language) (cadr (car (assoc 'language (voice.description current-voice))))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; Initializing and exiting singing mode ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; singing_init_func ;; (defvar singing_previous_eou_tree nil) (define (singing_init_func) "(singing_init_func) - Initialization for Singing mode" (if (not (symbol-bound? 'phoneme_durations)) (set! phoneme_durations '())) ;; use our intonation function (Parameter.set 'Int_Method 'General) (Parameter.set 'Int_Target_Method Int_Targets_General) (set! int_general_params `((targ_func ,singing_f0_targets))) (set! singing-last-f0 nil) ;; use our duration function (Parameter.set 'Duration_Method singing_duration_method) ;; set phoneme corrections for the current language (let ((language 'english ; (cadr (assoc 'language ; (cadr (voice.description current-voice)))) )) (set! phoneme_offsets* (cdr (assoc language phoneme_offsets)))) ;; avoid splitting to multiple utterances with insertion of unwanted pauses (set! singing_previous_eou_tree eou_tree) (set! eou_tree nil) ;; use our xml parsing function (set! singing_previous_elements xxml_elements) (set! xxml_elements singing_xml_elements)) ;; ;; singing_exit_func ;; (define (singing_exit_func) "(singing_exit_func) - Exit function for Singing mode" (set! eou_tree singing_previous_eou_tree) (set! xxml_elements singing_previous_elements)) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; Data tables ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (defvar note_names '((C 0) (C# 1) (Db 1) (D 2) (D# 3) (Eb 3) (E 4) (E# 5) (Fb 4) (F 5) (F# 6) (Gb 6) (G 7) (G# 8) (Ab 8) (A 9) (A# 10) (Bb 10) (B 11) (B# 12) (Cb 11))) ;; ;; The following list contains the offset into each phone that best ;; represents the perceptual onset of the phone. This is important ;; to know to get durations right in singing. For example, if the ;; offset for "t" is .060, and you want to start a "t" sound at ;; time 2.0 seconds, you should actually start the phone playing ;; at time 1.940 seconds in order for it to sound like the onset of ;; the "t" is really right at 2.0. ;; ;; These were derived empically by looking at and listening to the ;; waveforms of each phone for mwm's voice. ;; (defvar phoneme_offsets `((english (t 0.050) (T 0.050) (d 0.090) (D 0.090) (p 0.080) (b 0.080) (k 0.090) (g 0.100) (9r 0.050) ;; r (l 0.030) (f 0.050) (v 0.050) (s 0.040) (S 0.040) (z 0.040) (Z 0.040) (n 0.040) (N 0.040) (m 0.040) (j 0.090) (E 0.0) (> 0.0) (>i 0.0) (aI 0.0) (& 0.0) (3r 0.0) (tS 0.0) (oU 0.0) (aU 0.0) (A 0.0) (ei 0.0) (iU 0.0) (U 0.0) (@ 0.0) (h 0.0) (u 0.0) (^ 0.0) (I 0.0) (dZ 0.0) (i: 0.0) (w 0.0) (pau 0.0) (brth 0.0) (h# 0.0) ))) (defvar phoneme_offsets* nil) ;; ;; Declare the new mode to Festival ;; (set! tts_text_modes (cons `(singing ;; mode name ((init_func ,singing_init_func) (exit_func ,singing_exit_func) (analysis_type xml))) tts_text_modes)) (provide 'singing-mode) festival/lib/hts.scm0000644000175000017500000004270612442333465013066 0ustar awbawb;; ---------------------------------------------------------------- ;; ;; Nagoya Institute of Technology and ;; ;; Carnegie Mellon University ;; ;; Copyright (c) 2002 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and ;; ;; distribute this software and its documentation without ;; ;; restriction, including without limitation the rights to use, ;; ;; copy, modify, merge, publish, distribute, sublicense, and/or ;; ;; sell copies of this work, and to permit persons to whom this ;; ;; work is furnished to do so, subject to the following conditions: ;; ;; ;; ;; 1. The code must retain the above copyright notice, this list ;; ;; of conditions and the following disclaimer. ;; ;; ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; ;; ;; 3. Original authors' names are not deleted. ;; ;; ;; ;; 4. The authors' names are not used to endorse or promote ;; ;; products derived from this software without specific prior ;; ;; written permission. ;; ;; ;; ;; NAGOYA INSTITUTE OF TECHNOLOGY, CARNEGIE MELLON UNIVERSITY AND ;; ;; THE CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH ;; ;; REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF ;; ;; MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL NAGOYA INSTITUTE ;; ;; OF TECHNOLOGY, CARNEGIE MELLON UNIVERSITY NOR THE CONTRIBUTORS ;; ;; BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ;; ;; ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR ;; ;; PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER ;; ;; TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR ;; ;; PERFORMANCE OF THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; Generic HTS support code and specific features ;; ;; http://hts.ics.nitech.ac.jp ;; ;; Author : Alan W Black ;; ;; Date : August 2002 (and April 2004) ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;; ;; ;; Still has language specific features in here, that will have to ;; ;; move out to the voices ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (defvar hts_synth_pre_hooks nil) (defvar hts_synth_post_hooks nil) (defvar hts_engine_params nil) (defvar hts_duration_stretch 0) (defvar hts_f0_mean 0) (defvar hts_f0_std 1) (defvar hts_fw_factor 0.42) (defvar hts_total_length 0.0) (defvar hts_uv_threshold 0.5) (defvar hts_use_phone_align 0) (defSynthType HTS (let ((featfile (make_tmp_filename)) (mcepfile (make_tmp_filename)) (f0file (make_tmp_filename)) (wavfile (make_tmp_filename)) (labfile (make_tmp_filename))) (apply_hooks hts_synth_pre_hooks utt) (set! hts_output_params (list (list "-labelfile" featfile) (list "-om" mcepfile) (list "-of" f0file) (list "-or" wavfile) (list "-od" labfile)) ) (hts_dump_feats utt hts_feats_list featfile) (HTS_Synthesize utt) (delete-file featfile) (delete-file mcepfile) (delete-file f0file) (delete-file wavfile) (delete-file labfile) (apply_hooks hts_synth_post_hooks utt) utt) ) (define (hts_feats_output ofd s) "This is bad as it makes decisions about what the feats are" ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; SEGMENT ; boundary (format ofd "%10.0f %10.0f " (* 10000000 (item.feat s "segment_start")) (* 10000000 (item.feat s "segment_end"))) ; pp.name (format ofd "%s" (if (string-equal "0" (item.feat s "p.p.name")) "x" (item.feat s "p.p.name"))) ; p.name (format ofd "^%s" (if (string-equal "0" (item.feat s "p.name")) "x" (item.feat s "p.name"))) ; c.name (format ofd "-%s" (if (string-equal "0" (item.feat s "name")) "x" (item.feat s "name"))) ; n.name (format ofd "+%s" (if (string-equal "0" (item.feat s "n.name")) "x" (item.feat s "n.name"))) ; nn.name (format ofd "=%s" (if (string-equal "0" (item.feat s "n.n.name")) "x" (item.feat s "n.n.name"))) ; position in syllable (segment) (format ofd "@") (format ofd "%s" (if (string-equal "pau" (item.feat s "name")) "x" (+ 1 (item.feat s "pos_in_syl")))) (format ofd "_%s" (if (string-equal "pau" (item.feat s "name")) "x" (- (item.feat s "R:SylStructure.parent.R:Syllable.syl_numphones") (item.feat s "pos_in_syl")))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; SYLLABLE ;; previous syllable ; p.stress (format ofd "/A:%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "p.R:SylStructure.parent.R:Syllable.stress") (item.feat s "R:SylStructure.parent.R:Syllable.p.stress"))) ; p.accent (format ofd "_%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "p.R:SylStructure.parent.R:Syllable.accented") (item.feat s "R:SylStructure.parent.R:Syllable.p.accented"))) ; p.length (format ofd "_%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "p.R:SylStructure.parent.R:Syllable.syl_numphones") (item.feat s "R:SylStructure.parent.R:Syllable.p.syl_numphones"))) ;; current syllable ; c.stress (format ofd "/B:%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.R:Syllable.stress"))) ; c.accent (format ofd "-%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.R:Syllable.accented"))) ; c.length (format ofd "-%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.R:Syllable.syl_numphones"))) ; position in word (syllable) (format ofd "@%s" (if (string-equal "pau" (item.feat s "name")) "x" (+ 1 (item.feat s "R:SylStructure.parent.R:Syllable.pos_in_word")))) (format ofd "-%s" (if (string-equal "pau" (item.feat s "name")) "x" (- (item.feat s "R:SylStructure.parent.parent.R:Word.word_numsyls") (item.feat s "R:SylStructure.parent.R:Syllable.pos_in_word")))) ; position in phrase (syllable) (format ofd "&%s" (if (string-equal "pau" (item.feat s "name")) "x" (+ 1 (item.feat s "R:SylStructure.parent.R:Syllable.syl_in")))) (format ofd "-%s" (if (string-equal "pau" (item.feat s "name")) "x" (+ 1 (item.feat s "R:SylStructure.parent.R:Syllable.syl_out")))) ; position in phrase (stressed syllable) (format ofd "#%s" (if (string-equal "pau" (item.feat s "name")) "x" (+ 1 (item.feat s "R:SylStructure.parent.R:Syllable.ssyl_in")))) (format ofd "-%s" (if (string-equal "pau" (item.feat s "name")) "x" (+ 1 (item.feat s "R:SylStructure.parent.R:Syllable.ssyl_out")))) ; position in phrase (accented syllable) (format ofd "$%s" (if (string-equal "pau" (item.feat s "name")) "x" (+ 1 (item.feat s "R:SylStructure.parent.R:Syllable.asyl_in")))) (format ofd "-%s" (if (string-equal "pau" (item.feat s "name")) "x" (+ 1 (item.feat s "R:SylStructure.parent.R:Syllable.asyl_out")))) ; distance from stressed syllable (format ofd "!%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.R:Syllable.lisp_distance_to_p_stress"))) (format ofd "-%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.R:Syllable.lisp_distance_to_n_stress"))) ; distance from accented syllable (format ofd ";%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.R:Syllable.lisp_distance_to_p_accent"))) (format ofd "-%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.R:Syllable.lisp_distance_to_n_accent"))) ; name of the vowel of current syllable (format ofd "|%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.R:Syllable.syl_vowel"))) ;; next syllable (format ofd "/C:%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "n.R:SylStructure.parent.R:Syllable.stress") (item.feat s "R:SylStructure.parent.R:Syllable.n.stress"))) ; n.accent (format ofd "+%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "n.R:SylStructure.parent.R:Syllable.accented") (item.feat s "R:SylStructure.parent.R:Syllable.n.accented"))) ; n.length (format ofd "+%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "n.R:SylStructure.parent.R:Syllable.syl_numphones") (item.feat s "R:SylStructure.parent.R:Syllable.n.syl_numphones"))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ; WORD ;;;;;;;;;;;;;;;;;; ;; previous word ; p.gpos (format ofd "/D:%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "p.R:SylStructure.parent.parent.R:Word.gpos") (item.feat s "R:SylStructure.parent.parent.R:Word.p.gpos"))) ; p.lenght (syllable) (format ofd "_%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "p.R:SylStructure.parent.parent.R:Word.word_numsyls") (item.feat s "R:SylStructure.parent.parent.R:Word.p.word_numsyls"))) ;;;;;;;;;;;;;;;;; ;; current word ; c.gpos (format ofd "/E:%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.parent.R:Word.gpos"))) ; c.lenght (syllable) (format ofd "+%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.parent.R:Word.word_numsyls"))) ; position in phrase (word) (format ofd "@%s" (if (string-equal "pau" (item.feat s "name")) "x" (+ 1 (item.feat s "R:SylStructure.parent.parent.R:Word.pos_in_phrase")))) (format ofd "+%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.parent.R:Word.words_out"))) ; position in phrase (content word) (format ofd "&%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.parent.R:Word.content_words_in"))) (format ofd "+%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.parent.R:Word.content_words_out"))) ; distance from content word in phrase (format ofd "#%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.parent.R:Word.lisp_distance_to_p_content"))) (format ofd "+%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.parent.R:Word.lisp_distance_to_n_content"))) ;;;;;;;;;;;;;; ;; next word ; n.gpos (format ofd "/F:%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "n.R:SylStructure.parent.parent.R:Word.gpos") (item.feat s "R:SylStructure.parent.parent.R:Word.n.gpos"))) ; n.lenghte (syllable) (format ofd "_%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "n.R:SylStructure.parent.parent.R:Word.word_numsyls") (item.feat s "R:SylStructure.parent.parent.R:Word.n.word_numsyls"))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ; PHRASE ;;;;;;;;;;;;;;;;;;;; ;; previous phrase ; length of previous phrase (syllable) (format ofd "/G:%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "p.R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_syls_in_phrase") (item.feat s "R:SylStructure.parent.parent.R:Phrase.parent.p.lisp_num_syls_in_phrase"))) ; length of previous phrase (word) (format ofd "_%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "p.R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_words_in_phrase") (item.feat s "R:SylStructure.parent.parent.R:Phrase.parent.p.lisp_num_words_in_phrase"))) ;;;;;;;;;;;;;;;;;;;; ;; current phrase ; length of current phrase (syllable) (format ofd "/H:%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_syls_in_phrase"))) ; length of current phrase (word) (format ofd "=%s" (if (string-equal "pau" (item.feat s "name")) "x" (item.feat s "R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_words_in_phrase"))) ; position in major phrase (phrase) (format ofd "@%s" (+ 1 (item.feat s "R:SylStructure.parent.R:Syllable.sub_phrases"))) (format ofd "=%s" (- (item.feat s "lisp_total_phrases") (item.feat s "R:SylStructure.parent.R:Syllable.sub_phrases"))) ; type of tobi endtone of current phrase (format ofd "|%s" (item.feat s "R:SylStructure.parent.parent.R:Phrase.parent.daughtern.R:SylStructure.daughtern.tobi_endtone")) ;;;;;;;;;;;;;;;;;;;; ;; next phrase ; length of next phrase (syllable) (format ofd "/I:%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "n.R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_syls_in_phrase") (item.feat s "R:SylStructure.parent.parent.R:Phrase.parent.n.lisp_num_syls_in_phrase"))) ; length of next phrase (word) (format ofd "=%s" (if (string-equal "pau" (item.feat s "name")) (item.feat s "n.R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_words_in_phrase") (item.feat s "R:SylStructure.parent.parent.R:Phrase.parent.n.lisp_num_words_in_phrase"))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ; UTTERANCE ; length (syllable) (format ofd "/J:%s" (item.feat s "lisp_total_syls")) ; length (word) (format ofd "+%s" (item.feat s "lisp_total_words")) ; length (phrase) (format ofd "-%s" (item.feat s "lisp_total_phrases")) (format ofd "\n") ) (define (hts_dump_feats utt feats ofile) (let ((ofd (fopen ofile "w"))) (mapcar (lambda (s) (hts_feats_output ofd s)) (utt.relation.items utt 'Segment)) (fclose ofd) )) ;; ;; Extra features ;; From Segment items refer by ;; ;; R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_syls_in_phrase ;; R:SylStructure.parent.parent.R:Phrase.parent.lisp_num_words_in_phrase ;; lisp_total_words ;; lisp_total_syls ;; lisp_total_phrases ;; ;; The last three will act on any item (define (distance_to_p_content i) (let ((c 0) (rc 0 ) (w (item.relation.prev i "Phrase"))) (while w (set! c (+ 1 c)) (if (string-equal "1" (item.feat w "contentp")) (begin (set! rc c) (set! w nil)) (set! w (item.prev w))) ) rc)) (define (distance_to_n_content i) (let ((c 0) (rc 0) (w (item.relation.next i "Phrase"))) (while w (set! c (+ 1 c)) (if (string-equal "1" (item.feat w "contentp")) (begin (set! rc c) (set! w nil)) (set! w (item.next w))) ) rc)) (define (distance_to_p_accent i) (let ((c 0) (rc 0 ) (w (item.relation.prev i "Syllable"))) (while (and w (member_string (item.feat w "syl_break") '("0" "1"))) (set! c (+ 1 c)) (if (string-equal "1" (item.feat w "accented")) (begin (set! rc c) (set! w nil)) (set! w (item.prev w))) ) rc)) (define (distance_to_n_accent i) (let ((c 0) (rc 0 ) (w (item.relation.next i "Syllable"))) (while (and w (member_string (item.feat w "p.syl_break") '("0" "1"))) (set! c (+ 1 c)) (if (string-equal "1" (item.feat w "accented")) (begin (set! rc c) (set! w nil)) (set! w (item.next w))) ) rc)) (define (distance_to_p_stress i) (let ((c 0) (rc 0 ) (w (item.relation.prev i "Syllable"))) (while (and w (member_string (item.feat w "syl_break") '("0" "1"))) (set! c (+ 1 c)) (if (string-equal "1" (item.feat w "stress")) (begin (set! rc c) (set! w nil)) (set! w (item.prev w))) ) rc)) (define (distance_to_n_stress i) (let ((c 0) (rc 0 ) (w (item.relation.next i "Syllable"))) (while (and w (member_string (item.feat w "p.syl_break") '("0" "1"))) (set! c (+ 1 c)) (if (string-equal "1" (item.feat w "stress")) (begin (set! rc c) (set! w nil)) (set! w (item.next w))) ) rc)) (define (num_syls_in_phrase i) (apply + (mapcar (lambda (w) (length (item.relation.daughters w 'SylStructure))) (item.relation.daughters i 'Phrase)))) (define (num_words_in_phrase i) (length (item.relation.daughters i 'Phrase))) (define (total_words w) (length (utt.relation.items (item.get_utt w) 'Word))) (define (total_syls s) (length (utt.relation.items (item.get_utt s) 'Syllable))) (define (total_phrases s) (length (utt.relation_tree (item.get_utt s) 'Phrase))) (provide 'hts) festival/lib/etc/Makefile0000644000175000017500000000556007262605650013776 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Directory for various scripts (machine independent functions) ## ## Sub-directories of this will contain machine-dependent binaries ## ## ## ########################################################################### TOP=../.. DIRNAME=lib/etc FILTERS=email_filter FILES=Makefile $(FILTERS) include $(TOP)/config/common_make_rules festival/lib/etc/email_filter0000755000175000017500000000614107262605650014714 0ustar awbawb#!/bin/sh ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Email filter for tts text mode ## ## usage: email_filter email_message >filtered_message ## ## ## ## Extracts the From and Subject lines form the head and the body of ## ## of the message, I suppose it could also do signature extraction ## ## ## ########################################################################### grep "^From: " $1 echo grep "^Subject: " $1 echo # delete up to first blank line (i.e. the header) sed '1,/^$/ d' $1 festival/lib/multisyn/Makefile0000644000175000017500000000564710065564264015115 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 2004 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### # # # Makefile for lib/multisyn directory # # # ########################################################################### TOP=../.. DIRNAME=lib/multisyn PHONESETS = radio_phones_multisyn.scm GENERAL = multisyn.scm multisyn_pauses.scm target_cost.scm OTHERS = send_xwaves.scm FILES=Makefile $(PHONESETS) $(GENERAL) $(OTHERS) include $(TOP)/config/common_make_rules festival/lib/multisyn/radio_phones_multisyn.scm0000644000175000017500000001360510065567162020570 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997,2003, 2004 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; A definition of the radio phone set used in the BU RADIO FM ;;; corpus, some people call this the darpa set. This one ;;; has the closures removed and pauses added for multisyn ;;; (defPhoneSet radio_multisyn ;;; Phone Features (;; vowel or consonant (vc + -) ;; vowel length: short long dipthong schwa (vlng s l d a 0) ;; vowel height: high mid low (vheight 1 2 3 0) ;; vowel frontness: front mid back (vfront 1 2 3 0) ;; lip rounding (vrnd + - 0) ;; consonant type: stop fricative affricate nasal lateral approximant (ctype s f a n l r 0) ;; place of articulation: labial alveolar palatal labio-dental ;; dental velar glottal (cplace l a p b d v g 0) ;; consonant voicing (cvox + - 0) ) ;; Phone set members ( ;; multisyn extras (# - 0 0 0 0 0 0 -) ;; slience ... (B_10 - 0 0 0 0 0 0 -) ;; Pauses (B_20 - 0 0 0 0 0 0 -) ;; Pauses (B_30 - 0 0 0 0 0 0 -) ;; Pauses (B_40 - 0 0 0 0 0 0 -) ;; Pauses (B_50 - 0 0 0 0 0 0 -) ;; Pauses (B_100 - 0 0 0 0 0 0 -) ;; Pauses (B_150 - 0 0 0 0 0 0 -) ;; Pauses (B_200 - 0 0 0 0 0 0 -) ;; Pauses (B_250 - 0 0 0 0 0 0 -) ;; Pauses (B_300 - 0 0 0 0 0 0 -) ;; Pauses (B_400 - 0 0 0 0 0 0 -) ;; Pauses ;; Note these features were set by awb so they are wrong !!! (aa + l 3 3 - 0 0 0) ;; father (ae + s 3 1 - 0 0 0) ;; fat (ah + s 2 2 - 0 0 0) ;; but (ao + l 3 3 + 0 0 0) ;; lawn (aw + d 3 2 - 0 0 0) ;; how (ax + a 2 2 - 0 0 0) ;; about (axr + a 2 2 - r a +) (ay + d 3 2 - 0 0 0) ;; hide (b - 0 0 0 0 s l +) (ch - 0 0 0 0 a p -) (d - 0 0 0 0 s a +) (dh - 0 0 0 0 f d +) (dx - a 0 0 0 s a +) ;; ?? (eh + s 2 1 - 0 0 0) ;; get (el + s 0 0 0 l a +) (em + s 0 0 0 n l +) (en + s 0 0 0 n a +) (er + a 2 2 - r 0 0) ;; always followed by r (er-r == axr) (ey + d 2 1 - 0 0 0) ;; gate (f - 0 0 0 0 f b -) (g - 0 0 0 0 s v +) (hh - 0 0 0 0 f g -) (hv - 0 0 0 0 f g +) (ih + s 1 1 - 0 0 0) ;; bit (iy + l 1 1 - 0 0 0) ;; beet (jh - 0 0 0 0 a p +) (k - 0 0 0 0 s v -) (l - 0 0 0 0 l a +) (m - 0 0 0 0 n l +) (n - 0 0 0 0 n a +) (nx - 0 0 0 0 n d +) ;; ??? (ng - 0 0 0 0 n v +) (ow + d 2 3 + 0 0 0) ;; lone (oy + d 2 3 + 0 0 0) ;; toy (p - 0 0 0 0 s l -) (r - 0 0 0 0 r a +) (s - 0 0 0 0 f a -) (sh - 0 0 0 0 f p -) (t - 0 0 0 0 s a -) (th - 0 0 0 0 f d -) (uh + s 1 3 + 0 0 0) ;; full (uw + l 1 3 + 0 0 0) ;; fool (v - 0 0 0 0 f b +) (w - 0 0 0 0 r l +) (y - 0 0 0 0 r p +) (z - 0 0 0 0 f a +) (zh - 0 0 0 0 f p +) (pau - 0 0 0 0 0 0 -) (h# - 0 0 0 0 0 0 -) (brth - 0 0 0 0 0 0 -) ) ) (PhoneSet.silences '(# pau h# brth)) (provide 'radio_phones_multisyn) festival/lib/multisyn/multisyn.scm0000644000175000017500000001666312165025207016035 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 2003, 2004 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Multisyn top level scheme code (Korin Richmond and Rob Clark) ;;; ; Requires (require_module 'UniSyn) (require_module 'MultiSyn) (require 'multisyn_pauses) (require 'target_cost) ;; use a global parameter to specify which UnitSelection voice ;; to use to synthesise a given utterance for now, because the ;; standard Festival synthesis mainline doesn't accept a voice ;; parameter. (This should be set to the current voice object) (defvar currentMultiSynVoice nil) (defvar relp t) (defvar flattenVoice nil) ; extract utt list from a .data file (define (load_utt_list filename) "(load_utt_list filename) Loads a fextvox .data file and extracts an utterance list." (let (l entries) (set! entries (load filename t)) (mapcar (lambda (d) (set! l (cons (car d) l)) t) entries) l)) ;; SynthType definition, main entry point. (defSynthType MultiSyn ;(print "Multisyn unit selection synthesis") (defvar MultiSyn_module_hooks nil) (Param.def "unisyn.window_name" "hanning") (Param.def "unisyn.window_factor" 1.0) ;; Unisyn requires these to be set. (set! us_abs_offset 0.0) (set! us_rel_offset 0.0) (apply_hooks MultiSyn_module_hooks utt) ;; 4processing of diphone names ;; find appropriate unit sequence and put sythesis ;; parameters in the Unit relation of the utterance structure (voice.getUnits currentMultiSynVoice utt) ;(print "doing concat") (us_unit_concat utt) ;(print "doing raw concat") (utt.relation.create utt 'SourceSegments) (set! do_prosmod (du_voice.prosodic_modification currentMultiSynVoice)) (if do_prosmod (begin (if (not (member 'f0 (utt.relationnames utt))) (targets_to_f0 utt)) ;; temporary fix (if (utt.relation.last utt 'Segment) (set! pm_end (+ (item.feat (utt.relation.last utt 'Segment) "end") 0.02)) (set! pm_end 0.02)) (us_f0_to_pitchmarks utt 'f0 'TargetCoef pm_end) (us_mapping utt 'segment_single)) (begin (utt.copy_relation utt 'SourceCoef 'TargetCoef) (us_mapping utt "linear"))) ;(print "generating wave") ;; specify something else if you don't want lpc (us_generate_wave utt 'lpc) ) ; target cost scheme code (define (targetcost it1 it2) (Default_Target_Cost it1 it2)) ; Evil function which writes the functions to actually load and switch new voices. (define (make_voice_definition name srate config_function backoff_rules data_dir config) "(make_voice_definition NAME SRATE CONFIG_FUNCTION BACKOFF_RULES DATA_DIR CONFIG) Create the fuction definitions to load and unload a voice." (let ((voice_name (string-append "voice_" name)) (free_name (string-append "free_voice_" name)) (pre_config_function (string-append config_function "_pre")) (voice_variable (upcase (string-append "voice_" name)))) (eval (list 'defvar (intern voice_variable) nil)) (eval (list 'define (list (intern voice_name)) (list 'if (intern pre_config_function) (list (intern pre_config_function) (intern voice_variable))) (list 'if (list 'null (intern voice_variable)) (list 'set! (intern voice_variable) (list 'multisyn_load_voice_modules (list 'quote name) srate (list 'quote backoff_rules) data_dir (list 'quote config)))) (list (intern config_function) (intern voice_variable)) (list 'set! 'current-voice (list 'quote name)) (list 'define_current_voice_reset) (list 'set! 'currentMultiSynVoice (intern voice_variable)) )) (eval (list 'define (list (intern free_name)) (list 'cond (list (list 'null (intern voice_variable)) (list 'error "Voice not currently loaded!")) (list (list 'eq? 'currentMultiSynVoice (intern voice_variable)) (list 'error "Can't free current voice!")) (list 't (list set! (intern voice_variable) 'nil)))))) nil) (define (multisyn_load_voice_modules name srate backoff_rules base_dir module_list) "(multisyn_add_module voice name srate backoff_rules base_dir module_list) Add voice modules to a voice." (let (voice) (mapcar (lambda (module_entry) (let ((dirs (car module_entry)) (utt_list (load_utt_list (path-append base_dir (cadr module_entry))))) (if (null voice) (set! voice (make_du_voice utt_list dirs srate)) (voice.addModule voice utt_list dirs srate)))) module_list) (voice.setName voice name) (if flattenVoice (du_voice.setTargetCost voice "flat") (du_voice.setTargetCost voice t)) (du_voice.setJoinCost voice t) (format stderr "Please wait: Initialising multisyn voice.\n") (voice.init voice) (format stderr " Voice loaded successfully!\n") (du_voice.set_ob_pruning_beam voice 0.25) (du_voice.set_pruning_beam voice 0.25) (du_voice.setDiphoneBackoff voice backoff_rules) voice)) (define (define_current_voice_reset) "(define_current_voice_reset) Re-define (current_voice_reset) correctly." (eval (list 'define (list 'current_voice_reset) (list 'multisyn_reset_globals)))) (define (multisyn_reset_globals) "(multisyn_reset_globals) Reset multisyn specific global variables." (Param.set 'unisyn.window_symmetric 1)) (provide 'multisyn) festival/lib/multisyn/multisyn_pauses.scm0000644000175000017500000001157412442333465017417 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 2003, 2004 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Multisyn Pause module (Rob Clark and Korin Richmond) ;;; ;;; (defvar BB_Pause "B_300") (defvar B_Pause "B_150") (defvar mB_Pause "B_150") ; shouldn't be used (define (MultiSyn_Pauses utt) "(MultiSyn_Pauses UTT) Predict pause insertion in a Multisyn unit selection utterance structure." (let ((words (utt.relation.items utt 'Word)) lastword tpname) (if words (begin (insert_initial_pause utt) ;; always have a start pause (set! lastword (car (last words))) (mapcar (lambda (w) (let ((pbreak (item.feat w "pbreak")) (emph (item.feat w "R:Token.parent.EMPH"))) (cond ((string-equal pbreak "BB") (unitselection_pause_insert w BB_Pause)) ((string-equal pbreak "mB") (unitselection_pause_insert w mB_Pause)) ((string-equal pbreak "B") (unitselection_pause_insert w B_Pause))))) words) ;; The embarassing bit. Remove any words labelled as punc or fpunc (mapcar (lambda (w) (let ((pos (item.feat w "pos"))) (if (or (string-equal "punc" pos) (string-equal "fpunc" pos)) (let ((pbreak (item.feat w "pbreak")) (wp (item.relation w 'Phrase))) (if (and (string-matches pbreak "BB?") (item.relation.prev w 'Word)) (item.set_feat (item.relation.prev w 'Word) "pbreak" pbreak)) (item.relation.remove w 'Word) ;; can't refer to w as we've just deleted it (item.relation.remove wp 'Phrase))))) words))) ;(utt.relation.print utt 'Word) ;(utt.relation.print utt 'Segment) utt)) (define (unitselection_pause_insert word pause) "(pause_insert word pause) Insert segments needed for a pause." (let ((silence (car (cadr (car (PhoneSet.description '(silences)))))) (seg (item.relation (find_last_seg word) 'Segment)) pause_item) ;(format stderr " inserting pause after: %s.\n" (item.name seg)) ;(format stderr " Inserting pause\n") ; if next seg is not silence insert one. (if (or (not (item.next seg)) (not (string-equal (item.name (item.next seg)) silence))) (item.insert seg (list silence) 'after)) ; insert pause after that if not the end. (if (item.next (item.next seg)) (begin (set! pause_item (item.insert (item.next seg) (list pause) 'after)) ;if next seg after that is not silence add one. (if (not (string-equal (item.name (item.next pause_item)) silence)) (item.insert pause_item (list silence) 'after)))))) (provide 'multisyn_pauses) festival/lib/multisyn/target_cost.scm0000644000175000017500000003155410065567162016473 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 2003, 2004 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Multisyn scheme target cost (Rob Clark and Korin Richmond) ;;; ;;; (define (Default_Target_Cost targ cand) "(Default_Target_Cost targ cand) A Default Target Cost function." (let ((cost 0)) (mapcar (lambda (row) (set! cost (+ cost (tc_eval_row row targ cand)))) target_matrix) (set! cost (/ cost target_matrix_weight)) cost)) (define (tc_eval_row row targ cand) "(tc_eval_row row targ cand) Evaluate a target matrix row." (let ((weight (car row)) (func (cadr row)) (result 0)) (set! result (* weight (eval (list func targ cand)))) result)) ;; ;; Target cost Matrix ;; '(weight function) (define (get_matrix_weight m) (let ((w 0)) (mapcar (lambda (x) (set! w (+ w (car x)))) m) w)) (set! test_matrix_max_weight 1) (set! test_matrix '( (10 tc_stress ) (5 tc_syl_pos ) (5 tc_word_pos) (6 tc_partofspeech) (7 tc_phrase_pos) (4 tc_left_context) (3 tc_right_context) (25 tc_bad_f0) ;; set to equal 1/3 of total cost (so high because interaction with join) ; (0 tc_segment_score) ;; was 4. turned off until utterances are built for this. (10 tc_bad_duration) ;; was 6 )) (set! test_matrix_weight (* test_matrix_max_weight (get_matrix_weight test_matrix))) (set! target_matrix test_matrix) (set! target_matrix_weight test_matrix_weight) ;; ;; tc_stress ;; ;; Compares stress on any vowel which form part of the diphone. stress ;; conditions must match for a zero target cost. ;; (define (tc_stress targ cand) "(tc_stress targ cand) Target Cost stressed. 0 - stress patterns match [ compares: 0 unstressed vs. > 0 stressed ] 1 - stress miss-match. " (let ((c 0) cand_stress targ_stress) ;(format t "my_is_vowel %l\n" (my_is_vowel targ)) ;(format t "phone_is_silence %l\n" (phone_is_silence (item.feat targ 'name))) ;; For first segment (if (and (not (phone_is_silence (item.feat targ 'name))) (my_is_vowel targ)) (begin (set! cand_stress (item.feat cand "R:SylStructure.parent.stress")) (set! targ_stress (item.feat targ "R:SylStructure.parent.stress")) (if (or (and (eq? cand_stress 0) (> targ_stress 0)) (and (eq? targ_stress 0) (> cand_stress 0))) (set! c 1)))) ;; For second segment ;(format t "n.my_is_vowel %l\n" (my_is_vowel (item.next targ))) ;(format t "n.phone_is_silence %l\n" (phone_is_silence (item.feat targ 'n.name))) (if (and (not (phone_is_silence (item.feat targ 'n.name))) (my_is_vowel (item.next targ))) (begin (set! cand_stress (item.feat cand "n.R:SylStructure.parent.stress")) (set! targ_stress (item.feat targ "n.R:SylStructure.parent.stress")) (if (or (and (eq? cand_stress 0) (> targ_stress 0)) (and (eq? targ_stress 0) (> cand_stress 0))) (set! c 1)))) ; (format t "tc_stress: %l\n" c) c)) ;; ;; tc_syl_position ;; ;; Find and compare diphone position in syllabic structure. ;; Values are: inter - diphone crosses syllable boundary. ;; initial - diphone is syllable initial. ;; medial - diphone is syllable medial ;; final - diphone is syllable final ;; returns 0 for a match 1 for a mismatch. ;; (define (tc_syl_pos targ cand) "(tc_syl_pos targ cand) Score position in syllable." (let ((targ_pos "medial") (cand_pos "medial") (targ_syl (get_syl targ)) (targ_next_syl (get_syl (item.next targ))) (cand_syl (get_syl cand)) (cand_next_syl (get_syl (item.next cand)))) ;; target (cond ((not (equal? targ_syl targ_next_syl)) (set! targ_pos "inter")) ((not (equal? targ_syl (get_syl (item.prev targ)))) (set! targ_pos "initial")) ((not (equal? targ_next_syl (get_syl (item.next (item.next targ))))) (set! targ_pos "final"))) ;; candidate (cond ((not (equal? cand_syl cand_next_syl)) (set! cand_pos "inter")) ((not (equal? cand_syl (get_syl (item.prev cand)))) (set! cand_pos "initial")) ((not (equal? cand_next_syl (get_syl (item.next (item.next cand))))) (set! cand_pos "final"))) ; (format t "targ_syl: %l cand_syl %l\n" targ_pos cand_pos) (if (equal? targ_pos cand_pos) 0 1))) ;; ;; tc_word_position ;; ;; Find and compare diphone position in word structure ;; Values are: inter - diphone crosses word boundary. ;; initial - diphone is word initial. ;; medial - diphone is word medial ;; final - diphone is word final ;; returns 0 for a match 1 for a mismatch. ;; (define (tc_word_pos targ cand) "(tc_word_pos targ cand) Score position in word." (let ((targ_pos "medial") (cand_pos "medial") (targ_word (get_word targ)) (targ_next_word (get_word (item.next targ))) (cand_word (get_word cand)) (cand_next_word (get_word (item.next cand)))) ;; target (cond ((not (equal? targ_word targ_next_word)) (set! targ_pos "inter")) ((not (equal? targ_word (get_word (item.prev targ)))) (set! targ_pos "initial")) ((not (equal? targ_next_word (get_word (item.next (item.next targ))))) (set! targ_pos "final"))) ;; candidate (cond ((not (equal? cand_word cand_next_word)) (set! cand_pos "inter")) ((not (equal? cand_word (get_word (item.prev cand)))) (set! cand_pos "initial")) ((not (equal? cand_next_word (get_word (item.next (item.next cand))))) (set! cand_pos "final"))) ; (format t "targ_word: %l cand_word %l\n" targ_pos cand_pos) (if (equal? targ_pos cand_pos) 0 1))) ;; ;; tc_phrase_position ;; ;; Position (of word) in phrase ;; initial/medial/final ;; ;; 0 - match, 1 - mismatch ;; (define (tc_phrase_pos targ cand) "(tc_phrase_pos targ cand) Score position in phrase." (let ((targ_word (get_word targ)) (cand_word (get_word cand))) (cond ((and (null targ_word) (null cand_word)) 0) ((or (null targ_word) (null cand_word)) 1) ((string-equal (item.feat targ_word 'pbreak) (item.feat cand_word 'pbreak)) 0) (t 1)))) ;; ;; tc_partofspeech ;; ;; ;; (define (tc_partofspeech targ cand) "(tc_partofspeech targ cand) Score part of speech." (let ((targ_word (get_word targ)) (cand_word (get_word cand)) targ_pos cand_pos) (if targ_word (set! targ_pos (simple_pos (item.feat targ_word 'pos)))) (if cand_word (set! cand_pos (simple_pos (item.feat cand_word 'pos)))) ;(format t "targ_pos %l cand_pos %l\n" targ_pos cand_pos) (if (equal? targ_pos cand_pos) 0 1))) (define (score_contexts targ_context cand_context) "(score_contexts targ_context cand_context) If both context items are nil, then score is 0. If both context items are not nil, and are the same, then score is 0. Otherwise, score is 1." (if (and targ_context cand_context) (if (equal? (item.feat targ_context "name") (item.feat cand_context "name")) 0 1) (if (and (equal? targ_context nil) (equal? cand_context nil)) 0 1))) (define (tc_left_context targ cand) "(tc_left_context targ cand) Score left phonetic context." (let ((targ_context (item.prev targ)) (cand_context (item.prev cand))) (score_contexts targ_context cand_context))) ;; ;; tc_right_context ;; ;; ;; (define (tc_right_context targ cand) "(tc_right_context targ cand) Score right phonetic context." (let ((targ_context (item.next (item.next targ))) (cand_context (item.next (item.next cand)))) (score_contexts targ_context cand_context))) ;; ;; tc_segment_score ;; ;; This currently thresholds based on looking at the distributions of the scores. ;; A nice exp function may be better. (define (tc_segment_score targ cand) "tc_segment_score targ cand) A bad alignment score make a bad segment." (let ((score 0)) (if (not (phone_is_silence (item.feat cand "name"))) (set! score (+ score (item.feat cand 'score)))) (if (not (phone_is_silence (item.feat (item.next cand) "name"))) (set! score (+ score (item.feat (item.next cand) 'score)))) (cond ((> score -4000) ;2000 (x2) is 7.5% 0) ((> score -5000) ;2500 (x2) is 5.0% 0.5) (t 1)))) ;; ;; tc_bad_duration ;; ;; If the segment is marked as having a weird duration penalise it. ;; We allow bad_dur to be set on the target so resynthesis works ;; and so you could ask for really long/short segments. ;; (define (tc_bad_duration targ cand) (if (equal? (item.feat targ "bad_dur") (item.feat cand "bad_dur")) 0 1)) ;; ;; tc_bad_f0 ;; ;; If the candidate is deemed to have an inappropriate f0, then penalise it. ;; ;; Specifically, if the targ/cand segment type is expected to be voiced, then ;; an f0 of zero is bad (results from poor pitch tracking). In such a case, ;; the join cost would then favour other units with f0 (since the euclidean ;; distance between two zeros is very small ;) ;; We want to avoid that. ;; ;; Presumeably, we also want to penalise cases where supposedly voiceless ;; candidates have an f0 != 0 (either a consequence of bad pitch tracking ;; or bad labelling) but that's not done here yet... ;; ;; (the function itself has been implemented in C for convenience, and ;; this stub is left here just for this note ;) (define (tc_bad_f0 targ cand) (let ((score (temp_tc_bad_f0 targ cand)) (name (format nil "%s_%s" (item.feat targ "name") (item.feat (item.next targ) "name")))) (if (not (equal? score 0.0)) (format t "f0 score for %s is %f\n" name score)) score)) ;; ;; Is a segment a vowel? ( ph_is_a_vowel doesn't seem to work) ;; (define (my_is_vowel seg) (if seg (if (equal? (item.feat seg 'ph_vc) "+") t nil))) ;; get the syllable from sysstructure in normal utterance ;; (define (get_syl seg) (let (syl) (if seg (set! syl (item.relation.parent seg 'SylStructure))) syl)) ;; get the word from sylstructure in normal utterance ;; (define (get_word seg) (let ((syl (get_syl seg)) word) (if syl (set! word (item.parent syl))) word)) ;; simple pos ;; (define (simple_pos pos) (let (spos) (cond ((member_string pos '(vbd vb vbn vbz vbp vbg)) (set! spos "v")) ((member_string pos '(nn nnp nns nnps fw sym ls)) (set! spos "n")) ((member_string pos '(dt gin prp cc of to cd md pos wdt wp wrb ex uh pdt)) (set! spos "func")) ((member_string pos '(jj jjr jjs 1 2 rb rp rbr rbs)) (set! spos "other"))) spos)) ;; debugging (define (test_target_cost utt1 utt2) (let ((segs1 (utt.relation.items utt1 'Segment)) (segs2 (utt.relation.items utt2 'Segment)) (tc 0)) (while (and segs1 segs2) (set! tc (Default_Target_Cost (car segs1) (car segs2))) (format t "targ: %l cand: %l cost: %l\n" (item.name (car segs1)) (item.name (car segs2)) tc) (set! segs1 (cdr segs1)) (set! segs2 (cdr segs2))))) (provide 'target_cost) festival/lib/multisyn/send_xwaves.scm0000644000175000017500000002460510065567162016502 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 2003, 2004 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; xwaves interface for festival for multisyn (Rob Clark) ;;; ;;; This is never loaded by defualt. ;;; You'd need to change the paths here for this to currently work outside of CSTR. ;;; If anyone else ends up using it let me know and I'll make it more robust. ;;; ;; Send commands to xwaves (defvar send_xwaves_command "/cstr/linux/entropic/esps531.linux/bin/send_xwaves") (defvar spectrogram_command "/cstr/linux/entropic/esps531.linux/bin/sgram") (defvar data_path "/projects/cougar/data/cstr/nina") (set! xw_object_count 0) (set! xw_active_list nil) ;; ;; Display a synthesised utterance ;; (define (xwaves_display_utterance utt) "(xwaves_display_utterance utt) Display join and target information for an utterance." (let ((units (utt.relation.items utt 'Unit)) (object (xw_name_object)) wavfile specfile segfile diphfile joinfile targfile sourcefile timefile) (set! wavfile (xw_make_tmp_filename object)) (set! specfile (xw_make_tmp_filename object)) (set! segfile (xw_make_tmp_filename object)) (set! diphfile (xw_make_tmp_filename object)) (set! joinfile (xw_make_tmp_filename object)) (set! targfile (xw_make_tmp_filename object)) (set! sourcefile (xw_make_tmp_filename object)) (set! timefile (xw_make_tmp_filename object)) ; display resulting waveform (utt.save.wave utt wavfile 'riff) (xwaves_show_general object wavfile 1500 200 10 10) ; display resulting spectrogram (xw_genspec wavfile specfile) (xwaves_show_general object specfile 1500 400 10 260) ; segments (utt.save.unit_selection_segs utt segfile) (xwaves_show_labels object segfile specfile) ; Unit information (utt.save.unit_selection_info utt diphfile joinfile targfile sourcefile timefile) (xwaves_show_labels object timefile specfile) (xwaves_show_labels object sourcefile specfile) (xwaves_show_labels object targfile specfile) (xwaves_show_labels object joinfile specfile) (xwaves_show_labels object diphfile specfile) ; mark files (xw_register_active object (list wavfile specfile segfile diphfile joinfile sourcefile timefile)) )) ;; ;; Edit a diphone source ;; (define (xwaves_edit_diphone utt id) "(xwaves_edit_diphone utt id) Access the source diphone for label correction." (let ((diphone nil) segfilename wavefilename (utt (Utterance Text nil)) segs (seg nil) (start 0) end) ;; find unit. (mapcar (lambda (unit) (if (string-equal (format nil "_%s" id) (item.feat unit id)) (set! diphone unit))) (utt.relation.items utt 'Unit)) (if (null diphone) (error (format nil "Diphone with id _%s not found in utterance."))) (set! uttname (item.feat diphone "source_utt")) (set! end (item.feat diphone "source_end")) (set! segfilename (format nil "%s/lab/%s.lab" data_path uttname)) (set! wavefilename (format nil "%s/wav/%s.wav" data_path uttname)) (utt.relation.load utt 'Segment segfilename) (set! segs (utt.relation.items utt 'Segment)) (while (and segs (not (equal? (item.feat (car segs) "end") end))) (set! segs (cdr segs))) if null seg ... (if (item.prev diphone) (set! start (item.feat seg "start")) (set! start 0)) )) ;; ;; Interface with xwaves. ;; (define (xwaves_show_general object file width height xpos ypos) "(xwaves_show_general object file width height xpos ypos) Display an wave or track file." (xw_send (format nil "make name %s file %s width %d height %d loc_x %d loc_y %d" object file width height xpos ypos))) (define (xwaves_show_wave object file) "(xwaves_show_wave object file) Display a waveform." (xwaves_show_general object file 1500 200 10 10)) (define (xwaves_show_labels object file attachto) "(xwaves_show_labels object file attachto) Display a label file." (xw_send (format nil "send make signal %s name %s file %s color 125" attachto object file)) (xw_send "send activate fields 1 2 3 4 5")) (define (xwaves_attach_xlabel) "(xwaves_attach_xlabel) Attach xlabel to xwaves." (xw_send "attach function xlabel")) (define (xwaves_set_markers object left right) "(xwaves_set_markers object left right) Set the markers." (xw_send (format nil "%s set l_marker_time %f" object left)) (xw_send (format nil "%s set r_marker_time %f" object right))) (define (xwaves_bracket_markers object file) "(xwaves_bracket_markers object file) Bracket markers." (xw_send (format nil "%s bracket file %s " object file))) (define (xwaves_close_windows object) "(xwaves_close_windows object) Close currently open windows related to object or all if nil.." (cond ((null object) (xw_send "kill")) (t (xw_send (format nil "kill name %s" object)))) (xw_clear_active_list object)) (define (xwaves_wait) "(xwaves_wait) Wait for xwaves continue signal." (xw_send "pause")) ;; ;; Object naming ;; (define (xw_name_object) "(xw_name_object) Generate a name for this object." (let (name) (set! name (string-append "obj" xw_object_count)) (set! xw_object_count (+ xw_object_count 1)) name)) ;; ;; Temp file lists ;; (define (xw_clear_active_list object) "(xw_clear_active_list) Clear active list of specified object, or all if nil." (let (new_active_list) (mapcar (lambda (objectlist) (cond ((or (null object) (string-equal object (car objectlist))) (mapcar (lambda (file) (delete-file file)) (cadr objectlist))) (t (set! new_active_list (cons objectlist new_active_list))))) xw_active_list) (set! xw_active_list new_active_list)) nil) (define (xw_register_active object flist) "(xw_register_active object flist) Adds an object and its filenames to the active list." (set! xw_active_list (cons (cons object (list flist)) xw_active_list)) nil) (define (xw_make_tmp_filename object) "(xw_make_tmp_filename) make tmp file name which incorporates object name." (format nil "%s_%s" (make_tmp_filename) object)) ;; ;; Low level xwaves stuff. ;; (define (xw_genspec wavfile specfile) "(xw_genspec wavfile specfile) Generate a spectrogram file." (system (format nil "%s -dHAMMING -o8 -E0.94 -S2 -w8 %s %s\n" spectrogram_command wavfile specfile))) (define (xw_send command) "(xw_send command) Send a command to xwaves." (system (format nil "%s %s\n" send_xwaves_command command))) ;; ;; General Festival stuff. ;; (define (utt.save.unit_selection_segs utt filename) "(utt.save.unit_selection_segs utt filename) Save unit selection segments of UTT in a FILE in xlabel format." (let ((fd (fopen filename "w"))) (format fd "#\n") (mapcar (lambda (info) (format fd "%2.4f 100 %s\n" (car info) (car (cdr info)))) (utt.features utt 'Segment '(source_end name))) (fclose fd) utt)) (define (utt.save.unit_selection_info utt diphfile joinfile targfile sourcefile timefile) "(utt.save.unit_selection_info utt diphfile joinfile targfile sourcefile timefile) Save stuff in xlabel format." (let ((fdd (fopen diphfile "w")) (fdj (fopen joinfile "w")) (fdt (fopen targfile "w")) (fds (fopen sourcefile "w")) (fdx (fopen timefile "w")) real_join) (format fdd "#\n") (format fdj "#\n") (format fdt "#\n") (format fds "#\n") (format fdx "#\n") (mapcar (lambda (unit) (set! real_join "") (if (item.next unit) (if (not (string-equal (item.feat unit 'source_utt) (item.feat (item.next unit) 'source_utt))) (set! real_join "*"))) (format fdd "%2.4f 100 %s %s\n" (item.feat unit 'end) (item.feat unit 'name) real_join) (format fdj "%2.4f 100 %s\n" (item.feat unit 'end) (if (item.next unit) (item.feat (item.next unit) 'join_cost) 0)) (format fdt "%2.4f 100 %s\n" (item.feat unit 'end) (item.feat unit 'target_cost)) (format fds "%2.4f 100 %s\n" (item.feat unit 'end) (item.feat unit 'source_utt)) (format fdx "%2.4f 100 %s\n" (item.feat unit 'end) (item.feat unit 'source_end))) (utt.relation.items utt 'Unit)) (fclose fdd) (fclose fdj) (fclose fdt) (fclose fds) (fclose fdx) utt)) festival/examples/Makefile0000644000175000017500000000751312133505611014260 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=.. DIRNAME=examples BUILD_DIRS= ALL_DIRS=$(BUILD_DIRS) songs EXTEXTS = intro.text spintro.text benchmark.text EXAMPLES = webdemo.scm ex1.email ex1.ogi example.sable example2.sable \ tobi.stml example.th example.apml SCRIPTS = saytime.sh text2pos.sh latest.sh \ scfg_parse_text.sh text2wave.sh make_utts.sh dumpfeats.sh \ durmeanstd.sh powmeanstd.sh run-festival-script.sh text2utt.sh SHELL_SCRIPTS = benchmark festival_client.pl SCMS = toksearch.scm th-mode.scm addr-mode.scm SAMPLEC = festival_client.c festival_client.h OTHERS = apml.dtd FILES=Makefile $(EXAMPLES) $(EXTEXTS) $(SCRIPTS) $(SHELL_SCRIPTS) $(SCMS) $(SAMPLEC) $(OTHERS) speech_pm_1.0.tar ALL = $(SCRIPTS:.sh=) LOCAL_CLEAN = $(SCRIPTS:.sh=) include $(TOP)/config/common_make_rules $(ALL) : % : %.sh rm -f $@ @echo "#!/bin/sh" >$@ @echo "\"true\" ; exec "$(FESTIVAL_HOME)/bin/festival --script '$$0 $$*' >>$@ cat $< >>$@ chmod +x $@ festival_client: festival_client.o festival_client.h $(LINK_COMMAND) -o festival_client festival_client.o $(LIBS) festival_client.o: festival_client.c festival_client.h $(CC_COMMAND) -DSTANDALONE festival_client.c -o festival_client.o # Do this manually to make sure Festival.tar (perl module) is up to date. speech_pm_1.0.tar: $(wildcard speech_pm_1.0/*) $(wildcard speech_pm_1.0/*/*) $(wildcard speech_pm_1.0/*/*/*) -chmod +w speech_pm_1.0.tar tar cvf speech_pm_1.0.tar `cat speech_pm_1.0/MANIFEST|sed -e 's/^/speech_pm_1.0\//'` festival/examples/webdemo.scm0000644000175000017500000000572507262605650014764 0ustar awbawb;;; ;;; Sentences presynthesized on demo web page ;;; (set! utt1 (Utterance Text " This is a short introduction to the Festival Speech Synthesis System. Festival was developed by Alan Black and Paul Taylor, at the Centre for Speech Technology Research, University of Edinburgh. ")) (set! utt2 (Utterance Text " Festival currently uses a diphone synthesizer, both residual excited LPC and PSOLA methods are supported. The upper levels, duration and intonation, are generated from statistically trained models, built from databases of natural speech. The architecture of the system is designed to be flexible, including various tools, which allow new modules to be added easily. ")) (define (make_waves) "Synthesize the two examples and save them in the desired formats" (Synth utt1) (Parameter.set 'Wavefiletype 'riff) (utt.save.wave utt1 "intro.wav") (Parameter.set 'Wavefiletype 'ulaw) (utt.save.wave utt1 "intro.au") (Synth utt2) (Parameter.set 'Wavefiletype 'riff) (utt.save.wave utt2 "intro2.wav") (Parameter.set 'Wavefiletype 'ulaw) (utt.save.wave utt2 "intro2.au") ) (set! welsh1 (Utterance Text "Dwi'n gallu llefaru pob llinell heb atal, oherwydd does dim tafod gyda fi.")) (define (make_welsh) (voice_welsh_hl) (Synth welsh1) (Parameter.set 'Wavefiletype 'riff) (utt.save.wave welsh1 "welsh1.wav") (Parameter.set 'Wavefiletype 'ulaw) (utt.save.wave welsh1 "welsh1.au")) (set! spanish1 (Utterance Text "m'uchos 'a~nos despu'es, fr'ente al pelot'on de fusilami'ento, el coron'el aureli'ano buend'ia hab'ia de record'ar de aqu'el d'ia lej'ano, en que su p'adre lo llev'o a conoc'er el hi'elo.")) (define (make_spanish) (voice_spanish_el) (Synth spanish1) (Parameter.set 'Wavefiletype 'riff) (utt.save.wave spanish1 "spanish1.wav") (Parameter.set 'Wavefiletype 'ulaw) (utt.save.wave spanish1 "spanish1.au")) (set! utt_pos (Utterance Text "My cat who lives dangerously had nine lives. ")) (set! utt_Bdi (Utterance Text "He wanted to go for a drive in.")) (set! utt_Bditc (Utterance Text "He wanted to go for a drive in the country.")) (define (make_others) (Synth utt_pos) (Synth utt_Bdi) (Synth utt_Bditc) (Parameter.set 'Wavefiletype 'riff) (utt.save.wave utt_pos "cat.wav") (utt.save.wave utt_Bdi "Bdi.wav") (utt.save.wave utt_Bditc "Bditc.wav") (Parameter.set 'Wavefiletype 'ulaw) (utt.save.wave utt_pos "cat.au") (utt.save.wave utt_Bdi "Bdi.au") (utt.save.wave utt_Bditc "Bditc.au")) (set! utt_diph (Utterance Text "This is a short introduction to the Festival Speech Synthesis System.")) (set! utt_sucs (Utterance Text "This is a short introduction to the Festival Speech Synthesis System.")) (define (make_diphsbs) (Synth utt_diph) (Parameter.set 'Wavefiletype 'riff) (utt.save.wave utt_diph "diph1.wav") (Parameter.set 'Wavefiletype 'ulaw) (utt.save.wave utt_diph "diph1.au") (voice_gsw_450) (Synth utt_sucs) (Parameter.set 'Wavefiletype 'riff) (utt.save.wave utt_sucs "sbs1.wav") (Parameter.set 'Wavefiletype 'ulaw) (utt.save.wave utt_sucs "sbs1.au")) festival/examples/ex1.email0000644000175000017500000000214307262605650014333 0ustar awbawbFrom VM Wed Nov 27 15:33:27 1996 Content-Length: 482 Return-Path: awb@cstr.ed.ac.uk Received: from margo (margo.cstr.ed.ac.uk [192.41.114.73]) by liddell.cstr.ed.ac.uk (8.6.13/8.6.10) with ESMTP id PAA08491 for ; Wed, 27 Nov 1996 15:32:57 GMT Received: (awb@localhost) by margo (SMI-8.6/8.6.9) id PAA00818; Wed, 27 Nov 1996 15:32:54 GMT Message-Id: <199611271532.PAA00818@margo> In-Reply-To: <199611271531.PAA00812@margo> References: <199611271531.PAA00812@margo> From: Alan W Black To: Alan W Black Subject: Example mail message Date: Wed, 27 Nov 1996 15:32:54 GMT Alan W. Black writes on 27 November 1996: > > > I'm looking for a demo mail message for Festival, but can't seem to > find any suitable. It should at least have some quoted text, and > have some interesting tokens like a URL or such like. > > Alan Well I'm not sure exactly what you mean but awb@cogsci.ed.ac.uk has an interesting home page at http://www.cstr.ed.ac.uk/~awb/ which might be what you're looking for. Alan > PS. Will you attend the course? I hope so bye for now festival/examples/ex1.ogi0000644000175000017500000000072311363106376014022 0ustar awbawb This is an example. This is a slow example. This is a very slow example. a normal one and a fast talking example. Maybe this one is too fast. My name is Mike, . My telephone number is Mike here. Chris here. This is Gordon. I'm most terribly sorry to interrupt at this time, but my name is Roger. A good day to you. festival/examples/example.sable0000644000175000017500000000405307262605650015272 0ustar awbawb My telephone number is festival/examples/example2.sable0000644000175000017500000000135507262605650015356 0ustar awbawb Homographs are words that are written the same but have different pronunciations, such as lives and lives. You say either, while I say either. We can say things fast. and slowly. And then at normal speed. festival/examples/tobi.stml0000644000175000017500000000137307262605650014467 0ustar awbawb I to go on a . wanted to go on a , but wanted to go festival/examples/example.th0000644000175000017500000000030507262605650014613 0ustar awbawb*smile* Good morning. This is an example of Festival using a talking head. This example uses a Festival text mode *frown* which can be difficult to understand, *smile* but it does help me talk. festival/examples/example.apml0000644000175000017500000000655707611241332015137 0ustar awbawb Good morning Mr Smith I'm sorry to tell you that you have been diagnosed as suffering from a mild form of what we call angina pectoris. This is a spasm of the chest , resulting from overexertion when the heart is diseased To solve this problem , there are two drugs I would like you to take. The first one is Aspirin, which is an analgesic. that is, it relieves the pain. I have prescribed it to cure your angina The only problem is that this drug can be associated with some sideeffects . festival/examples/intro.text0000644000175000017500000000030407262605650014663 0ustar awbawb This is a short introduction to the Festival Speech Synthesis System. Festival was developed by Alan Black and Paul Taylor, at the Centre for Speech Technology Research, University of Edinburgh. festival/examples/spintro.text0000644000175000017500000000024107262605650015226 0ustar awbawb Bienvenido a Festival, nuestro conversor de texto a voz. Festival es un conversor multilenguaje de texto a voz, desarrollado en la Universidad de Edimburgo. festival/examples/benchmark.text0000644000175000017500000000523207262605650015467 0ustar awbawb FESTIVAL(1) User Commands FESTIVAL(1) NAME festival - a text-to-speech system. SYNOPSIS festival [options] [file0] [file1] ... DESCRIPTION Festival is a general purpose text-to-speech system. As well as simply rendering text as speech it can be used in an interactive command mode for testing and developing various aspects of speech synthesis technology. Festival has two major modes, command and tts (text-to- speech). When in command mode input (from file or interac- tively) is interpreted by the command interpreter. When in tts mode input is rendered as speech. OPTIONS -h print help information -q Start without loading any initialization files. --libdir PATH Specify alternate to default library directory (used in initializing the variable load-path, and for loading most intinialisation files) -b or --batch Run in batch mode. In batch mode no input is read from standard input -i or --interactive Run in interactive mode. In intteractive mode input (commands or text) is read from standard input. When in command mode (the default) a readline based Lisp read-eval-print command interpreter is presented. --server Run in server mode. Any file arguments are loaded 9and interpreted, before going into server mode. In server mode Festival waits for clients on port 1314 (by default). Note server mode can give unauthor- ised access to your machine, please read the section in the manual entitled Server/client API before using this mode. --tts Run in tts mode. All files argument are treated as text files to be said. Unless interactive mode SunOS 5.5.1 Last change: 4th Feb 1996 1 FESTIVAL(1) User Commands FESTIVAL(1) is explicitly specified no input is read from stan- dard input, unless no files are specified. --language LANG Where LANG is one of english (default) spanish or welsh. Select that language for basic operation, command or tts. This may be change during a session with the command select_language. -v Display version number and exit. BUGS More than you can imagine. A manual with much detail (though not complete) is available in info (or html) format. AUTHOR Alan W Black and Paul Taylor (C) Centre for Speech Technology Research University of Edinburgh 80 South Bridge Edinburgh EH1 1HN SunOS 5.5.1 Last change: 4th Feb 1996 2 festival/examples/saytime.sh0000644000175000017500000001331707262605650014641 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;-*-mode:scheme-*- ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: wasting time one August morning in 1996 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Here is a short example of a Festival program that speaks the ;;; current time. It uses UNIX date to get the time then builds ;;; a string with an expression of the current time. ;;; ;;; The string generated for synthesis is of the form ;;; The time is now ;;; ;;; Because this is a --script type file I has to explicitly ;;; load the initfiles: init.scm and user's .festivalrc (load (path-append libdir "init.scm")) (define (get-the-time) "Returns a list of hour and minute and second, for later processing" (let (date) (system "date | awk '{print $4}' | tr : ' ' >/tmp/saytime.tmp") (set! date (load "/tmp/saytime.tmp" t)) ;; loads the file unevaluated (system "rm /tmp/saytime.tmp") date) ) (define (round-up-time time) "Rounds time up/down to nearest five minute interval" (let ((hour (car time)) (min (car (cdr time))) (sec (car (cdr (cdr time))))) (set! min (round-min (+ 2 min))) (list hour min sec))) (define (round-min min) "Returns minutes rounded down to nearest 5 minute interval" (cond ((< min 5) 0) (t (+ 5 (round-min (- min 5)))))) (define (approx time) "Returns a string stating the approximation of the time. exactly -- within a minute either side almost -- 1-2 minutes before just after - 1-2 minutes after a little after 2-3 minutes after " (let ((rm (round-min (car (cdr time)))) (min (car (cdr time)))) (cond ((or (< (- min rm) 1) (> (- min rm) 3)) "exactly ") ((< (- min rm) 2) "just after ") ((< (- min rm) 3) "a little after ") (t "almost ")))) (define (hour-string time) "Return description of hour" (let ((hour (car time))) (if (> (car (cdr time)) 30) (set! hour (+ 1 hour))) (cond ((or (eq hour 0) (eq hour 24)) "midnight ") ((> hour 12) (string-append (- hour 12) ", ")) (t (string-append hour ", "))))) (define (minute-string time) "Return description of minute" (let ((min (car (cdr time)))) (cond ((or (eq min 0) (eq min 60)) " ") ((eq min 5) "five past ") ((eq min 10) "ten past ") ((eq min 15) "quarter past ") ((eq min 20) "twenty past ") ((eq min 25) "twenty-five past ") ((eq min 30) "half past ") ((eq min 35) "twenty-five to ") ((eq min 40) "twenty to ") ((eq min 45) "quarter to ") ((eq min 50) "ten to ") ((eq min 55) "five to ") (t "something else ")))) (define (ampm-string time) "Return morning/afternoon or evening string" (let ((hour (car time))) (cond ((or (eq hour 0) (eq hour 12) (eq hour 24)) " ") ((< hour 12) "in the morning. ") ((< hour 18) "in the afternoon. ") (t "in the evening. ")))) ;;; ;;; Now with all the functions defined we can get the time ;;; (set! actual-time (get-the-time)) (set! round-time (round-up-time actual-time)) ;;; Construct the time expression (set! time-string (string-append "The time is now, " (approx actual-time) (minute-string round-time) (hour-string round-time) (ampm-string round-time))) (format t "%s\n" time-string) ;;; Synthesize it (SayText time-string) festival/examples/text2pos.sh0000644000175000017500000000743210253404655014753 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;-*-mode:scheme-*- ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: August 1996 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Reads in text from stdin and outputs text/pos on stdout ;;; ;;; Designed to show how simple filters can be written in Festival ;;; ;;; First we defined a function that processes an utterance enough ;;; to predict part of speech, namely, tokenize it, find the words ;;; and then run the POS tagger on it. ;;; Then we define a function to extract the word and pos tag itself ;;; ;;; We redefine the basic functions run on utterances during text to ;;; speech to be our two newly-defined function and then simply ;;; run tts on standard input. ;;; ;;; Because this is a --script type file I has to explicitly ;;; load the initfiles: init.scm and user's .festivalrc (load (path-append libdir "init.scm")) (define (find-pos utt) "Main function for processing TTS utterances. Predicts POS and prints words with their POS" (Token utt) (POS utt) ) (define (output-pos utt) "Output the word/pos for each word in utt" (mapcar (lambda (word) (format t "%l/%l\n" (item.feat word "name") (item.feat word "pos"))) (utt.relation.items utt 'Word))) ;;; ;;; Redefine what happens to utterances during text to speech ;;; (set! tts_hooks (list find-pos output-pos)) ;;; Stop those GC messages (gc-status nil) ;;; Do the work (tts_file "-") festival/examples/latest.sh0000644000175000017500000001177107262605650014464 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;-*-mode:scheme-*- ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: September 1996 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Gets the a summary of the latest news from Pathfinder.com (Time ;;; magazine and synthesizes it). My first web based speech application. ;;; ;;; This is far too dependent on Time's latest news pages format and ;;; not really very general, but its a start. Also they seem to change ;;; both the format and the name of the pages regularly so this probably ;;; no longer works. ;;; ;;; Note the news in Copyright Reuters, and should not be used except ;;; for personal use. This program can be viewed simply as a web ;;; browser (for one particular page) and does not itself contain any ;;; information under Time or Reuter copyright. ;;; ;;; Because this is a --script type file I has to explicitly ;;; load the initfiles: init.scm and user's .festivalrc (load (path-append libdir "init.scm")) (audio_mode 'async) ;; play waves while continuing synthesis ;;; Give short introduction, so something can happen while we're ;;; getting the news (SayText "And here is the news.") (SayText "News stories are courtesy of Time Warners' Path finder Magazine and Reuters News Media.") (format t "Getting news from Pathfinder Magazine ... \n") (fflush nil) ;;; First get the page (set! tmpfile (make_tmp_filename)) (set! tmpfile2 (string-append tmpfile "_2")) (get_url "http://www.pathfinder.com/news/latest" tmpfile) (format t "done\n") ;; This has to be powerful awk, not the original awk. GNU awk or nawk ;; are what I'm looking for, but they have such random names, and may or ;; may not be on your system. (if (string-matches *ostype* ".*Linux.*") (defvar GOOD_AWK "awk") (defvar GOOD_AWK "nawk")) ;; Should now use some HTML to SSML conversion but hack it just now (system (string-append GOOD_AWK " '{ if ($1 == \"
    \") inlist = 1; if (inlist == 1) { if ($1 == \"
    \") # title { getline # skip href getline line = $0 sub(/^.*/,\"\",line); sub(/ *<.b>.*$/,\"\",line); printf(\"%s, \",line); } else if ($1 == \"
    \") # summary { getline line = $0 sub(/\(.. ... .... ..:.. ...\)/,\"\",line) # remove time stamp printf(\"%s\\n\\n\",line); } else if ($1 == \"
    \") inlist = 0; } }' < " tmpfile " > " tmpfile2)) ;; Say the news (tts_file tmpfile2 nil) (system (string-append "rm -f " tmpfile " " tmpfile2)) (audio_mode 'close) ;; close gracefully festival/examples/scfg_parse_text.sh0000644000175000017500000001271307262605650016345 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;-*-mode:scheme-*- ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: October 1997 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Parse arbitrary text using the given SCFG. ;;; ;;; Tokenizes given text file and runs the part of speech tagger on it ;;; Parses it with respect to a grammar trained from the UPenn WSJ ;;; tree bank (you may optionall specify a different grammar). ;;; ;;; This may be slow for long sentences as there is |w|^3 factor ;;; involved in parsing algorithm. ;;; Because this is a --script type file I has to explicitly ;;; load the initfiles: init.scm and user's .festivalrc (load (path-append libdir "init.scm")) (require 'scfg) ;;; Process command line arguments (define (scfg_parse_text_help) (format t "%s\n" "scfg_parse_text [options] textfile Parse arbitrary text Options -o ofile File to save parses (default is stdout). -grammar ifile Alternative grammar, default uses standard grammar from Festival distribution. -full_parse Output full parse with probabilities rather than simplified form (which is the default).") (quit)) ;;; No gc messages (gc-status nil) ;;; Default argument values (defvar grammarfile (path-append libdir "scfg_wsj_wp20.gram")) (defvar outfile "-") (defvar outfd t) (defvar parse_type 'brackets_only) (defvar text_files '("-")) ;;; Get options (define (get_options) (let ((files nil) (o argv)) (if (or (member_string "-h" argv) (member_string "-help" argv) (member_string "--help" argv) (member_string "-?" argv)) (scfg_parse_text_help)) (while o (begin (cond ((string-equal "-o" (car o)) (if (not (cdr o)) (scfg_error "no output file specified")) (set! outfile (car (cdr o))) (set! outfd (fopen outfile "w")) (set! o (cdr o))) ((string-equal "-grammar" (car o)) (if (not (cdr o)) (scfg_error "no grammar file specified")) (set! grammarfile (car (cdr o))) (set! o (cdr o))) ((string-equal "-full_parse" (car o)) (set! parse_type 'full_parse)) (t (set! files (cons (car o) files)))) (set! o (cdr o)))) (if files (set! text_files (reverse files))))) (define (scfg_error message) (format stderr "%s: %s\n" "scfg_parse_text" message) (scfg_parse_text_help)) ;;; Functions that do the work (define (find-parse utt) "Main function for processing TTS utterances. Tokenizes, predicts POS and then parses." (Token utt) (POS utt) (Phrasify utt) ;; cause it maps the POS tags (ProbParse utt) ) (define (output-parse utt) "Output the parse tree for each utt" (if (equal? parse_type 'brackets_only) (pprintf (scfg_simplify_relation_tree (utt.relation_tree utt 'Syntax)) outfd) (pprintf (utt.relation_tree utt 'Syntax) outfd)) (format outfd "\n") utt) ;;; ;;; Redefine what happens to utterances during text to speech ;;; (set! tts_hooks (list find-parse output-parse)) (define (main) (get_options) ;; Load the grammar (set! scfg_grammar (load grammarfile t)) ;; Parse the files (mapcar (lambda (f) (tts_file f)) text_files)) ;;; Do the work (main) festival/examples/text2wave.sh0000755000175000017500000001441007720777707015131 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;-*-mode:scheme-*- ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: November 1997 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Text to a single waveform like festival_client but without ;;; starting hte server ;;; ;;; Because this is a --script type file I has to explicitly ;;; load the initfiles: init.scm and user's .festivalrc (load (path-append libdir "init.scm")) ;;; Process command line arguments (define (text2wave_help) (format t "%s\n" "text2wave [options] textfile Convert a textfile to a waveform Options -mode Explicit tts mode. -o ofile File to save waveform (default is stdout). -otype Output waveform type: ulaw, snd, aiff, riff, nist etc. (default is riff) -F Output frequency. -scale Volume factor -eval File or lisp s-expression to be evaluated before synthesis. ") (quit)) ;;; No gc messages (gc-status nil) ;;; Default argument values (defvar outfile "-") (defvar output_type 'riff) (defvar frequency nil) ;; default is no frequency modification (defvar text_files '("-")) (defvar mode nil) (defvar volume "1.0") (defvar wavefiles nil) ;;; Get options (define (get_options) (let ((files nil) (o argv)) (if (or (member_string "-h" argv) (member_string "-help" argv) (member_string "--help" argv) (member_string "-?" argv)) (text2wave_help)) (while o (begin (cond ((string-equal "-o" (car o)) (if (not (cdr o)) (text2wave_error "no output file specified")) (set! outfile (car (cdr o))) (set! o (cdr o))) ((string-equal "-otype" (car o)) (if (not (cdr o)) (text2wave_error "no output filetype specified")) (set! output_type (car (cdr o))) (set! o (cdr o))) ((or (string-equal "-f" (car o)) ;; for compatibility and memory loss (string-equal "-F" (car o))) (if (not (cdr o)) (text2wave_error "no frequency specified")) (set! frequency (car (cdr o))) (set! o (cdr o))) ((string-equal "-scale" (car o)) (if (not (cdr o)) (text2wave_error "no scale specified")) (set! volume (car (cdr o))) (set! o (cdr o))) ((string-equal "-mode" (car o)) (if (not (cdr o)) (text2wave_error "no mode specified")) (set! mode (car (cdr o))) (set! o (cdr o))) ((string-equal "-eval" (car o)) (if (not (cdr o)) (text2wave_error "no file specified to load")) (if (string-matches (car (cdr o)) "^(.*") (eval (read-from-string (car (cdr o)))) (load (car (cdr o)))) (set! o (cdr o))) (t (set! files (cons (car o) files)))) (set! o (cdr o)))) (if files (set! text_files (reverse files))))) (define (text2wave_error message) (format stderr "%s: %s\n" "text2wave" message) (text2wave_help)) (define (save_record_wave utt) "Saves the waveform and records its so it can be joined into a a single waveform at the end." (let ((fn (make_tmp_filename))) (utt.save.wave utt fn) (set! wavefiles (cons fn wavefiles)) utt)) (define (combine_waves) "Join all the waves together into the desired output file and delete the intermediate ones." (let ((wholeutt (utt.synth (Utterance Text "")))) (mapcar (lambda (d) (utt.import.wave wholeutt d t) (delete-file d)) (reverse wavefiles)) (if frequency (utt.wave.resample wholeutt (parse-number frequency))) (if (not (equal? volume "1.0")) (begin (utt.wave.rescale wholeutt (parse-number volume)))) (utt.save.wave wholeutt outfile output_type) )) ;;; ;;; Redefine what happens to utterances during text to speech ;;; (set! tts_hooks (list utt.synth save_record_wave)) (define (main) (get_options) ;; do the synthesis (mapcar (lambda (f) (if mode (tts_file f mode) (tts_file f (tts_find_text_mode f auto-text-mode-alist)))) text_files) ;; Now put the waveforms together at again (combine_waves) ) ;;; Do the work (main) festival/examples/make_utts.sh0000644000175000017500000004605510045556464015170 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;-*-mode:scheme-*- ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: November 1997 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Build a utterance from a number of stream label files including ;;; building the links between the stream items ;;; ;;; This used to be a shell script but that was just soooo slow ;;; and inflexible it was better to do it in Festival ;;; ;;; Because this is a --script type file it has to explicitly ;;; load the initfiles: init.scm and user's .festivalrc (if (not (symbol-bound? 'caar)) (load (path-append libdir "init.scm"))) ;;; Some parts are potentially editable (defvar basic_relations '((Phrase segmental ()) (Word segmental (Phrase)) (Syllable segmental (Word) ) (Segment segmental (Syllable)) (IntEvent point (Syllable) ) (Target point (Segment)) ;; virtually unused ) "The basic relations that exist and need to be combined into a single utterance. Also their type, that is if they describe segments of the utterance or points in the utterance.") (require 'tilt) (define (make_utts_help) (format t "%s\n" "make_utts [options] festival/relations/Segment/*.Segment Build utterance forms for sets of stream_item labels Options -utt_dir Directory where utterances will be saved (default is festival/utts/) -label_dir The directory which contains subdirectories containing label files for each relation, default is festival/relations/ -style What style of utterances, classic or unisyn -tilt_events IntEvent files are tilt event so uses special information in the syllink feature to link to syllables. -eval Load in scheme file with run specific code, if file name starts with a left parent the string itsefl is interpreted -tokens Overly non-general method to load in Tokens and markup -pos Do part of speech assignment -phoneset Specify the phoneset name, this is required for -tilt_events. ") (quit)) ;;; Default options values (defvar seg_files nil) ;; files to build from (defvar label_dir "festival/relations/") (defvar style 'classic) (defvar tilt_events nil) (defvar with_tokens nil) (defvar unisyn_build_with_silences t) (defvar do_pos nil) (defvar do_syn nil) (defvar utt_dir "festival/utts/") ;; may be redefined by user (define (make_utts_user_function utt) utt) ;;; Get options (define (get_options) (let ((files nil) (o argv)) (if (or (member_string "-h" argv) (member_string "-help" argv) (member_string "--help" argv) (member_string "-?" argv)) (make_utts_help)) (while o (begin (cond ((string-equal "-label_dir" (car o)) (if (not (cdr o)) (make_utts_error "no label_dir file specified")) (set! label_dir (car (cdr o))) (set! o (cdr o))) ((string-equal "-utt_dir" (car o)) (if (not (cdr o)) (make_utts_error "no utt_dir file specified")) (set! utt_dir (car (cdr o))) (set! o (cdr o))) ((string-equal "-phoneset" (car o)) (if (not (cdr o)) (make_utts_error "no phoneset specified")) (load_library (string-append (car (cdr o)) "_phones.scm")) (set! o (cdr o))) ((string-equal "-eval" (car o)) (if (not (cdr o)) (make_utts_error "no file specified to load")) (if (string-matches (car (cdr o)) "^(.*") (eval (read-from-string (car (cdr o)))) (load (car (cdr o)))) (set! o (cdr o))) ((string-equal "-tilt_events" (car o)) (set! tilt_events t)) ((string-equal "-style" (car o)) (if (not (cdr o)) (make_utts_error "no style specified")) (set! style (car (cdr o))) (set! o (cdr o))) ((string-equal "-tokens" (car o)) (set! with_tokens t)) ((string-equal "-pos" (car o)) (set! do_pos t)) (t (set! files (cons (car o) files)))) (set! o (cdr o)))) (if files (set! seg_files (reverse files))))) (define (make_utts_error message) (format stderr "%s: %s\n" "make_utts" message) (make_utts_help)) ;;; No gc messages (gc-status nil) (define (make_utt name relations) "(make_utt dir name relations) Build an utterance from the stream_item label files in dir/RELATION/name.RELATION and return it. This also creates relations between each base relation." (let (utt) (cond ((equal? 'classic style) (set! utt (make_utt_classic name relations))) ((equal? 'unisyn style) (set! utt (make_utt_unisyn name relations))) (t (err "make_utts: unknown style" style))) (utt.set_feat utt "fileid" name) (if do_pos (find_pos utt)) (if do_syn (find_syn utt)) utt) ) ;; These should probably be somewhere else (defvar met_insertion 7) (defvar met_deletion 7) (defvar met_substitution 7) (define (make_utt_unisyn name relation) "(make_utt_classic name relations) Build utterance, with xml, metrical tree and tilt relations." (let ((utt (Utterance nil nil))) (set! utt (wsi_build utt name label_dir tilt_events)) (add_xml_relation utt (string-append label_dir "xml/rhet/" name ".xrhet")) (add_xml_relation utt (string-append label_dir "xml/syn/" name ".xsyn")) (add_xml_relation utt (string-append label_dir "xml/anaph/" name ".xana")) (add_xml_relation utt (string-append label_dir "xml/info/" name ".xinf")) (syntax_to_metrical_words utt) (extend_tree utt (list 'MetricalTree 'MetricalWord 'WordStructure 'Syllable)) (extend_tree utt (list 'ProsodicTree 'MetricalTree 'SylStructure 'Segment)) (add_match_features utt) utt) ) (define (wsi_build utt file pathname do_il) (add_trans_word utt (string-append pathname "wrd/" file ".wrd")) (add_trans_segment utt (string-append pathname "lab/" file ".lab")) (if do_il (add_trans_intonation utt (string-append pathname "tilt/" file ".tilt")) nil ) utt ) (define (make_utt_classic name relations) "(make_utt_classic name relations) Build utterance in classic style for name and named relations." (let (utt) (if with_tokens (set! utt (utt.load nil (string-append label_dir "/" "Token" "/" name ".Token"))) (set! utt (Utterance Text nil))) (set! current_utt utt) (mapcar (lambda (s) (utt.relation.load utt (car s) (string-append label_dir "/" (car s) "/" name "." (car s)))) relations) ;; Now link them (make_syl_structure utt) (make_target_links utt) (make_phrase_structure utt) (if tilt_events (tilt_link_syls utt) (intevent_link_syls utt)) (if with_tokens (relate_tokens_to_words utt)) (make_utts_user_function utt) utt) ) (define (find_pos utt) "(find pos utt) Assign part of speech using standard POS tagger. Also produce standard reduced tagset in phr_pos. Irrelevant extra POS features are removed. This assumes a POS tagger is set up at this point, this can most easily be done be setting up a relevant voice." (POS utt) (mapcar (lambda (w) (item.set_feat w "phr_pos" (map_pos (item.feat w "pos") english_pos_map_wp39_to_wp20)) (item.remove_feature w "pos_index") (item.remove_feature w "pos_index_score") ) (utt.relation.items utt 'Word)) utt) (define (map_pos pos map) (cond ((null map) pos) ((member_string pos (car (car map))) (car (cdr (car map)))) (t (map_pos pos (cdr map))))) (define (make_target_links utt) "(make_target_links utt) Make targets that fall within a segment. Targets contains all segments and have that actual Targets as daughters." (let ((targets (utt.relation.items utt 'Target)) (segs (utt.relation.items utt 'Segment)) tt1) (utt.relation.create utt 'TTarget) (mapcar (lambda (tt) (set! tt1 (utt.relation.append utt 'TTarget)) ;; covert the target values to the newer naming convention (item.set_feat tt1 "pos" (item.feat tt "end")) (item.set_feat tt1 "f0" (parse-number (item.feat tt "name"))) (item.relation.remove tt 'Target)) targets) (set! targets (utt.relation.items utt 'TTarget)) (set! TARGSEGFACTOR 0.010) (while segs (utt.relation.append utt 'Target (car segs)) (while (and targets (< (item.feat (car targets) "pos") (+ (item.feat (car segs) "end") TARGSEGFACTOR))) (item.relation.append_daughter (car segs) 'Target (car targets)) (set! targets (cdr targets))) (set! segs (cdr segs))) (utt.relation.delete utt 'TTarget) )) (define (make_phrase_structure utt) "(make_phrase_structure utt) Add words into phrases." (let ((phrases (utt.relation.items utt 'Phrase)) (words (utt.relation.items utt 'Word))) (set! WORDPHRASEFACTOR 0.200) (while phrases (while (and words (< (item.feat (car words) 'end) (+ (item.feat (car phrases) 'end) WORDPHRASEFACTOR))) (item.relation.append_daughter (car phrases) 'Phrase (car words)) (set! words (cdr words))) (set! phrases (cdr phrases))))) (define (relate_tokens_to_words utt) "(relate_tokens_to_words utt) A specific function for aligning the token stream to word stream." (convert_token_stream utt) (let ((tokens (utt.relation.items utt 'Token)) (words (utt.relation.items utt 'Word))) (link_tokens_words tokens words) utt) ) (define (convert_token_stream utt) "(convert_token_stream utt) Replace Token Stream with Token relation. -- won't be needed when things are properly converted." (utt.relation.create utt 'Token) (mapcar (lambda (tok) (utt.relation.append utt 'Token tok)) (utt.stream utt 'Token)) (utt.stream.delete utt 'Token) ) (define (link_tokens_words tokens words) "(link_tokens_words tokens words) Advance through the tokens and words aligning them as required." (cond ((null words) t) ((null tokens) (error (format nil "Extra words: %l\n" (mapcar item.name words)))) ((or (string-equal "1" (item.feat (car tokens) "punct-elem")) (member_string (item.name (car tokens)) '("(" ")"))) (link_tokens_words (cdr tokens) words)) ((string-equal "SPEECH-OMITTED" (item.feat (car tokens) "R:SOLEML.parent.TYPE")) (link_tokens_words (cdr tokens) words)) ((and (string-matches (item.name (car words)) ".*'.*") (string-equal "APOSTROPHE" (item.feat (car tokens) "R:Token.n.TYPE"))) (item.relation.append_daughter (car tokens) 'Token (car words)) (item.relation.append_daughter (car (cdr tokens)) 'Token (car words)) (if (string-matches (item.name (car words)) ".*'") (link_tokens_words (cdr (cdr tokens)) (cdr words)) (begin (item.relation.append_daughter (car (cdr (cdr tokens))) 'Token (car words)) (link_tokens_words (cdr (cdr (cdr tokens))) (cdr words))))) ((string-equal (downcase (item.name (car tokens))) (downcase (item.name (car words)))) (item.relation.append_daughter (car tokens) 'Token (car words)) (link_tokens_words (cdr tokens) (cdr words))) ;; there going to be more here !!! (t (error (format nil "Mismatch of tokens and words \n %l\n %l\n" (mapcar item.name tokens) (mapcar item.name words)))))) (define (do_utt name) (let ((utt (make_utt name basic_relations))) (utt.save utt (string-append utt_dir "/" name ".utt") 'est_ascii) t)) (define (make_syl_structure utt) "(make_syl_structure utt) Make SylStructure relation linking Words, Syllables and Segments." (let ((words (utt.relation.items utt 'Word)) (syls (utt.relation.items utt 'Syllable)) (segs (utt.relation.items utt 'Segment))) (set! SYLWORDFACTOR 0.025) (set! SEGSYLFACTOR 0.02) (utt.relation.create utt 'SylStructure) (while words (utt.relation.append utt 'SylStructure (car words)) (while (and syls (< (item.feat (car syls) 'end) (+ (item.feat (car words) 'end) SYLWORDFACTOR))) (item.relation.append_daughter (car words) 'SylStructure (car syls)) (while (and segs (< (item.feat (car segs) 'end) (+ (item.feat (car syls) 'end) SEGSYLFACTOR))) (if (not (phone_is_silence (item.name (car segs)))) (item.relation.append_daughter (car syls) 'SylStructure (car segs))) (set! segs (cdr segs))) (set! syls (cdr syls))) (set! words (cdr words))))) (define (tilt_link_syls utt) "(tilt_link_syls utt) Link syls to IntEvents, for Tilt. In this case the feature syllink specifies the word.sylnum that the event should be linked to." (let ((syls (utt.relation.items utt 'Syllable))) (utt.relation.create utt 'Intonation) (mapcar (lambda (ie) (let ((name (item.name ie)) (syllink (item.feat ie "syllink")) syl) (cond ((member_string name '("phrase_start" "phrase_end")) ;; relate this IntEvent to silence segment ; (if (string-equal name "phrase_start") ; (set! syl (find_ie_phrase_syl utt ie 'syllable_start)) ; (set! syl (find_ie_phrase_syl utt ie 'syllable_end))) ; (utt.relation.append utt 'Intonation syl) ; (item.relation.append_daughter syl 'Intonation ie) ) ((and (string-equal (item.feat ie "int_event") "1") (set! syl (find_related_syl utt syls syllink))) (if (not (member 'Intonation (item.relations syl))) (utt.relation.append utt 'Intonation syl)) (item.relation.append_daughter syl 'Intonation ie) (set_rel_peak_pos utt ie syl))))) (utt.relation.items utt 'IntEvent)) ;; the IntEvents )) (define (intevent_link_syls utt) "(intevent_link_syls utt) Non-tilt link of syllables to intevents through the Intonation relation." (let ((syls (utt.relation.items utt 'Syllable))) (utt.relation.create utt 'Intonation) (mapcar (lambda (ie) (let ((syl (find_container_syl ie syls))) (if (not (member 'Intonation (item.relations syl))) (utt.relation.append utt 'Intonation syl)) (item.relation.append_daughter syl 'Intonation ie))) (utt.relation.items utt 'IntEvent)) ;; the IntEvents )) (define (find_container_syl ie syls) "(find_container_syl ie syls) Find the syl thats cloests to the time on this ie." (let ((pos (item.feat ie 'end)) (ss syls) syl) (while (and ss (not syl)) (let ((ss_start (item.feat (car ss) 'syllable_start)) (ss_end (item.feat (car ss) 'syllable_end))) (if (and (> pos ss_start) (< pos (+ ss_end 0.030))) (set! syl (car ss))) (set! ss (cdr ss)))) (if (not syl) (error "Failed to find related syllable for IntEvent at" pos)) syl)) (define (find_ie_phrase_syl utt ie direction) "(find_ie_phrase_syl utt ie pos direction) Find the syllable that should be related to this IntEvent. As at this stage no real relations can be relied on this blindly searches the Syllable stream for a segment at the right time point." (let ((syls (utt.relation.items utt 'Syllable)) (pos (item.feat ie 'position)) syl) (while (and syls (not syl)) (if (or (approx-equal? pos (item.feat (car syls) direction) 0.04) (and (not (item.relation.next ie 'IntEvent)) (not (cdr syls)))) (set! syl (car syls))) (set! syls (cdr syls))) (if (not syl) (error "Failed to find related syllable for phrase IntEvent at" pos)) syl)) (define (set_rel_peak_pos utt ie syl) "(set_rel_peak_pos ie syl) Set the feature tilt:rel_pos to the distance from the start of of the vowel in syl" (item.set_feat ie "tilt:rel_pos" (- (- (item.feat ie 'end) (* (- 1.0 (item.feat ie 'tilt:tilt)) (item.feat ie 'tilt:dur) 0.5)) (syl_vowel_start syl)))) (define (find_related_syl utt syls syllink) "(find_related_syl utt syls syllink) Find the syllable name by sylllink, which is of the form x[.y]. x the word number and y is the syllable number." (unwind-protect (let (wordlab sylnum word syls syl) (if (string-matches syllink ".*\\..*") (begin (set! wordlab (string-before syllink ".")) (set! sylnum (- (parse-number (string-after syllink ".")) 1))) (begin (set! wordlab syllink) (set! sylnum 0))) (set! word (find_word_labelled utt (utt.relation.items utt 'Word) wordlab)) (if (not word) (error "Failed to find word labelled:" wordlab)) (set! syls (item.relation.daughters word 'SylStructure)) (set! syl (nth sylnum syls)) (if syl syl (car (last syls)))) (begin (error "Failed to find syllable labelled:" syllink)))) (define (find_word_labelled utt words lab) "(find_word_labelled words lab) Find the word whose label is lab." (cond ((null words) nil) ((string-equal lab (item.feat (car words) "wordlab")) (car words)) (t (find_word_labelled utt (cdr words) lab)))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; The main work ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (main) (get_options) (mapcar (lambda (f) (format t "%s\n" f) (unwind-protect (do_utt (path-basename f)) (format stderr "utterance build or save failed\n"))) seg_files)) (main) festival/examples/dumpfeats.sh0000644000175000017500000001664611110017036015144 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;-*-mode:scheme-*- ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: December 1997 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Dump features from a list of utterances ;;; ;;; Because this is a --script type file it has to explicitly ;;; load the initfiles: init.scm and user's .festivalrc (load (path-append libdir "init.scm")) (define (dumpfeats_help) (format t "%s\n" "Usage: dumpfeats [options] ... Dump features from a set of utterances Options -relation Relation from which the features have to be dumped from -output If output parameter contains a %s its treated as a skeleton e.g feats/%s.feats and multiple files will be created one each utterance. If output doesn't contain %s the output is treated as a single file and all features and dumped in it. -feats If argument starts with a \"(\" it is treated as a list of features to dump, otherwise it is treated as a filename whose contents contain a set of features (without parenetheses). -eval A scheme file to be loaded before dumping. This may contain dump specific features etc. If filename starts with a left parenthis it it evaluated as lisp. -from_file A file with a list of utterance files names (used when there are a very large number of files. ") (quit)) ;;; Default options values (defvar utt_files nil) ;; utterance files to dump from (defvar desired_relation nil) (defvar output "-") (defvar desired_features nil) (defvar extra-file nil) ;;; Get options (define (get_options) (let ((files nil) (o argv)) (if (or (member_string "-h" argv) (member_string "-help" argv) (member_string "--help" argv) (member_string "-?" argv)) (dumpfeats_help)) (while o (begin (cond ((string-equal "-relation" (car o)) (if (not (cdr o)) (dumpfeats_error "no stream file specified")) (set! desired_relation (car (cdr o))) (set! o (cdr o))) ((string-equal "-output" (car o)) (if (not (cdr o)) (dumpfeats_error "no output file/skeleton specified")) (set! output (car (cdr o))) (set! o (cdr o))) ((string-equal "-feats" (car o)) (if (not (cdr o)) (dumpfeats_error "no feats list/file specified")) (if (string-matches (car (cdr o)) "^(.*") (set! desired_features (read-from-string (car (cdr o)))) (set! desired_features (load (car (cdr o)) t))) (set! o (cdr o))) ((string-equal "-from_file" (car o)) (if (not (cdr o)) (durmeanstd_error "no file of utts names file specified")) (set! files (append (reverse (load (car (cdr o)) t)) files)) (set! o (cdr o))) ((string-equal "-eval" (car o)) (if (not (cdr o)) (dumpfeats_error "no file specified to load")) (if (string-matches (car (cdr o)) "^(.*") (eval (read-from-string (car (cdr o)))) (load (car (cdr o)))) (set! o (cdr o))) (t (set! files (cons (car o) files)))) (set! o (cdr o)))) (if files (set! utt_files (reverse files))))) (define (dumpfeats_error message) (format stderr "%s: %s\n" "dumpfeats" message) (dumpfeats_help)) ;;; No gc messages (gc-status nil) (define (dump_all_features relname feats names outskeleton) "(dump_all_features relname feats names outskeleton) Dump all names features in RELNAME from utterances in NAMES to a files or files specified by outskeleton." (let (fd) (if (not (string-matches outskeleton ".*%s.*")) (set! fd (fopen outskeleton "w"))) (mapcar (lambda (uttfile) (if (cdr names) ;; only output the utt name if there is more than one (format stderr "%s\n" uttfile)) ;; change fd to new file if in skeleton mode (if (string-matches outskeleton ".*%s.*") (set! fd (fopen (format nil outskeleton (string-before (basename uttfile) ".")) "w"))) (unwind-protect (extract_feats relname feats (utt.load nil uttfile) fd) nil) (if (string-matches outskeleton ".*%s.*") (fclose fd)) t) names) (if (not (string-matches outskeleton ".*%s.*")) (fclose fd)))) (define (extract_feats relname feats utt outfd) "(extract_feats relname feats utt outfd) Extract the features and write them to the file descriptor." (mapcar (lambda (si) (mapcar (lambda (f) (set! fval (unwind-protect (item.feat si f) "0")) (if (or (string-equal "" fval) (string-equal " " fval)) (format outfd "%l " fval) (format outfd "%s " fval))) feats) (format outfd "\n") t) (utt.relation.items utt relname)) t) (define (get_utt fname) (utt.load nil fname)) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; The main work ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (main) (get_options) (dump_all_features desired_relation desired_features utt_files output) ) (main) festival/examples/durmeanstd.sh0000644000175000017500000001573310605676216015341 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;-*-mode:scheme-*- ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: December 1997 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Find the means and standard deviations of the durations of all ;;; phones in the given utterances. ;;; ;;; Because this is a --script type file it has to explicitly ;;; load the initfiles: init.scm and user's .festivalrc (load (path-append libdir "init.scm")) (define (durmeanstd_help) (format t "%s\n" "durmeanstd [options] festival/utts/*.utts Find means and standard deviation of phone durations in utterances Options -output File to save output in -relation Relation for phones (default Segment) -log Take log of durations first -from_file file with list of utt names ") (quit)) ;;; Default options values (defvar utt_files nil) (defvar outfile "durs.meanstd") (defvar log_domain nil) (defvar durrelation 'Segment) ;;; Get options (define (get_options) (let ((files nil) (o argv)) (if (or (member_string "-h" argv) (member_string "-help" argv) (member_string "--help" argv) (member_string "-?" argv)) (durmeanstd_help)) (while o (begin (cond ((string-equal "-output" (car o)) (if (not (cdr o)) (durmeanstd_error "no output file specified")) (set! outfile (car (cdr o))) (set! o (cdr o))) ((string-equal "-relation" (car o)) (if (not (cdr o)) (durmeanstd_error "no relation file specified")) (set! durrelation (car (cdr o))) (set! o (cdr o))) ((string-equal "-from_file" (car o)) (if (not (cdr o)) (durmeanstd_error "no file of utts names file specified")) (set! files (append (load (car (cdr o)) t) files)) (set! o (cdr o))) ((string-equal "-log" (car o)) (set! log_domain t)) (t (set! files (cons (car o) files)))) (set! o (cdr o)))) (if files (set! utt_files (reverse files))))) (define (durmeanstd_error message) (format stderr "%s: %s\n" "durmeanstd" message) (durmeanstd_help)) ;;; No gc messages (gc-status nil) ;;; A simple sufficient statistics class (define (suffstats.new) (list 0 ;; n 0 ;; sum 0 ;; sumx )) (define (suffstats.set_n x n) (set-car! x n)) (define (suffstats.set_sum x sum) (set-car! (cdr x) sum)) (define (suffstats.set_sumx x sumx) (set-car! (cdr (cdr x)) sumx)) (define (suffstats.n x) (car x)) (define (suffstats.sum x) (car (cdr x))) (define (suffstats.sumx x) (car (cdr (cdr x)))) (define (suffstats.reset x) (suffstats.set_n x 0) (suffstats.set_sum x 0) (suffstats.set_sumx x 0)) (define (suffstats.add x d) (suffstats.set_n x (+ (suffstats.n x) 1)) (suffstats.set_sum x (+ (suffstats.sum x) d)) (suffstats.set_sumx x (+ (suffstats.sumx x) (* d d))) ) (define (suffstats.mean x) (/ (suffstats.sum x) (suffstats.n x))) (define (suffstats.variance x) (cond ((or (< (suffstats.n x) 2 ) (equal? (* (suffstats.n x) (suffstats.sumx x)) (* (suffstats.sum x) (suffstats.sum x)))) ;; avoid 0 variance (/ (suffstats.mean x) 10.0)) (t (/ (- (* (suffstats.n x) (suffstats.sumx x)) (* (suffstats.sum x) (suffstats.sum x))) (* (suffstats.n x) (- (suffstats.n x) 1)))))) (define (suffstats.stddev x) (sqrt (suffstats.variance x))) ;;; Index for each phone (defvar phonelist nil) ;; index of phone to suffstats (define (get_phone_data phone) (let ((a (car (cdr (assoc phone phonelist))))) (if a a (begin ;; first time for this phone (set! phonelist (cons (list phone (suffstats.new)) phonelist)) (car (cdr (assoc phone phonelist))))))) (define (duration i) (if (item.prev i) (- (item.feat i "end") (item.feat i "p.end")) (item.feat i "end"))) (define (cummulate_seg_durs utt_name) (let ((utt (utt.load nil utt_name))) (mapcar (lambda (s) (suffstats.add (get_phone_data (item.name s)) (if log_domain (log (duration s)) (duration s)))) (utt.relation.items utt durrelation)))) (define (output_dur_data data outfile) (let ((fd (fopen outfile "w"))) (mapcar (lambda (d) (format fd "(%s %f %f)\n" (car d) (suffstats.mean (car (cdr d))) (suffstats.stddev (car (cdr d))))) data) (fclose fd))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; The main work ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (main) (get_options) (mapcar (lambda (u) (unwind-protect (cummulate_seg_durs u) nil)) utt_files) (output_dur_data phonelist outfile) ) (main) festival/examples/powmeanstd.sh0000644000175000017500000001376410036506076015350 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;-*-mode:scheme-*- ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Author: Alan W Black ;;; Date: December 1997 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; Find the means and standard deviations of the power of all ;;; phones in the given utterances. ;;; ;;; Because this is a --script type file it has to explicitly ;;; load the initfiles: init.scm and user's .festivalrc (load (path-append libdir "init.scm")) (define (powmeanstd_help) (format t "%s\n" "powmeanstd [options] festival/utts/*.utts Find means and standard deviation of phone power in utterances Options -output File to save output in -log Take log of power first ") (quit)) ;;; Default options values (defvar utt_files nil) (defvar outfile "pow.meanstd") (defvar log_domain nil) ;;; Get options (define (get_options) (let ((files nil) (o argv)) (if (or (member_string "-h" argv) (member_string "-help" argv) (member_string "--help" argv) (member_string "-?" argv)) (powmeanstd_help)) (while o (begin (cond ((string-equal "-output" (car o)) (if (not (cdr o)) (powmeanstd_error "no output file specified")) (set! outfile (car (cdr o))) (set! o (cdr o))) ((string-equal "-log" (car o)) (set! log_domain t)) (t (set! files (cons (car o) files)))) (set! o (cdr o)))) (if files (set! utt_files (reverse files))))) (define (powmeanstd_error message) (format stderr "%s: %s\n" "powmeanstd" message) (powmeanstd_help)) ;;; No gc messages (gc-status nil) ;;; A simple sufficient statistics class (define (suffstats.new) (list 0 ;; n 0 ;; sum 0 ;; sumx )) (define (suffstats.set_n x n) (set-car! x n)) (define (suffstats.set_sum x sum) (set-car! (cdr x) sum)) (define (suffstats.set_sumx x sumx) (set-car! (cdr (cdr x)) sumx)) (define (suffstats.n x) (car x)) (define (suffstats.sum x) (car (cdr x))) (define (suffstats.sumx x) (car (cdr (cdr x)))) (define (suffstats.reset x) (suffstats.set_n x 0) (suffstats.set_sum x 0) (suffstats.set_sumx x 0)) (define (suffstats.add x d) (suffstats.set_n x (+ (suffstats.n x) 1)) (suffstats.set_sum x (+ (suffstats.sum x) d)) (suffstats.set_sumx x (+ (suffstats.sumx x) (* d d))) ) (define (suffstats.mean x) (/ (suffstats.sum x) (suffstats.n x))) (define (suffstats.variance x) (/ (- (* (suffstats.n x) (suffstats.sumx x)) (* (suffstats.sum x) (suffstats.sum x))) (* (suffstats.n x) (- (suffstats.n x) 1)))) (define (suffstats.stddev x) (sqrt (suffstats.variance x))) ;;; Index for each phone (defvar phonelist nil) ;; index of phone to suffstats (define (get_phone_data phone) (let ((a (car (cdr (assoc phone phonelist))))) (if a a (begin ;; first time for this phone (set! phonelist (cons (list phone (suffstats.new)) phonelist)) (car (cdr (assoc phone phonelist))))))) (define (cummulate_seg_pow utt_name) (let ((utt (utt.load nil utt_name))) (mapcar (lambda (s) (suffstats.add (get_phone_data (item.name s)) (if log_domain (log (item.feat s "power")) (item.feat s "power")))) (utt.relation.items utt 'Segment)))) (define (output_pow_data data outfile) (let ((fd (fopen outfile "w"))) (mapcar (lambda (d) (format fd "(%s %f %f)\n" (car d) (suffstats.mean (car (cdr d))) (suffstats.stddev (car (cdr d))))) data) (fclose fd))) ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; The main work ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; (define (main) (get_options) (mapcar cummulate_seg_pow utt_files) (output_pow_data phonelist outfile) ) (main) festival/examples/run-festival-script.sh0000755000175000017500000000564311223442762017110 0ustar awbawb#!/bin/sh #####################################################-*-mode:shell-script-*- ## ## ## Carnegie Mellon University ## Copyright (c) 2005 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## CARNEGIE MELLON UNIVERSITY AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL CARNEGIE MELLON UNIVERSITY NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## Author: Alan W Black ## ## Date: Nov 2008 ## ########################################################################### ## Run a festival script with an explicit libpath so an implementation ## ## can be moved from where it was compiled and the scripts can still be ## ## be used ## ########################################################################### FESTIVAL=$1 FESTLIBDIR=$2 SCRIPT=$3 shift shift shift exec $FESTIVAL --libdir $FESTLIBDIR --script $3 $* festival/examples/text2utt.sh0000755000175000017500000000730512165025207014764 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;-*-mode:scheme-*- ;; ;; ;; Centre for Speech Technology Research ;; ;; University of Edinburgh, UK ;; ;; Copyright (c) 1996,1997 ;; ;; All Rights Reserved. ;; ;; ;; ;; Permission is hereby granted, free of charge, to use and distribute ;; ;; this software and its documentation without restriction, including ;; ;; without limitation the rights to use, copy, modify, merge, publish, ;; ;; distribute, sublicense, and/or sell copies of this work, and to ;; ;; permit persons to whom this work is furnished to do so, subject to ;; ;; the following conditions: ;; ;; 1. The code must retain the above copyright notice, this list of ;; ;; conditions and the following disclaimer. ;; ;; 2. Any modifications must be clearly marked as such. ;; ;; 3. Original authors' names are not deleted. ;; ;; 4. The authors' names are not used to endorse or promote products ;; ;; derived from this software without specific prior written ;; ;; permission. ;; ;; ;; ;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;; THIS SOFTWARE. ;; ;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Because this is a --script type file I has to explicitly ;;; load the initfiles: init.scm and user's .festivalrc (load (path-append libdir "init.scm")) ;;; Process command line arguments (define (text2utt_help) (format t "%s\n" "text2utt textfile Convert a textfile to a uttrancefile ") (quit)) ;;; No gc messages (gc-status nil) ;;; Default argument values (defvar outfile "-") (defvar text_files '("-")) ;;; Get options (define (get_options) (let ((files nil) (o argv)) (if (or (member_string "-h" argv) (member_string "-help" argv) (member_string "--help" argv) (member_string "-?" argv)) (text2utt_help)) (while o (begin (cond (t (set! files (cons (car o) files)))) (set! o (cdr o)))) (if files (set! text_files (reverse files))))) (define (text2utt_error message) (format stderr "%s: %s\n" "text2utt" message) (text2utt_help)) (define (save_record_utt utt) "Saves the utterancefiles" (let ((fn (make_tmp_filename))) (utt.save utt outfile 'est_ascii) utt)) ;;; ;;; Redefine what happens to utterances during text to speech ;;; (set! tts_hooks (list utt.synth save_record_utt)) (define (main) (get_options) ;; do the synthesis (mapcar (lambda (f) (tts_file f (tts_find_text_mode f auto-text-mode-alist)) ) text_files) ) ;;; Do the work (main) festival/examples/benchmark0000755000175000017500000000300107262605650014477 0ustar awbawb#!/bin/sh default_libdir="/projects/festival/lib" while true do case "$1" in -f ) festival="${2}" shift 2 ;; -l ) libdir="$2" shift 2 ;; * ) break;; esac done text=${1-"$HOME/projects/festival/examples/benchmark.text"} for i in . src/main ../src/main $HOME/projects/festival/src/main /cstr/bin do if [ -n "$festival" ] then break; fi if [ -x "$i/festival" ] then festival="$i/festival" fi done [ -n "$festival" ] || { echo "Can't find festival" exit 1 } if [ -z "$libdir" ] then case $festival in *main/festival ) libdir=`dirname $festival`/../../lib;; * ) libdir=$default_libdir;; esac fi echo Using $festival start_flag_file="/tmp/fest_start_$$" end_flag_file="/tmp/fest_end_$$" script="/tmp/fest_script_$$" echo -n > $flag_file; cat > $script <<__END__ (set! libdir "$libdir/") (set! lexdir "$default_libdir/dicts/") (set! voiced_dir "$default_libdir/voices/") (load (string-append libdir "init.scm")) (if (probe_file (format nil "%s/.festivalrc" (getenv "HOME"))) (load (format nil "%s/.festivalrc" (getenv "HOME")))) (audio_mode 'async) (set! tts_hooks (list utt.synth)) (puts "start...\n" nil) (fclose (fopen "$start_flag_file" "w")) (tts_file "$text" (quote text)) (fclose (fopen "$end_flag_file" "w")) (puts "...end\n" nil) (audio_mode 'close) (quit) __END__ eval $festival --script $script perl -e 'print "running time = ", (stat($ARGV[1]))[8]-(stat($ARGV[0]))[8], " seconds\n";' $start_flag_file $end_flag_file /bin/rm -f $start_flag_file $end_flag_file $script festival/examples/festival_client.pl0000644000175000017500000001136707262605650016345 0ustar awbawb#!/usr/local/bin/perl # festival_client.pl - a perl socket client for festival # # Copyright (C) 1997 # Kevin A. Lenzo (lenzo@cs.cmu.edu) 7/97 # All rights reserved. # # The authors hereby grant permission to use, copy, modify, distribute, # and license this software and its documentation for any purpose, provided # that existing copyright notices are retained in all copies and that this # notice is included verbatim in any distributions. No written agreement, # license, or royalty fee is required for any of the authorized uses. # Modifications to this software may be copyrighted by their authors # and need not follow the licensing terms described here, provided that # the new terms are clearly indicated on the first page of each file where # they apply. # # IN NO EVENT SHALL THE AUTHORS OR DISTRIBUTORS BE LIABLE TO ANY PARTY # FOR DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES # ARISING OUT OF THE USE OF THIS SOFTWARE, ITS DOCUMENTATION, OR ANY # DERIVATIVES THEREOF, EVEN IF THE AUTHORS HAVE BEEN ADVISED OF THE # POSSIBILITY OF SUCH DAMAGE. # # THE AUTHORS AND DISTRIBUTORS SPECIFICALLY DISCLAIM ANY WARRANTIES, # INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY, # FITNESS FOR A PARTICULAR PURPOSE, AND NON-INFRINGEMENT. THIS SOFTWARE # IS PROVIDED ON AN "AS IS" BASIS, AND THE AUTHORS AND DISTRIBUTORS HAVE # NO OBLIGATION TO PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR # MODIFICATIONS. # ######################################################################### # # A good deal of this was taken from the example in the # perlipc man page. The rest is simply the Festival-specific # stuff. use IO::Socket; package Festival; my ($audio_command, $audio_file, $file_stuff_key); my ($host, $port, $kidpid, $handle, $line); $wave_type = "nist"; # the type of the audio files $audio_command = "na_play"; # your local audio play command # uncomment the next line if your audio command requires a file # $audio_file = "/tmp/client_tmp$$.nist"; # temp file for waveforms $file_stuff_key = "ft_StUfF_key"; # defined in speech tools $host = shift || 'localhost'; $port = shift || 1314; if ($host eq "-h") { print STDOUT " Usage: $0 [ []] OR $0 -h perl client for the Festival text-to-speech server Lines that do not begin with a ( are treated as text to be spoken. Those that do begin with a ( are assumed to Scheme commands to festival and are sent without alteration. Use `quit' or `exit' to quit. Note that a server must be started before the client will work. Use the `-h' (help) option for this message. "; exit(1); } # create a tcp connection to the specified host and port $handle = IO::Socket::INET->new(Proto => "tcp", PeerAddr => $host, PeerPort => $port) or die " Can't connect to port $port on $host: $! (Are you sure the server is running and accepting connections?) "; $handle->autoflush(1); # so output gets there right away print STDERR "[Connected to $host:$port]\n"; # tell the server to send us back a 'file' of the right type print $handle "(Parameter.set 'Wavefiletype '$wave_type)\n"; # split the program into two processes, identical twins die "can't fork: $!" unless defined($kidpid = fork()); # the if{} block runs only in the parent process if ($kidpid) { # the parent handles the input so it can exit on quit while (defined ($line = )) { last if ($line =~ /^(quit|exit)$/); if ($line =~ /^\(/) { # just send it wholesale if it's a ( ) print $handle $line; } else { # otherwise assume it's text to be spoken chomp $line; print $handle "(tts_textall \"$line\" 'file)\n"; } } kill("TERM", $kidpid);# send SIGTERM to child } # the else{} block runs only in the child process else { # the child is forked off to get the results from the server undef $line; while (($line = $remains) || defined ($line = <$handle>)) { undef $remains; if ($line eq "WV\n") { # we have a waveform coming undef $result; if ($audio_file) { open(AUDIO, ">$audio_file"); } else { open(AUDIO, "| $audio_command"); } while ($line = <$handle>) { if ($line =~ s/$file_stuff_key(.*)$//s) { $remains = $1; print AUDIO $line; last; } print AUDIO $line; } close AUDIO; if ($audio_file) { # call the command if we weren't piping system("$audio_command $audio_file"); # remove the evidence unlink($audio_file); } } elsif ($line eq "LP\n") { while ($line = <$handle>) { if ($line =~ s/$file_stuff_key(.*)$//s) { $remains = $1; print STDOUT $line; last; } print STDOUT $line; } } else { # if we don't recognize it, echo it print STDOUT $line; } } } festival/examples/toksearch.scm0000644000175000017500000001171207262605650015316 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; A search for token occurrences in buckets of text ;;; ;;; This is only an example to aid you, this actually depends on ;;; the availability of databases we don't have permission to ;;; distribute. (set! text_dir "/home/awb/data/text/") ;;; The databases themselves are identified by a file which names all ;;; the files in that databases. e.g. This expects bin/gutenberg.files ;;; to exists which should contain something like ;;; gutenberg/etext90/bill11.txt ;;; gutenberg/etext90/const11.txt ;;; gutenberg/etext90/getty11.txt (set! db_names '("gutenberg" ;; books from gutenberg 21906570 "desktopshop" ;; books, documents etc 23090463 "time" ;; Time Magazine 1990-1994 6770175 "hutch" ;; Hutchinson Encyclopedia 1715268 "dicts" ;; Dictionaries and Encyclopedias 4248109 "stw-ref" ;; Standard Reference libraries 3330448 "treebank" ;; WSJ articles from PENN treebank 1109895 "email" ;; awb's email )) ;;; Identify the tokens you want extracted ;;; Tokens may be regular expressions (set! desired_tokens '(lead wound tear axes Jan bass Nice Begin Chi Colon St Dr III IV V X VII II "[0-9]+")) ;;; First pass: to get examples and context for labelling (set! desired_feats '(filepos p.p.p.p.name p.p.p.name p.p.name p.name name n.name nn.name n.n.n.name n.n.n.n.name)) ;;; Second: pass to get desried features for tree building ;;; Typically this has to be specific for a particular homograph ;;; so you'll probably want to do multiple second passes one for each ;;; homograph type ;(set! desired_feats ; '(filepos ; lisp_tok_rex ; p.punc ; punc ; n.punc ; pp.cap p.cap n.cap nn.cap ; )) (define (tok_search_db dbname) "Search through DB for named tokens and save found occurrences." (let ((outfile (string-append text_dir "fullhgs/" dbname ".out"))) (delete-file outfile) (mapcar (lambda (fname) ;; for each file in the database (extract_tokens ;; call internal function to extract tokens (string-append text_dir fname) ;; full pathname to extract from (mapcar ;; list of tokens and features (lambda (t) ;; to extract (cons t desired_feats)) desired_tokens) outfile)) (load (string-append text_dir "bin/" dbname ".files") t)) t)) (define (tok_do_all) "Search all dbs for desired tokens." (mapcar (lambda (db) (print db) (tok_search_db db)) db_names) t) festival/examples/th-mode.scm0000644000175000017500000001715707262616703014702 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; This contains an example of how to use Festival with a ;;; a talking head. This is a combination of various answers given ;;; to different groups who have been using Festival with talking ;;; heads ;;; ;;; This version has not actually been used by any talking head ;;; but just serves as an example. ;;; ;;; The basic mode produces /tmp/th.ph (phone info) and /tmp/th.com ;;; (commands: smile frown) for each utterance in the file. These ;;; files are produced and the program makefaces called before ;;; waveform synthesis for each utterance. The play command then ;;; calls xanim with the generate animation and waveform. ;;; ;;; There are probabaly better way to do this. Using Festival as a ;;; server to generate the phone and command files might ;;; be more reasonable. Note festival not supports the returning ;;; of Lisp data to the client as well as waveform data. ;;; In that case you'd want to change th_output_info to use ;;; the send_cleint command and package the phone info into an ;;; s-expression. (defvar th-prepare-prog "makefaces" " A program that takes phones and other data and produces the animated face.") (define (utt.save.phonedata utt filename) "(utt.save.mydata UTT FILE) Saves phone, duration, stress, F0 word pos." (let ((fd (fopen filename "w"))) (mapcar (lambda (seg) (format fd "%s %2.4f %s %s" (item.feat seg "name") (item.feat seg "segment_duration") (item.feat seg "R:SylStructure.parent.stress") (item.feat seg "R:Target.daughter1.name")) ;; output word name and part of speech if start of word (if (and (not (item.relation.next seg "SylStructure")) (not (item.next (item.relation.parent seg "SylStructure")))) (format fd " %s %s" (item.feat seg "R:SylStructure.parent.parent.name") (item.feat seg "R:SylStructure.parent.parent.pos"))) (format fd "\n")) (utt.relation.items utt 'Segment)) (fclose fd) utt)) (define (utt.save.commands utt filename) "(utt.save.commands UTT FILE) Save commands with time stamps. Commands are those tokens which start and end with an asterisk." (let ((fd (fopen filename "w"))) (format fd "#\n") (mapcar (lambda (tok_item) (if (string-matches (item.name tok_item) "\\*.+\\*") (format fd "%2.4f 100 %s\n" (find_com_time utt tok_item) (item.name tok_item)))) (utt.relation.items utt 'Token)) (fclose fd) utt)) (define (find_com_time utt tok_item) "Returns time of tok_item. Looks backward for first token that is related to a word and returns the end time of that word." (cond ((item.daughtern tok_item) (item.feat (item.daughtern tok_item) "word_end")) ((not (item.prev tok_item)) ;; start of stream 0.0) (t (find_com_time utt (item.prev tok_item))))) (define (th_output_info utt) "(th_output_info utt) This is called after linguistic analysis but before waveform synthesis. It collects the phone and duration data and also any th commands found in the utterance. The file names are then passed to some external program which will process them for the talking head." (set! th-current-file "/tmp/th") ;; this should have a process id in it (utt.save.phonedata utt (string-append th-current-file ".ph")) (utt.save.commands utt (string-append th-current-file ".com")) ;; It would be good to background this process as long as you ;; resync at play time (system (format nil "%s %s %s" th-prepare-prog (string-append th-current-file ".ph") (string-append th-current-file ".ph"))) utt) ;;; ;;; Define a new text mode for talking heads ;;; (define (th_init_func) "Called on starting talking head text mode." (set! th_previous_t2w_func token_to_words) (set! th_previous_after_analysis_hooks after_analysis_hooks) (set! after_analysis_hooks (list th_output_info)) (set! english_token_to_words th_token_to_words) (set! token_to_words th_token_to_words) ;; We assume the prepare talking head program generates a movie ;; that can be played by something, so we redefie the audio ;; player to play the generated animation and waveform. (set! th_previous_Parameter Parameter) (audio_mode 'sync) ;; ensure new Audio command gets passed to new audiosp (Parameter.set 'Audio_Required_Format 'riff) (Parameter.set 'Audio_Command "xanim /tmp/th.anime $FILE") (Parameter.set 'Audio_Method 'Audio_Command) (audio_mode 'async) ) (define (th_exit_func) "Called on exit talking head text mode." (set! token_to_words th_previous_t2w_func) (set! english_token_to_words th_previous_t2w_func) (set! after_analysis_hooks th_previous_after_analysis_hooks) (audio_mode 'sync) ;; so we can reset the audio (set! Parameter th_previous_Parameter) ) (define (th_token_to_words token name) "(th_token_to_words TOKEN NAME) Talking head specific token to word rules." (cond ((string-matches name "\\*.*\\*") ;; Symbols started and ended with an asterisk as treated as commands ;; and not rendered as speech nil) (t (th_previous_t2w_func token name)))) (set! tts_text_modes (cons (list 'th ;; mode name (list ;; ogimarkup mode params (list 'init_func th_init_func) (list 'exit_func th_exit_func))) tts_text_modes)) (provide 'th-mode) festival/examples/addr-mode.scm0000644000175000017500000003052311641345451015164 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1998 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; An address mode for reading lists of names, addresses and ;;; telephone numbers. Takes quite an aggressive view of the data. ;;; ;;; This was used for the CSTR's entry in the evaluations at the ;;; ESCA workshop on Speech Synthesis in Janolen Caves, Blue Mountains, ;;; NSW, Australia. ;;; ;;; This can read things like ;;; Brown, Bill, 6023 Wiser Rd, Austin, TX 76313-2837, 817-229-7849 ;;; Green, Bob, 3076 Wabash Ct, Fort Worth, TX 76709-1368, (817)292-9015 ;;; Smith, Bobbie Q, 3337 St Laurence St, Fort Worth, TX 71611-5484, (817)839-3689 ;;; Jones, Billy, 5306 Dr Dana Lynn Dr, Fort Worth, TX 71637-2547, 817 845 6154 ;;; Henderson, Bryan J, 5808 Sycamore Creek Rd Apt R, Fort Worth, TX 76134-1906, (817)239-4634 ;;; Black, Alan W, 130 S 18th St #3, Pittsburgh, PA 15205, (412)268-8189 ;;; Bowman, K, 2610 W Bowie St, El Paso, TX 76019-1712, (817)268-7257 ;;; Sydney, Aaron A, 1521 NW Ballard Way, Seattle, WA 91807-4712, (206)783-8645 ;;; Anderson, A, 12012 Pinehurst Way NE, Seattle, NE 98125-5108, (212)404-9988 ;; New lines without trailing continuation punctuation signal EOU (defvar addr_eou_tree '((n.whitespace matches ".*\n.*") ;; any new line ((punc in ("," ":")) ((0)) ((1))))) (set! addr_phrase_cart_tree ' ((pbreak is "B") ((B)) ((pbreak is "BB") ((BB)) ((lisp_token_end_punc in ("?" "." ":" "'" "\"" "," ";")) ((B)) ((n.name is 0) ;; end of utterance ((BB)) ((NB))))))) (define (addr_init_func) "Called on starting addr text mode." (Parameter.set 'Phrase_Method 'cart_tree) (set! phrase_cart_tree addr_phrase_cart_tree) (set! int_lr_params '((target_f0_mean 105) (target_f0_std 12) (model_f0_mean 170) (model_f0_std 34))) (Parameter.set 'Duration_Stretch 1.1) (set! addr_previous_t2w_func english_token_to_words) (set! english_token_to_words addr_token_to_words) (set! token_to_words addr_token_to_words) (set! addr_previous_eou_tree eou_tree) (set! eou_tree addr_eou_tree)) (define (addr_exit_func) "Called on exit addr text mode." (Parameter.set 'Duration_Stretch 1.0) (set! token_to_words addr_previous_t2w_func) (set! english_token_to_words addr_previous_t2w_func) (set! eou_tree addr_previous_eou_tree)) (set! addr_regex_ZIPCODE2 "[0-9][0-9][0-9][0-9][0-9]-[0-9][0-9][0-9][0-9]") (set! addr_regex_USPHONE3 "[0-9][0-9][0-9])[0-9][0-9][0-9]-[0-9][0-9][0-9][0-9]") (set! addr_regex_USPHONE1 "[0-9][0-9][0-9]-[0-9][0-9][0-9]-[0-9][0-9][0-9][0-9]") (set! addr_tiny_break (list '(name "") '(pbreak mB))) (set! addr_small_break (list '(name "") '(pbreak B))) (set! addr_large_break (list '(name "") '(pbreak BB))) (define (addr_within_name_part token) "(addr_within_name_part token) Heuristic guess if we are still in the name (i.e. pre-address) this is desgined to stop Mr W Smith becoming West." (cond ((addr_preceding_number token) ;; any preceding token with a digit nil) (t t))) (define (addr_preceding_number tok) (cond ((null tok) nil) ((string-matches (item.name tok) ".*[0-9].*") t) (t (addr_preceding_number (item.prev tok))))) (define (addr_succeeding_non_number tok) (cond ((null tok) nil) ((string-matches (item.name tok) ".*[A-Za-z].*") t) (t (addr_succeeding_non_number (item.next tok))))) (define (addr_token_to_words token name) "(addr_token_to_words token name) Address specific text reading mode. Lots of address specific abbreviations and phrasing etc." (set! utt_addr (item.get_utt token)) (let ((type (item.feat token "token_type"))) (cond ((string-matches name "A") (list (list '(name "a") '(pos nn)))) ((addr_within_name_part token) (builtin_english_token_to_words token name) ) ((string-matches name "\\([dD][Rr]\\|[Ss][tT]\\)") (if (string-equal (item.feat token "token_pos") "street") (if (string-matches name "[dD][rR]") (list "drive") (list "street")) (if (string-matches name "[dD][rR]") ;; default on title side (list "doctor") (list "saint")))) ((string-matches name addr_regex_ZIPCODE2) ;; Zip code (item.set_feat token "token_pos" "digits") (append (builtin_english_token_to_words token (string-before name "-")) (list addr_small_break) (builtin_english_token_to_words token (string-after name "-")) (list addr_large_break))) ((string-matches name addr_regex_USPHONE3) (item.set_feat token "token_pos" "digits") (append (builtin_english_token_to_words token (string-before name ")")) (list addr_small_break) (builtin_english_token_to_words token (string-after (string-before name "-") ")")) (list addr_small_break) (builtin_english_token_to_words token (string-after name "-")))) ((string-matches name addr_regex_USPHONE1) (item.set_feat token "token_pos" "digits") (append (builtin_english_token_to_words token (string-before name "-")) (list addr_small_break) (builtin_english_token_to_words token (string-before (string-after name "-") "-")) (list addr_small_break) (builtin_english_token_to_words token (string-after (string-after name "-") "-")))) ((string-equal name "NE") (cond ((string-matches (item.feat token "n.name") addr_regex_ZIPCODE2) (list "Nebraska")) ;; could check if there is a state following it (t (list "North" "East")))) ((set! full (addr_undo_abbrev name addr_addr_abbrevs)) (cdr full)) ((string-matches name "#.*") (cons "number" (builtin_english_token_to_words token (string-after name "#")))) ((string-matches name "[0-9][0-9][0-9][0-9][0-9]+") ;; long number (item.set_feat token "token_pos" "digits") (builtin_english_token_to_words token name)) ((or (string-matches name "[0-9]0[0-9][0-9]") (string-matches name "[0-9][0-9]0[0-9]")) (item.set_feat token "token_pos" "digits") (mapcar (lambda (a) (if (string-equal a "zero") "oh" a)) (builtin_english_token_to_words token name))) ((and (addr_succeeding_non_number token) (string-matches name "[0-9][0-9][0-9][0-9]")) ;; four digit number (let (block number) (item.set_feat token "token_pos" "number") (set! block (builtin_english_token_to_words token (substring name 0 2))) (if (string-equal (nth 2 (symbolexplode name)) "0") (item.set_feat token "token_pos" "digits") (item.set_feat token "token_pos" "number")) (set! number (builtin_english_token_to_words token (substring name 2 2))) (append block (list addr_tiny_break) number))) ((and (addr_succeeding_non_number token) (string-matches name "[0-9][0-9][0-9]")) ;; four digit number (let (block number) (item.set_feat token "token_pos" "number") (set! block (builtin_english_token_to_words token (substring name 0 1))) (if (string-equal (nth 1 (symbolexplode name)) "0") (item.set_feat token "token_pos" "digits") (item.set_feat token "token_pos" "number")) (set! number (builtin_english_token_to_words token (substring name 1 2))) (append block number))) ((string-matches name "[0-9]+") (item.set_feat token "token_pos" "digits") (builtin_english_token_to_words token name)) (t ;; for all other cases (addr_previous_t2w_func token name))))) (define (addr_undo_abbrev name abbrevs) "(addr_undo_abbrev name abbrevs) General abbreviation undoer. Looks for name in reverse assoc list and returns the value." (cond ((null abbrevs) nil) ((member_string name (car (car abbrevs))) (car abbrevs)) (t (addr_undo_abbrev name (cdr abbrevs))))) (set! tts_text_modes (cons (list 'addr ;; mode name (list ;; addr mode params (list 'init_func addr_init_func) (list 'exit_func addr_exit_func))) tts_text_modes)) (set! addr_us_states '((( AL Ala ) Alabama ) (( AK ) Alaska ) (( AR ) Arkansas ) (( AZ Ariz ) Arizona ) (( CA Cal Calif ) California ) (( CO Colo ) Colorado ) (( CT Conn ) Connecticutt ) (( DC ) DC ) (( DE Dela ) Delaware ) (( FL Fla ) Florida ) (( GA ) Georgia ) (( HI ) Hawaii ) (( IA Ind ) Indiana ) (( ID ) Idaho ) (( IL Ill ) Illinois ) (( KS Kans ) Kansas ) (( KY ) Kentucky ) (( LA Lou Lous) Louisiana ) (( MA Mass ) Massachusetts ) (( MD ) Maryland ) (( ME ) Maine ) (( MI Mich ) Michigan ) (( MN Minn ) Minnesota ) (( MS Miss ) Mississippi ) (( MT ) Montana ) (( MO ) Missouri ) (( NC ) North Carolina ) (( ND ) North Dakota ) (( NE Neb ) Nebraska ) (( NH ) New Hampshire) (( NV Nev ) Nevada ) (( NY ) New York ) (( OH ) Ohio ) (( OK Okla ) Oklahoma ) (( Or Ore ) Oregon ) (( PA Penn ) Pennsylvania ) (( RI ) Rhode Island ) (( SC ) Sourth Carolina ) (( SD ) Sourth Dakota ) (( TN Tenn ) Tennessee ) (( TX Tex ) Texas ) (( UT ) Utah ) (( VA Vir ) Virginia ) (( VT ) Vermont ) (( WA Wash ) Washington ) (( WI Wisc ) Wisconsin ) (( WV ) West Virginia ) (( WY Wyom ) Wyoming ) (( PR ) Puerto Rico ) )) (set! addr_compass_points '(((S So Sth) South) ((N No Nor) North) ((E) East) ((W) West) ((NE) North East) ((NW) North West) ((SE) South East) ((SW) South West))) (set! addr_streets '(((Hwy) Highway) ((Rt Rte) Root) ((Ct) Court) ((Pl) Place) ((Blvd Bld) Boulevard) ((Ave) Avenue) ((Rd) Road) ((Apt App Appt) Apartment) ((Cntr Ctr) Center) ((Ter Terr Tr) Terrace) ((Ln) Lane) ((PO) pea oh) )) (set! addr_uk_counties '((( Hants ) Hampshire) (( Soton ) Southampton ) (( Berks ) Berkshire ) (( Yorks ) Yorkshire ) (( Leics ) Leicestershire ) (( Shrops ) Shropshire ) (( Cambs ) Cambridgeshire ) (( Oxon ) Oxfordshire ) (( Notts ) Nottinghamshire ) (( Humbers ) Humberside ) (( Glams ) Glamorganshire ) (( Pembs ) Pembrookeshire ) (( Lancs ) Lancashire ) (( Berwicks ) Berwickshire ) )) (set! addr_addr_abbrevs (append addr_us_states addr_compass_points addr_streets)) (provide 'addr-mode) festival/examples/festival_client.c0000644000175000017500000003265707262605651016162 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1999 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black (awb@cstr.ed.ac.uk) */ /* Date : March 1999 */ /*-----------------------------------------------------------------------*/ /* */ /* Client end of Festival server API in C designed specifically for */ /* Galaxy Communicator use though might be of use for other things */ /* */ /* This is a standalone C client, no other Festival or Speech Tools */ /* libraries need be link with this. Thus is very small. */ /* */ /* Compile with (plus socket libraries if required) */ /* cc -o festival_client -DSTANDALONE festival_client.c */ /* */ /* Run as */ /* festival_client -text "hello there" -o hello.snd */ /* */ /* */ /* This is provided as an example, it is quite limited in what it does */ /* but is functional compiling without -DSTANDALONE gives you a simple */ /* API */ /* */ /*=======================================================================*/ #include #include #include #include #include #include #include #include #include #include "festival_client.h" /* For testing endianness */ int fapi_endian_loc = 1; static char *socket_receive_file_to_buff(int fd,int *size); void delete_FT_Wave(FT_Wave *wave) { if (wave != 0) { if (wave->samples != 0) free(wave->samples); free(wave); } } int save_FT_Wave_snd(FT_Wave *wave, const char *filename) { FILE *fd; struct { unsigned int magic; /* magic number */ unsigned int hdr_size; /* size of this header */ int data_size; /* length of data (optional) */ unsigned int encoding; /* data encoding format */ unsigned int sample_rate; /* samples per second */ unsigned int channels; /* number of interleaved channels */ } header; short sw_short; int i; if ((filename == 0) || (strcmp(filename,"stdout") == 0) || (strcmp(filename,"-") == 0)) fd = stdout; else if ((fd = fopen(filename,"wb")) == NULL) { fprintf(stderr,"save_FT_Wave: can't open file \"%s\" for writing\n", filename); return -1; } header.magic = (unsigned int)0x2e736e64; header.hdr_size = sizeof(header); header.data_size = 2 * wave->num_samples; header.encoding = 3; /* short */ header.sample_rate = wave->sample_rate; header.channels = 1; if (FAPI_LITTLE_ENDIAN) { /* snd is always sparc/68000 byte order */ header.magic = SWAPINT(header.magic); header.hdr_size = SWAPINT(header.hdr_size); header.data_size = SWAPINT(header.data_size); header.encoding = SWAPINT(header.encoding); header.sample_rate = SWAPINT(header.sample_rate); header.channels = SWAPINT(header.channels); } /* write header */ if (fwrite(&header, sizeof(header), 1, fd) != 1) return -1; if (FAPI_BIG_ENDIAN) fwrite(wave->samples,sizeof(short),wave->num_samples,fd); else { /* have to swap */ for (i=0; i < wave->num_samples; i++) { sw_short = SWAPSHORT(wave->samples[i]); fwrite(&sw_short,sizeof(short),1,fd); } } if (fd != stdout) fclose(fd); return 0; } void delete_FT_Info(FT_Info *info) { if (info != 0) free(info); } static FT_Info *festival_default_info() { FT_Info *info; info = (FT_Info *)malloc(1 * sizeof(FT_Info)); info->server_host = FESTIVAL_DEFAULT_SERVER_HOST; info->server_port = FESTIVAL_DEFAULT_SERVER_PORT; info->text_mode = FESTIVAL_DEFAULT_TEXT_MODE; info->server_fd = -1; return info; } static int festival_socket_open(const char *host, int port) { /* Return an FD to a remote server */ struct sockaddr_in serv_addr; struct hostent *serverhost; int fd; fd = socket(AF_INET, SOCK_STREAM, IPPROTO_TCP); if (fd < 0) { fprintf(stderr,"festival_client: can't get socket\n"); return -1; } memset(&serv_addr, 0, sizeof(serv_addr)); if ((serv_addr.sin_addr.s_addr = inet_addr(host)) == -1) { /* its a name rather than an ipnum */ serverhost = gethostbyname(host); if (serverhost == (struct hostent *)0) { fprintf(stderr,"festival_client: gethostbyname failed\n"); return -1; } memmove(&serv_addr.sin_addr,serverhost->h_addr, serverhost->h_length); } serv_addr.sin_family = AF_INET; serv_addr.sin_port = htons(port); if (connect(fd, (struct sockaddr *)&serv_addr, sizeof(serv_addr)) != 0) { fprintf(stderr,"festival_client: connect to server failed\n"); return -1; } return fd; } static int nist_get_param_int(char *hdr, char *field, int def_val) { char *p; int val; if (((p=strstr(hdr,field)) != NULL) && (strncmp(" -i ",p+strlen(field),4) == 0)) { sscanf(p+strlen(field)+4,"%d",&val); return val; } else return def_val; } static int nist_require_swap(char *hdr) { char *p; char *field = "sample_byte_format"; if ((p=strstr(hdr,field)) != NULL) { if (((strncmp(" -s2 01",p+strlen(field),7) == 0) && FAPI_BIG_ENDIAN) || ((strncmp(" -s2 10",p+strlen(field),7) == 0) && FAPI_LITTLE_ENDIAN)) return 1; } return 0; /* if unknown assume native byte order */ } static char *client_accept_s_expr(int fd) { /* Read s-expression from server, as a char * */ char *expr; int filesize; expr = socket_receive_file_to_buff(fd,&filesize); expr[filesize] = '\0'; return expr; } static FT_Wave *client_accept_waveform(int fd) { /* Read waveform from server */ char *wavefile; int filesize; int num_samples, sample_rate, i; FT_Wave *wave; wavefile = socket_receive_file_to_buff(fd,&filesize); wave = 0; /* I know this is NIST file and its an error if it isn't */ if (filesize >= 1024) { num_samples = nist_get_param_int(wavefile,"sample_count",0); sample_rate = nist_get_param_int(wavefile,"sample_rate",16000); if ((num_samples*sizeof(short))+1024 == filesize) { wave = (FT_Wave *)malloc(sizeof(FT_Wave)); wave->num_samples = num_samples; wave->sample_rate = sample_rate; wave->samples = (short *)malloc(num_samples*sizeof(short)); memmove(wave->samples,wavefile+1024,num_samples*sizeof(short)); if (nist_require_swap(wavefile)) for (i=0; i < num_samples; i++) wave->samples[i] = SWAPSHORT(wave->samples[i]); } } if (wavefile != 0) /* just in case we've got an ancient free() */ free(wavefile); return wave; } static char *socket_receive_file_to_buff(int fd,int *size) { /* Receive file (probably a waveform file) from socket using */ /* Festival key stuff technique, but long winded I know, sorry */ /* but will receive any file without closeing the stream or */ /* using OOB data */ static char *file_stuff_key = "ft_StUfF_key"; /* must == Festival's key */ char *buff; int bufflen; int n,k,i; char c; bufflen = 1024; buff = (char *)malloc(bufflen); *size=0; for (k=0; file_stuff_key[k] != '\0';) { n = read(fd,&c,1); if (n==0) break; /* hit stream eof before end of file */ if ((*size)+k+1 >= bufflen) { /* +1 so you can add a NULL if you want */ bufflen += bufflen/4; buff = (char *)realloc(buff,bufflen); } if (file_stuff_key[k] == c) k++; else if ((c == 'X') && (file_stuff_key[k+1] == '\0')) { /* It looked like the key but wasn't */ for (i=0; i < k; i++,(*size)++) buff[*size] = file_stuff_key[i]; k=0; /* omit the stuffed 'X' */ } else { for (i=0; i < k; i++,(*size)++) buff[*size] = file_stuff_key[i]; k=0; buff[*size] = c; (*size)++; } } return buff; } /***********************************************************************/ /* Public Functions to this API */ /***********************************************************************/ FT_Info *festivalOpen(FT_Info *info) { /* Open socket to server */ if (info == 0) info = festival_default_info(); info->server_fd = festival_socket_open(info->server_host, info->server_port); if (info->server_fd == -1) return NULL; return info; } FT_Wave *festivalStringToWave(FT_Info *info,char *text) { FT_Wave *wave; FILE *fd; char *p; char ack[4]; int n; if (info == 0) return 0; if (info->server_fd == -1) { fprintf(stderr,"festival_client: server connection unopened\n"); return 0; } fd = fdopen(dup(info->server_fd),"wb"); /* Copy text over to server, escaping any quotes */ fprintf(fd,"(tts_textall \"\n"); for (p=text; p && (*p != '\0'); p++) { if ((*p == '"') || (*p == '\\')) putc('\\',fd); putc(*p,fd); } fprintf(fd,"\" \"%s\")\n",info->text_mode); fclose(fd); /* Read back info from server */ /* This assumes only one waveform will come back, also LP is unlikely */ wave = 0; do { for (n=0; n < 3; ) n += read(info->server_fd,ack+n,3-n); ack[3] = '\0'; if (strcmp(ack,"WV\n") == 0) /* receive a waveform */ wave = client_accept_waveform(info->server_fd); else if (strcmp(ack,"LP\n") == 0) /* receive an s-expr */ client_accept_s_expr(info->server_fd); else if (strcmp(ack,"ER\n") == 0) /* server got an error */ { fprintf(stderr,"festival_client: server returned error\n"); break; } } while (strcmp(ack,"OK\n") != 0); return wave; } int festivalClose(FT_Info *info) { if (info == 0) return 0; if (info->server_fd != -1) close(info->server_fd); return 0; } #ifdef STANDALONE int main(int argc, char **argv) { char *server=0; int port=-1; char *text=0; char *output=0; char *mode=0; int i; FT_Info *info; FT_Wave *wave; for (i=1; i < argc; i++) { if (strcmp(argv[i],"-server") == 0) server = argv[++i]; else if (strcmp(argv[i],"-port") == 0) port = atoi(argv[++i]); else if (strcmp(argv[i],"-text") == 0) text = argv[++i]; else if (strcmp(argv[i],"-mode") == 0) mode = argv[++i]; else if (strcmp(argv[i],"-o") == 0) output = argv[++i]; } if (i > argc) { fprintf(stderr,"missing argument\n"); exit(1); } info = festival_default_info(); if (server != 0) info->server_host = server; if (port != -1) info->server_port = port; if (mode != 0) info->text_mode = mode; info = festivalOpen(info); wave = festivalStringToWave(info,text); if (wave != 0) save_FT_Wave_snd(wave,output); festivalClose(info); return 0; } #endif festival/examples/festival_client.h0000644000175000017500000001056207262605651016156 0ustar awbawb/*************************************************************************/ /* */ /* Centre for Speech Technology Research */ /* University of Edinburgh, UK */ /* Copyright (c) 1999 */ /* All Rights Reserved. */ /* */ /* Permission is hereby granted, free of charge, to use and distribute */ /* this software and its documentation without restriction, including */ /* without limitation the rights to use, copy, modify, merge, publish, */ /* distribute, sublicense, and/or sell copies of this work, and to */ /* permit persons to whom this work is furnished to do so, subject to */ /* the following conditions: */ /* 1. The code must retain the above copyright notice, this list of */ /* conditions and the following disclaimer. */ /* 2. Any modifications must be clearly marked as such. */ /* 3. Original authors' names are not deleted. */ /* 4. The authors' names are not used to endorse or promote products */ /* derived from this software without specific prior written */ /* permission. */ /* */ /* THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK */ /* DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING */ /* ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT */ /* SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE */ /* FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES */ /* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN */ /* AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, */ /* ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF */ /* THIS SOFTWARE. */ /* */ /*************************************************************************/ /* Author : Alan W Black (awb@cstr.ed.ac.uk) */ /* Date : March 1999 */ /*-----------------------------------------------------------------------*/ /* */ /* Client end of Festival server API (in C) designed specifically for */ /* Galaxy Communicator use, though might be of use for other things */ /* */ /*=======================================================================*/ #ifndef _FESTIVAL_CLIENT_H_ #define _FESTIVAL_CLIENT_H_ #define FESTIVAL_DEFAULT_SERVER_HOST "localhost" #define FESTIVAL_DEFAULT_SERVER_PORT 1314 #define FESTIVAL_DEFAULT_TEXT_MODE "fundamental" typedef struct FT_Info { int encoding; char *server_host; int server_port; char *text_mode; int server_fd; } FT_Info; typedef struct FT_Wave { int num_samples; int sample_rate; short *samples; } FT_Wave; void delete_FT_Wave(FT_Wave *wave); void delete_FT_Info(FT_Info *info); #define SWAPSHORT(x) ((((unsigned)x) & 0xff) << 8 | \ (((unsigned)x) & 0xff00) >> 8) #define SWAPINT(x) ((((unsigned)x) & 0xff) << 24 | \ (((unsigned)x) & 0xff00) << 8 | \ (((unsigned)x) & 0xff0000) >> 8 | \ (((unsigned)x) & 0xff000000) >> 24) /* Sun, HP, SGI Mips, M68000 */ #define FAPI_BIG_ENDIAN (((char *)&fapi_endian_loc)[0] == 0) /* Intel, Alpha, DEC Mips, Vax */ #define FAPI_LITTLE_ENDIAN (((char *)&fapi_endian_loc)[0] != 0) /*****************************************************************/ /* Public functions to interface */ /*****************************************************************/ /* If called with NULL will attempt to access using defaults */ FT_Info *festivalOpen(FT_Info *info); FT_Wave *festivalStringToWave(FT_Info *info,char *text); int festivalClose(FT_Info *info); #endif festival/examples/apml.dtd0000644000175000017500000000230507707512466014261 0ustar awbawb festival/examples/speech_pm_1.0.tar0000644000175000017500000014400207262616703015660 0ustar awbawbspeech_pm_1.0/MANIFEST100444 65206 76603 304 6777424661 12641 0ustar rjccstrMANIFEST README.txt Makefile.PL Changes Speech/Festival.pm Speech/Festival/Synthesiser.pm Speech/Synthesiser.pm Audio/FileWave.pm festival_panel.prl festival_test.prl synthesiser_test.prl test.pl speech_pm_1.0/README.txt100444 65206 76603 3111 6777424661 13225 0ustar rjccstr Speech::Synthesiser - Speech Output for Perl ============================================ Richard Caley -------------------------------- This is a simple perl extension module which provides a way for perl programs to speak using a speech synthesiser. The interface is intended to allow any synthesiser to be plugged in. An implementation using the festival speech synthesiser is included. For more information about festival, see: http://www.cstr.ed.ac.uk/projects/festival.html If you have problemsw with or suggestions about this module, please sign up for and mail to the festival-talk mailing list, see: http://www.cstr.ed.ac.uk/projects/festival/lists.html In addition to the generic Speech::Synthesiser interface specific interface to festival is provided as Speech::Festival, thi allows more control of the synthesiser. Some simple example scripts are included to show how it works. festival_test.prl Uses the scheme interface. festival_test [auto|loop] scheme1 scheme2 ... Sends each of the scheme forms to festival synthesiser_test.prl Uses the text to speech interface. synthesiser_test text1 text2 ... Synthesizes each text in turn, choosing voices randomly. festival_panel.prl Is a Tk graphical interface allowing you to enter text, select a voice and hear the speech. Run with -q if you don't want it to be quite so chatty. Dispite the name, it's not actually festival specific, if someone implements an interface to another synthesiser it would be trivial to allow access to it from this script. speech_pm_1.0/Makefile.PL100444 65206 76603 622 6777424661 13465 0ustar rjccstruse ExtUtils::MakeMaker; # See lib/ExtUtils/MakeMaker.pm for details of how to influence # the contents of the Makefile that is written. WriteMakefile( 'NAME' => 'Speech::Synthesiser', 'AUTHOR' => 'Richard Caley ', 'ABSTRACT' => 'Speech synthesis from Perl', 'PMLIBDIRS' => [ 'Speech', 'Audio' ], 'VERSION_FROM' => 'Speech/Synthesiser.pm', # finds $VERSION ); speech_pm_1.0/Changes100444 65206 76603 240 6777424661 13002 0ustar rjccstrRevision history for Perl extension Festival. 1.0 Fri Oct 1, rjc - First full version. 0.01 Fri Oct 1 - original version; created by h2xs 1.1.1.1.2.1 speech_pm_1.0/Speech/Festival.pm100444 65206 76603 24030 6777425355 15075 0ustar rjccstrpackage Speech::Festival; $rcsId=' $Id: speech_pm_1.0.tar,v 1.2 2001/04/04 13:12:35 awb Exp $ '; ########################################################################### # # # Interface to festival server. # # # ########################################################################### use strict "subs"; use vars qw($VERSION @ISA @EXPORT @EXPORT_OK); require Exporter; use Socket; @ISA = qw(Exporter); @EXPORT = qw( ); # bootstrap Festival $VERSION; $Speech::Festival::nextstream='festival00000'; $Speech::Festival::speech_error=''; *speech_error = *main::synth_error; $Speech::Festival::end_key='ft_StUfF_key'; $Speech::Festival::OK='OK'; $Speech::Festival::ERROR='ER'; $Speech::Festival::SCHEME='LP'; $Speech::Festival::WAVE='WV'; sub new { my ($class, $host, $port) = @_; $host ||= 'localhost'; $port ||= 1314; my ($self) = [ $host, $port, $Speech::Festival::nextstream++, {} ]; return bless $self, $class; } sub conn { my ($self) = @_; my ($host, $port, $s, $prop) = @$self; my($iaddr, $paddr, $proto); unless ($iaddr = inet_aton($host)) { $speech_error = "no host: $host - $!"; return 0; } $paddr = sockaddr_in($port, $iaddr); $proto = getprotobyname('tcp'); unless(socket($s, PF_INET, SOCK_STREAM, $proto)) { $speech_error = "socket: $!"; return 0; } unless(connect($s, $paddr)) { $speech_error = "connect: $!"; return 0; } my ($old) = select($s); $|=1; select($old); $prop{C}=1; return $s; } sub disconnect { my ($self) = @_; my ($host, $port, $s, $prop) = @$self; if (defined($prop{C}) && $prop{C}) { eval { local $SIG{PIPE} = 'IGNORE'; close($s); } } $prop{C}=0; } sub detach { my ($self) = @_; &DESTROY($self); bless $self, "destroyed Festival"; } sub DESTROY { my ($self) = @_; print "disconnect\n"; disconnect $self; } sub request { my ($self, $scheme, $handler, @info) = @_; my ($host, $port, $s) = @$self; print $s "$scheme\n"; if (defined($handler)) { return handle_results($s, $handler, @info); } } sub wait_for_result { my ($self, $time) = @_; my ($host, $port, $s) = @$self; return waitforsomething($s, $time); } sub result_waiting { my ($self) = @_; my ($host, $port, $s) = @$self; return waitforsomething($s, 0); } sub get_result { my ($self) = @_; my ($host, $port, $s); if (ref($self)) { ($host, $port, $s) = @$self; } else { $s = $self; } my ($type) = ''; if (myread_n($s, $type, 3) != 3) { $speech_error = "Error reading type - $!"; return undef; } chomp $type; return ($type, 'void') if $type eq $OK || $type eq $ERROR; my ($data) = ''; if (myread_upto($s, $data, $end_key) < 0) { $speech_error = "Error reading data - $!"; return undef; } return ($type, $data); } sub handle_results { my ($s, $handler, @info) = @_; my ($nres)=0; my ($state); while (1) { my ($type, $data) = get_result $s; if (!defined($type)) { return undef; } $state = &$handler($type, $data, @info); if ($type eq $OK) { $state=$nres; last; } elsif ($type eq $ERROR) { $state=-1; last; } $nres++; } return $state; } # simple look-ahead IO $Speech::Festival::buffer=''; $Speech::Festival::bend=0; sub waitforsomething { my ($s, $time) = @_; if (length($buffer) > 0) { return 1; } my ($rin, $rout) = ''; vec($rin, fileno($s), 1) = 1; return select($rout = $rin, undef, undef, $time); } sub myread_n { my ($s, $b, $n) = @_; while ($bend < $n) { my ($nr) = sysread($s, $buffer, 1000-$bend, $bend); $bend += $nr if defined($nr); } $_[1] = substr($buffer, 0, $n); $buffer = substr($buffer, $n); $bend -= $n; return $n; } sub myread_upto { my ($s, $b, $key) = @_; my ($checkfrom, $keyat) = 0; while (($keyat=index($buffer, $key, $checkfrom)) <0) { $checkfrom = $bend-length($key) if $bend > length($key); my ($nr) = sysread($s, $buffer, 10000, $bend); $bend += $nr; } $_[1] = substr($buffer, 0, $keyat); $buffer = substr($buffer, $keyat+length($key)); $bend -= $keyat+length($key); return length($_[1]); } # parse scheme $Speech::Festival::scheme_token = '^\\s*(("([^\\]"|[^"]|\s)*")|([-a-zA-Z0-9_+]+)|(\')|(\()|(\)))\\s*'; sub parse_scheme { my ($text) = @_; my ($scheme); if ($text eq '') { return (undef, ""); } elsif ($text =~ /$scheme_token/om) { my ($tok, $str, $strcont, $atom, $sq, $open, $close, $tail) = ($1, $2, $3, $4, $5, $6, $7, $'); # print "XXX", join("//", ($tok, $str, $strcont, $atom, $sq, $open, $close, $tail)), "\n"; if (defined($str)) { return ($str, $tail); } elsif (defined($atom)) { return ($atom, $tail); } elsif (defined($sq)) { my ($quoted, $ttail) = parse_scheme($tail); return ([ 'quote', $quoted ], $ttail); } elsif (defined($open)) { my ($list) = []; while (1) { my ($item, $ttail) = parse_scheme($tail); $tail = $ttail; last if !defined($item) || $item eq ')'; push(@$list, $item); } return ($list, $tail); } elsif (defined($close)) { return ($close, $tail); } } return (undef, substr($text,1)); } # Autoload methods go after =cut, and are processed by the autosplit program. 1; __END__ # Below is the stub of documentation for your module. You better edit it! =head1 NAME Speech::Festival - Communicate with a festival server process. =head1 SYNOPSIS use Festival; $festival = new Festival -host => 'serverhost', -port => 1314; conn $festival; disconnect $festival; request $festival '(some scheme)'; request $festival '(some scheme)', \&result_handler, $args, $for, $handler; if (result_waiting $festival) { # process it } wait_for_result $festival, $timeout; ($type, $data) = get_result $festival; =head1 DESCRIPTION This package provides a simple interface to an instance of the festival speech synthesis system running as a server on some machine. If you want a simple speech-ouput module using this interface see L. Since festival can return an unpredictable number of results from a single request, and since it is useful to process them as they arrive, something a little more complex than a simple remote-procedure-call interface is needed. There are basicly three ways to organise your application's interaction with festival. In the simplest cases you can pass a result handling procedure along with your request. For more control you can process one value at a time by using L to poll for results or L to block until a result is available. In any case results consist of a type and some data. The types are =over 4 =item $Speech::Festival::SCHEME The data is a Scheme expression, for instance a number or a list. =item $Speech::Festival::WAVE The data is a waveform. what format this is in willbe determined by what you have previously told festival. =item $Speech::Festival::OK All the results for this request have been sent. No data. =item $Speech::Festival::ERROR Festival has reported an error. No data. Unfortunatly festival doesn't sen any information about the error, so for details you will have to check the server's log. =back A single festival session (between calls to L and L) talks to a single Scheme interpreter in festival, and hence state will be remembered between requests. =over 4 =item $festival = B Festival 'serverhost', 1314; Create a new festival session which will connect to the given host and port. If ommitted these default to I and I<1314>. =item B $festival; Connect to festival. Returns true if all is well, false otherwise. In the event of an error the variable I<$speech_error> conatains a description of it. =item B $festival; Disconnect from festival. The connection mat be reopened later with L, but any state will have been lost. =item B $festival '(some scheme)'; Send the given Scheme to festival. You should use L, L and L to process the results of the request. =item B $festival '(some scheme)', \&result_handler, $args, $for, $handler; Send the given Scheme to festival asking it to call I<&result_handler> on each result returned. The handler is called as: =back result_handler($type, $data, $args, $for, $handler) =over 4 =item if (B $festival) { # process it } Look to see if there are any results waiting to be processed. This only guarantees that the start of the result has arrived, so a subsequent call to L may not return instantly, but should not block for an extended time. =item B $festival, $timeout; Blocks until festival sends a result. Timeout is in seconds, if can be omitted, in which case the call waits for an unbounded time. Returns false if he call timed out, true if there is a request waiting. Again, this only guarantees that a result has started to arrive. =item ($type, $data) = get_result $festival; Reads a single result. =back =head1 EXAMPLES The code below does some arithmatic the hard way. use Speech::Festival; $festival = new Speech::Festival -host => 'festival-server.mynet'; conn $festival || die "can't talk to festival - $speech_error"; request $festival '(+ 123 456)', sub { my ($type, $data) = @_; print "Scheme Result=$data" if $type eq $Speech::Festival::SCHEME; print "ERROR!\n" if $type eq $Speech::Festival::ERROR; }; =head1 AUTHOR Richard Caley, R.Caley@ed.ac.uk =head1 SEE ALSO L, L, perl(1), festival(1), Festival Documentation =cut speech_pm_1.0/Speech/Festival/Synthesiser.pm100444 65206 76603 13472 6777424661 17424 0ustar rjccstruse Speech::Festival; package Speech::Festival::Synthesiser; use strict "subs"; use vars qw($VERSION @ISA @EXPORT @EXPORT_OK); require Exporter; @ISA = qw(Exporter DynaLoader); # Items to export into callers namespace by default. Note: do not export # names by default without a very good reason. Use EXPORT_OK instead. # Do not simply export all your public functions/methods/constants. @EXPORT = qw( ); @ISA = qw(Speech::Festival); sub new { my ($class, $args) = @_; my ($host) = $$args{-host} || "localhost"; my ($port) = $$args{-port} || 1314; my ($self) = Speech::Festival->new($host, $port); bless $self, $class; } sub set { my ($self, $key, $val) = @_; my ($host, $port, $s, $props) = @$self; $$props{$key} = $val; } sub start { my ($self) = @_; my ($host, $port, $s, $props) = @$self; my ($res) = Speech::Festival::conn($self); print "connected\n"; if ($res) { request $self "(Parameter.set 'Wavefiletype 'riff)", \&shandler; request $self "(tts_return_to_client)", \&shandler; } $res; } sub stop { my ($self) = @_; Speech::Festival::disconnect($self); } sub wavetype { my ($self, $waveimp, @args) = @_; if ($waveimp eq 'Audio::FileWave' && $#args < 0) { @args = ("riff"); } $self->set('WaveImp', $waveimp); $self->set('WaveArgs', [@args]); eval "use $waveimp;"; die $@ if $@; } sub voice_list { my ($self) = @_; my (@res) = (); request $self "(voice.list)", \&ret_lp_handler, \@res; my ($scheme) = Speech::Festival::parse_scheme($res[0]); # @res = split(/\s+/, ($res[0] =~ /\(\s*(.*?)\s*\)/)[0]); return @$scheme; } sub voice { my ($self, $voice) = @_; return request $self "(voice_$voice)", \&shandler; } sub voice_description { my ($self, $voice) = @_; my ($res) = []; request $self "(voice.description '$voice)", \&ret_lp_handler, $res; return Speech::Festival::parse_scheme($$res[0]); } sub speak { my ($self, $text) = @_; my ($host, $port, $s, $props) = @$self; my ($imp, $args) = ( $$props{WaveImp}, $$props{WaveArgs} ); $text =~ s/\\/\\\\/g; $text =~ s/"/\\"/g; my ($scheme) = '(tts_text "' . $text . '" "text")'; my ($n) = request $self $scheme, \&whandler, $imp, $args; $n-- if $n>0; return $n; } sub synth_description { my ($self) = @_; my ($host, $port, $s, $props) = @$self; my ($imp, $args) = ( $$props{WaveImp}, $$props{WaveArgs} ); my ($n) = request $self '(intro)', \&whandler, $imp, $args; $n-- if $n>0; return $n; } sub shandler { my ($type, $data, @info) = @_; chomp $data; print "shandler got $type:",substr($data,0,10),"\n" if defined($main::verbose) && $main::verbose>0; } sub whandler { my ($type, $data, $imp, $args) = @_; chomp $data; print "whandler got $type:",substr($data,0,10),"\n" if defined($main::verbose) && $main::verbose>0; if ($type eq $Speech::Festival::WAVE) { my ($wave) = $imp->new($data, @$args); play $wave; } } sub ret_lp_handler { my ($type, $data, $res) = @_; chomp $data; print "ret_lp_handler got $type:",substr($data,0,10),"\n" if defined($main::verbose) && $main::verbose>0; if ($type eq $Speech::Festival::SCHEME) { push(@$res, $data); } } 1; __END__ =head1 NAME Speech::Festival::Synthesiser - Simple text-to-speech using festival. =head1 SYNOPSIS use Speech::Speech::Festival::Synthesiser; $festival = new Speech::Festival::Synthesiser 'serverhost', 1314; start $festival; stop $festival; @voices = voice_list $festival; voice $festival "myvoice"; voice_description $festival; synth_description $festival; speak $festival $text; =head1 DESCRIPTION L provides a simple way to add speech to a perl application. It is a sub-class of L and so you can use any of the methods in that class to manipulate festival if necessary. This package conform to the interface defined by L and so provides a synthesiser for use throughthat interface. Actual sound output is provided by an auiliary class, by default L which runs an external program to play sound, but you can replace it with another class if you have a better way of playing sounds (eg a perl extension providing sound output), see the documentation for L for the interface an alternative should provide should implement. If you do use L you may need to set up the command it uses to play sounds, see the documentation for L. =over 4 =item $festival = B Speech::Festival::Synthesiser -host => 'serverhost', -port => 1314, ... Create a new festival session which will connect to the given host and port. If ommitted these default to I and I<1314>. =item B $festival; Connect to festival and inistilise the session. Returns true if all is well, false otherwise. In the event of an error the variable I<$festival_error> conatains a description of it. =item B $festival; Disconnect from festival. The connection may be reopened later with L, but any state will have been lost. =item @voices = B $festival; Return a list of the voices available from this server. =item B $festival "myvoice"; Select a voice. =item B $festival; Returns the description of the current voice. =item B $festival; Synthesize the standard festival introductory text. =item B $festival $text; Speak the given text. =back =head1 AUTHOR Richard Caley, R.Caley@ed.ac.uk =head1 SEE ALSO L, L, L, perl(1), festival(1), Festival Documentation =cut speech_pm_1.0/Speech/Synthesiser.pm100444 65206 76603 11620 6777424661 15640 0ustar rjccstr package Speech::Synthesiser; $VERSION = '1.0'; use strict "subs"; use vars qw($VERSION @ISA @EXPORT @EXPORT_OK); require Exporter; @ISA = qw(Exporter DynaLoader); # Items to export into callers namespace by default. Note: do not export # names by default without a very good reason. Use EXPORT_OK instead. # Do not simply export all your public functions/methods/constants. @EXPORT = qw( ); *synth_error = *main::synth_error; sub new { my ($class, %props) = @_; my ($type) = (defined($props{-type}) ? "Speech::$props{-type}::Synthesiser" : "Speech::Dummy::Synthesiser"); eval "use $type;" unless $type eq "Speech::Dummy::Synthesiser"; die "$@" if $@; my ($self) = $type->new(\%props); my ($waveimp) = $props{-waveimp} || "Audio::FileWave"; my ($waveargs) = $props{-waveargs} || []; wavetype $self $waveimp, @waveargs; $self; } sub set { my ($self, $key, $val) = @_; } sub start { my ($self) = @_; 0; } sub stop { my ($self) = @_; 0; } sub wavetype { my ($self, $waveimp, @args) = @_; $self->set('WaveImp', $waveimp); $self->set ('WaveArgs', [@args]); eval "use $waveimp;"; die $@ if $@; } sub voice_list { my ($self) = @_; return ("default"); } sub voice { my ($self, $voice) = @_; } sub voice_description { my ($self, $voice) = @_; return "Default voice." } sub speak { my ($self, $text) = @_; print STDOUT "Imagine I just said '$text'\n"; return 1; } sub synth_description { my ($self) = @_; print STDOUT "This is some synthesiser or other.\n"; return 1; } package Speech::Dummy::Synthesiser; @ISA=( "Speech::Synthesiser" ); sub new { my ($class, @args) = @_; $self = []; bless $self, $class; } sub start { return 1; } 1; __END__ =head1 NAME Speech::Synthesiser - Generic speech syntheiser interface =head1 SYNOPSIS use Speech::Synthesiser; $synth = new Speech::Synthesiser -type => 'SynthName', # other args ; start $synth; stop $synth; @voices = voice_list $synth; voice $synth "myvoice"; intro $synth; speak $synth $text; =head1 DESCRIPTION L provides a simple way to add speech to a perl application. It is a generic class which can be used to talk to any speech synthesiser given a suitable interface module. Actual sound output is provided by an auiliary class, by default L which runs an external program to play sound, but you can replace it with another class if you have a better way of playing sounds (eg a perl extension providing sound output), see the documentation for L for the interface an alternative should provide should implement. If you do use L you may need to set up the command it uses to play sounds, see the documentation for L. =over 4 =item $synth = B Speech::Synthesiser -type => 'SynthName', ARGS; Create a synthesiser of the named type. Looks for a package C. All arguments are passed to the creation function for that class. The following arguments have special meaning to the C C method. =over 4 =item -waveimp CLASS CLASS is the name of a perl package which implements wave playing. If not given it defaults to C. =item -waveargs [ @ARGS ] @ARGS are passed to the C method of the wave class. For C this defaults to C<( "riff" )>. =back =item B $synth; Do whatever is ncecessary to prepar ethe synthesiser fo work. Returns true if all is well, false otherwise. In the event of an error the variable I<$synth_error> conatains a description of it. =item B $synth; Close down the synthesiser, releasing any resources it holds. The synthesiser may be restarted with L, but any state may have been lost. =item @voices = B $synth; Return a list of the voices available from this synthesiser. =item B $synth "myvoice"; Select a voice. =item B $synth; Returns a description of the voice. =item B $synth; Synthesize a description of the synthesiser, =item B $synth $text; Speak the given text. Not much more to be said really:-). =back =head1 EXAMPLE The following should talk to you, all else being equal. Uses the festival synthesiser, so you will need to run a festival server on the named machine. use Speech::Synthesiser; $synth = new Speech::Synthesiser -type => 'Festival', -host => 'festival-server.mynet'; start $synth || die "can't talk to festival - $synth_error"; speak $synth "We are perl, prepare for assimilation."; =head1 AUTHOR Richard Caley, R.Caley@ed.ac.uk =head1 SEE ALSO L, L, L, perl(1), festival(1), Festival Documentation =cut speech_pm_1.0/Audio/FileWave.pm100444 65206 76603 4734 6777424661 14644 0ustar rjccstr package Audio::FileWave; $Audio::FileWave::tmpname="/tmp/${$}fwv0000000"; %Audio::FileWave::command = ( "DEFAULT" => "na_play WAVE" ); sub import { my ($class) = @_; } sub new { my ($class, $data, $type) = @_; my ($file) = $tmpname++; $type ||= 'riff'; my ($self) = [$file, $type]; local(*WV); open(WV, ">$file") || return "can't write to $file - $!"; syswrite(WV, $data, length($data)); close(WV); bless $self, $class; } sub set_play_command { my ($class, $type, $command) = @_; $command{$type} = $command; } sub free { my ($self) = @_; &DESTROY($self); bless $self, "Freed ".ref($self); } sub DESTROY { my ($self) = @_; my ($file) = @$self; unlink($file); } sub play { my ($self) = @_; my ($file, $type) = @$self; my ($command) = $command{$type}; $command = $command{DEFAULT} unless defined($command); $command =~ s/WAVE/$file/g; system($command); } 1; __END__ =head1 NAME Audio::FileWave - Simple waveform abstraction for use with Audio::TTS =head1 SYNOPSIS $wave = new Audio::FileWave $type, $data; set_play_command Audio::FileWave $type $command; play $wave; free $wave; =head1 DESCRIPTION This is a simple class which represents a waveform. It is designed for use by the L module, if you want to play waveforms some other way you need to supply another class which implements this interface. This class maintains a table mapping file types to play commands. When the command is called the string `I' is replaced by the filename of the data. If you don't set up a command for the waveform type L requests, the default command `I' is used. The L command is part of the I and so if you have festival you should also have it. =over 4 =item B Festival::FileWave $type $command; Sets the play command to be used for files of the given type. =item $wave = B Festival::FileWave $type, $data; Create a waveform from the given data which is of type $type. Actually saves it to a file. =item B $wave; Free the resources used for the waveform. (unlinks the file). =item B $wave; Play the waveform. Calls the command associated with the file type. =back =head1 AUTHOR Richard Caley, R.Caley@ed.ac.uk =head1 SEE ALSO L, perl(1), festival(1), Festival Documentation =cut speech_pm_1.0/festival_panel.prl100444 65206 76603 22414 6777424661 15271 0ustar rjccstr#!/usr/local/bin/perl -w ########################################################################### # # # Simple Tk panel which allows you to synthesise text. # # # ########################################################################### sub usage { my ($ret) = @_; print <=0 && $ARGV[0] =~ /^-/) { if ($ARGV[0] =~ /^-+s(erver)?/) { shift @ARGV; $server = shift @ARGV; } elsif ($ARGV[0] =~ /^-+p(ort)?/) { shift @ARGV; $port = shift @ARGV; } elsif ($ARGV[0] =~ /^-+t(imeout)?/) { shift @ARGV; $timeout = shift @ARGV; } elsif ($ARGV[0] =~ /^-+c(ommand)?/) { shift @ARGV; $command = shift @ARGV; } elsif ($ARGV[0] =~ /^-+q(uiet)?/) { shift @ARGV; $quiet=1; } elsif ($ARGV[0] =~ /^-+[h?](elp)?/) { shift @ARGV; usage(0); } else { usage(1); } } if ($#ARGV >=0) { $text = join("\n", @ARGV); } $festival=undef; $main = new MainWindow; $icon_im = load_image('Tk/tranicon.gif'); $powered_im = load_image('Tk/anim.gif'); if (defined($icon_im)) { $icon_w=$main->Toplevel( -background => undef, -width => $icon_im->width(), -height => $icon_im->height() ); $icon_lab = $icon_w->Label(-image => $icon_im); $icon_lab->pack(); $icon_lab->bind('' => sub { $main->deiconify;}); $main->iconwindow($icon_w); } $main->title('Festival Panel'); $text_field = $main->Text( -width => 20, -height => 4, -background => 'antiquewhite' ); $text_field->insert("0.0", $text); $fp_frame = $main->Frame(); $festival_frame = $fp_frame->Frame(); $festival_lab = $festival_frame->Label( -font => '-*-helvetica-bold-r-*-*-*-120-*-*-*-*-*-*', -background => $main->cget(-background), -text => 'Server' ); $festival_field = $festival_frame->Entry( -width => 15, -textvariable => \$server, -background => 'antiquewhite' ); $festival_field->bind('' => sub { $server_last_value = $server; }); $festival_field->bind('' => sub { server_changed() unless $server_last_value eq $server; }); $port_frame = $fp_frame->Frame(); $port_lab = $port_frame->Label( -font => '-*-helvetica-bold-r-*-*-*-120-*-*-*-*-*-*', -background => $main->cget(-background), -text => 'Port' ); $port_field = $port_frame->Entry( -width => 5, -textvariable => \$port, -background => 'antiquewhite' ); $port_field->bind('' => sub { $port_last_value = $port; }); $port_field->bind('' => sub { server_changed() unless $port_last_value eq $port; }); if(defined($powered_im)) { $powered_frame=$fp_frame->Frame(); $powered_lab=$powered_frame->Label(-image => $powered_im); } $message_lab = $main->Label(-anchor => 'w', -font => '-*-helvetica-bold-r-*-*-*-120-*-*-*-*-*-*', -background => $main->cget(-background), -textvariable => \$message, -relief => 'groove', -borderwidth => 2, -width => 40 ); $buttons_frame = $main->Frame( ); $voice_sel = $buttons_frame->Menubutton(-text => 'Voice', -relief => 'raised', -indicatoron => 1, -width => 12 ); $quit_button = $buttons_frame->Button(-text => 'Quit', -command => sub{exit} ); $intro_button = $buttons_frame->Button(-text => 'Intro', -command => \&intro ); $describe_button = $buttons_frame->Button(-text => 'Describe', -command => \&describe ); $speak_button = $buttons_frame->Button(-text => 'Speak', -command => \&speak ); # $main->pack(-side => 'top', -fill => 'x'); # $title->pack(-side => 'top', -fill => 'x'); $text_field->pack(-side => 'top', -fill => 'x'); $fp_frame->pack(-side => 'top', -fill => 'x'); $festival_frame->pack(-side => 'left', -fill => 'x'); $festival_lab->pack(-side =>'top'); $festival_field->pack(-side =>'bottom'); $port_frame->pack(-side => 'left', -fill => 'x'); $port_lab->pack(-side =>'top'); $port_field->pack(-side =>'bottom'); if(defined($powered_im)) { $powered_frame->pack(-side => 'right', -fill => 'x'); $powered_lab->pack(-side => 'bottom', -fill => 'x'); } $message_lab->pack(-side => 'top', -fill => 'x'); $buttons_frame->pack(-side => 'top', -fill => 'x'); $voice_sel->pack(-side =>'left'); $describe_button->pack(-side => 'left'); $intro_button->pack(-side => 'left'); $quit_button->pack(-side =>'right'); $speak_button->pack(-side => 'right'); server_changed(); MainLoop; exit(0); sub message { my ($text) = @_; if (!defined($text)) { $text = ''; } $message=$text; } sub server_changed { stop $festival if defined($festival); $festival = new Speech::Synthesiser -type => 'Festival', -host => $server, -port => $port; set_play_command Audio::FileWave "riff", $command; eval { local $SIG{ALRM} = sub { die "alarm clock restart" }; alarm $timeout; if (!start $festival) { die "couldn't connect $synth_error"; } }; alarm 0; if ($@) { if ($@ =~ /alarm clock restart/) { message("couldn't connect in $timeout seconds"); $festival=undef; festival_ok(); return; } else { message("couldn't connect $synth_error"); $festival=undef; festival_ok(); return; } } my (@voices) = voice_list $festival; my ($menu) = $voice_sel->cget(-menu); $menu->delete(0,'last') if defined($menu); for $v (@voices) { my ($v) = $v; $voice_sel->command(-label => $v, -command => sub {set_voice($v);} ); } set_voice($voices[0]); festival_ok(); } sub speak { my ($text) = @_; if (defined($festival)) { $speak_button->configure(-relief => 'sunken'); $speak_button->update(); my ($n) = $festival->speak(defined($text)?$text:get_text()); if ($n >= 0) { message("$n Chunk".($n==1?"":"s")); } else { message("Festival error"); } $speak_button->configure(-relief => 'raised'); $speak_button->update(); } } sub intro { if (defined($festival)) { $intro_button->configure(-relief => 'sunken'); $intro_button->update(); $speak_button->configure(-relief => 'sunken'); $speak_button->update(); my ($n) = $festival->synth_description(); message("$n Chunk".($n==1?"":"s")); $speak_button->configure(-relief => 'raised'); $speak_button->update(); $intro_button->configure(-relief => 'raised'); $intro_button->update(); } } sub set_voice { my ($v) = @_; if (defined($festival)) { $voice_sel->configure(-relief => 'sunken'); $voice_sel->update(); message("selecting $v"); voice $festival $v; $current_voice = $v; $voice_sel->configure(-text => $v); my ($vname) = $v; $vname =~ s/([^_]*)_(.*)/{my ($a, $b) = ($1, $2); $a =~ s|(.)|$1 |g; "$a _ $b"; }/e; speak("voice is $vname") unless $quiet; $voice_sel->configure(-relief => 'raised'); $voice_sel->update(); describe($v) unless $quiet; } } sub festival_ok { if (defined($festival)) { $voice_sel->configure(-state => 'normal'); $intro_button->configure(-state => 'normal'); $speak_button->configure(-state => 'normal'); } else { $voice_sel->configure(-state => 'disabled'); $intro_button->configure(-state => 'disabled'); $speak_button->configure(-state => 'disabled'); } } sub get_text { return $text_field->get("0.0", "end"); } sub load_image { my ($fn) = @_; foreach $d (@INC) { if ( -f "$d/$fn") { my ($im) = $main->Photo('-format' => 'GIF', -file => "$d/$fn"); return $im; } } return undef; } sub describe { my ($voice) = @_; $voice = $current_voice unless defined($voice); if (defined($festival)) { $describe_button->configure(-relief => 'sunken'); $describe_button->update(); my ($desc) = voice_description $festival $voice; my ($d) = $desc->[1]->[3]->[1]; speak($d); $describe_button->configure(-relief => 'raised'); $describe_button->update(); } } speech_pm_1.0/festival_test.prl100444 65206 76603 2753 6777424661 15135 0ustar rjccstr#!/usr/local/bin/perl -w ########################################################################### # # # Simple test of festival interface module. # # # ########################################################################### use Speech::Festival; $f = new Speech::Festival; if (!conn $f) { die "error connecting - $synth_error"; } $which = $ARGV[0]; shift @ARGV; while ($#ARGV>=0) { if ($which eq 'loop') { print "sending $ARGV[0]\n"; request $f "$ARGV[0]"; while (1) { if (result_waiting $f) { print "result waiting\n"; } else { print "waiting\n"; wait_for_result $f; } print "read\n"; my ($type, $data) = get_result $f; die "get error - $!" unless defined($type); print "type=$type data=$data\n"; last if $type eq $Speech::Festival::OK || $type eq $Speech::Festival::ERROR; } } else { print "sending\n"; my ($nres) = request $f "$ARGV[0]", \&handler; print "NRESULTS=$nres\n"; } shift @ARGV; } disconnect $f; exit(0); sub handler { my ($type, $data) = @_; if ($type eq $Speech::Festival::SCHEME) { chomp $data; print "LISP RESULT='$data'\n"; } elsif ($type eq $Speech::Festival::OK) { } elsif ($type eq $Speech::Festival::ERROR) { print "ERROR FROM FESTIVAL\n"; } return 0; } speech_pm_1.0/synthesiser_test.prl100444 65206 76603 2252 6777424662 15673 0ustar rjccstr#!/usr/local/bin/perl -w ########################################################################### # # # Simple test of Festival:TTS module. # # # ########################################################################### BEGIN { unshift @INC, "."; } use Speech::Synthesiser; $synth = new Speech::Synthesiser -type => 'Festival', -host => "localhost" ; # You will need to replace this with youre preferred play command. set_play_command Audio::FileWave "riff", "na_play WAVE "; if (!start $synth) { die "error starting synthesiser - $synth_error"; } @voices = voice_list $synth; print "Available Voices: ", join(" ", @voices), "\n"; while ($#ARGV>=0) { print "\n"; $voice = $voices[rand($#voices+1)]; print "Selecting $voice\n"; voice $synth $voice; print "Sending: $ARGV[0]\n"; my ($nres) = speak $synth $ARGV[0]; print "Number of Chunks: $nres\n"; shift @ARGV; } stop $synth; exit(0); speech_pm_1.0/test.pl100444 65206 76603 1215 6777424661 13046 0ustar rjccstr# Before `make install' is performed this script should be runnable with # `make test'. After `make install' it should work as `perl test.pl' ######################### We start with some black magic to print on failure. # Change 1..1 below to 1..last_test_to_print . # (It may become useful if the test is moved to ./t subdirectory.) BEGIN { $| = 1; print "1..1\n"; } END {print "not ok 1\n" unless $loaded;} use Festival; $loaded = 1; print "ok 1\n"; ######################### End of black magic. # Insert your test code below (better if it prints "ok 13" # (correspondingly "not ok 13") depending on the success of chunk 13 # of the test code): festival/examples/songs/Makefile0000644000175000017500000000524707603105322015414 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=../.. DIRNAME=examples/songs BUILD_DIRS= ALL_DIRS=$(BUILD_DIRS) SONGS = america1.xml america2.xml america3.xml america4.xml \ daisy.xml doremi.xml lochlomond.xml \ spice1.xml spice2.xml spice3.xml spice4.xml FILES=Makefile $(SONGS) include $(TOP)/config/common_make_rules festival/examples/songs/america1.xml0000644000175000017500000000145207603077776016177 0ustar awbawb oh beautiful for spacious skies for amber waves of grain festival/examples/songs/america2.xml0000644000175000017500000000145007603077776016176 0ustar awbawb oh beautiful for spacious skies for amber waves of grain festival/examples/songs/america3.xml0000644000175000017500000000145307603077776016202 0ustar awbawb oh beautiful for spacious skies for amber waves of grain festival/examples/songs/america4.xml0000644000175000017500000000145507603077776016205 0ustar awbawb oh beautiful for spacious skies for amber waves of grain festival/examples/songs/daisy.xml0000644000175000017500000000546707603077776015640 0ustar awbawb Daisy Daisy Give me your answer do I'm half crazy all for the love of you we don't need a stylish marriage I can't afford a carriage but you'll look sweet upon the seat of a bicycle built for two festival/examples/songs/doremi.xml0000644000175000017500000000120107603077776015764 0ustar awbawb doe ray me fah sew lah tee doe festival/examples/songs/lochlomond.xml0000644000175000017500000000514107603077776016652 0ustar awbawb Oh you'll take the high road and I'll take the low road and I'll be in Scotland before ye but me and my true love will never meet again on the bonnie bonnie banks of lock lahman festival/examples/songs/spice1.xml0000644000175000017500000000277707603077776015714 0ustar awbawb if you wanna be my lover you gotta get with my friends make it last forever friendship never in n festival/examples/songs/spice2.xml0000644000175000017500000000311307603077776015676 0ustar awbawb if you wanna be my lover you gotta get with my gotta get with my friends forever friendship never in n festival/examples/songs/spice3.xml0000644000175000017500000000301107603077776015674 0ustar awbawb if you wanna be my lover you gotta get with my friends make it last forever friendship never in n festival/examples/songs/spice4.xml0000644000175000017500000000300407603077776015677 0ustar awbawb if you wanna be my lover you gotta get with my friends make it last forever friendship never in n festival/bin/Makefile0000644000175000017500000000705707607314565013235 0ustar awbawb ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1994,1995,1996 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## ## Author: Richard Caley (rjc@cstr.ed.ac.uk) ## ## Date: June 1997 ## ## --------------------------------------------------------------------- ## ## Makefile for bin directory. ## ## ## ## If we are staticly linked, we link to things in main, otherwise we ## ## write scripts which set LD_LIBRARY_PATH. ## ## ## ## Things in scripts are preprocessed. ## ## ## ########################################################################### TOP=.. DIRNAME=bin FILES= Makefile VCLocalRules ALL = .remove_links .link_main .process_scripts text2wave include $(TOP)/config/common_make_rules include $(EST)/config/rules/bin_process.mak text2wave: @ cp -p $(TOP)/examples/text2wave . festival/bin/VCLocalRules0000644000175000017500000000010007320045414013756 0ustar awbawb# don't make text2wave text2wave: @echo "no text to wave" festival/doc/Makefile0000644000175000017500000001142407262605651013217 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=.. DIRNAME=doc DOCNAME=festival DSSL_SANE_DB=$(EST)/doc/sane_to_docbook.dsl DSSSL_HTML=$(EST)/doc/cstr.dssl DSSSL=$(DSSSL_HTML) # Temporarilly we explicitly list the programs which have been documented MAIN_TO_DOCUMENT= EXAMPLE_TO_DOCUMENT= FESTIVAL=$(TOP)/bin/festival --libdir $(TOP)/lib # Include some of EST documentation. DOCXX_EXTRA_FILES = DOCXXIMAGES = edcrest.gif cstr.gif \ est.jpg est_small.jpg \ festival.jpg festival_small.jpg festival_tiny.jpg DOCXXFILES= classHeader.inc hierHeader.inc indexHeader.inc \ banner.inc $(DOCXXIMAGES) MANPAGES = festival.head festival.tail \ festival_client.head festival_client.tail #SGMLFILES = festival.sgml \ # introductory.sgml basics.sgml core.sgml advanced.sgml programming.sgml FILES=Makefile $(MANPAGES) festival.texi $(SGMLFILES) refcard.tex $(DOCXXFILES) LOCAL_CLEAN = *.aux *.cp *.fn *.ky *.log *.pg *.toc *.tp *.vr ALL = festival.1 festival_client.1 include $(TOP)/config/common_make_rules include $(EST)/config/rules/doc.mak %.1 : %.head %.options %.tail cat $^ >$@ %.options : $(TOP)/src/main/% $(TOP)/bin/$* -man_options >$@ %.options : $(TOP)/src/main/%.exe $(TOP)/bin/$* -man_options >$@ festival.info: festival.texi festfunc.texi festvars.texi festfeat.texi @ if [ ! -d info ] ; \ then mkdir -p info ; fi @ sed 's/@url{/@file{/g' info/festival.texi @ cp festfunc.texi info/festfunc.texi @ cp festvars.texi info/festvars.texi @ cp festfeat.texi info/festfeat.texi ( cd info; makeinfo festival.texi ) @ rm info/*.texi # texi2html is available from http://wwwcn.cern.ch/dci/texi2html/ festival.html: festival.texi festfunc.texi festvars.texi @ if [ ! -d html ] ; \ then mkdir -p html ; fi (cd html; texi2html -number -split_chapter ../festival.texi) # give the html files background color of white @ for i in html/*.html ; \ do \ sed 's///' $$i >ttt.html; \ mv ttt.html $$i ; \ done festival.ps: festival.dvi dvips -f festival.dvi >festival.ps festival.dvi: festival.texi festfunc.texi festvars.texi tex festival.texi texindex festival.cp tex festival.texi doc: festival.ps festival.html festival.info festfunc.texi festvars.texi festfeat.texi: $(TOP)/src/main/festival echo "(load_library \"festdoc.scm\") (make-doc)" | $(FESTIVAL) refcard.dvi: refcard.tex latex refcard.tex refcard.ps: refcard.dvi dvips -f -t landscape refcard.dvi >refcard.ps festival/doc/festival.head0000644000175000017500000000132207262605651014213 0ustar awbawb.TH FESTIVAL 1 "6th Apr 1998" .SH NAME festival \- a text-to-speech system. .SH SYNOPSIS .B festival .I [options] .I [file0] .I [file1] .I ... .SH DESCRIPTION Festival is a general purpose text-to-speech system. As well as simply rendering text as speech it can be used in an interactive command mode for testing and developing various aspects of speech synthesis technology. Festival has two major modes, command and tts (text-to-speech). When in command mode input (from file or interactively) is interpreted by the command interpreter. When in tts mode input is rendered as speech. When in command mode filenames that start with a left paranthesis are treated as literal commands and evaluated. .SH OPTIONS festival/doc/festival.tail0000644000175000017500000000112307262605651014242 0ustar awbawb.SH BUGS More than you can imagine. A manual with much detail (though not complete) is available in distributed as part of the system and is also accessible at .br http://www.cstr.ed.ac.uk/projects/festival/manual/ Although we cannot guarantee the time required to fix bugs, we would appreciated it if they were reported to .br festival-bug@cstr.ed.ac.uk .SH AUTHOR Alan W Black, Richard Caley and Paul Taylor .br (C) Centre for Speech Technology Research, 1996-1998 .br University of Edinburgh .br 80 South Bridge .br Edinburgh EH1 1HN .br http://www.cstr.ed.ac.uk/projects/festival.html festival/doc/festival_client.head0000644000175000017500000000100707262605651015551 0ustar awbawb.TH FESTIVAL 1 "1st Aug 1997" .SH NAME festival_client \- client access to festiva; text-to-speech server mode .SH SYNOPSIS .B festival_client .I [options] .I [file0] .I [file1] .I ... .SH DESCRIPTION Allows sending commands and/or test to a festival test-to-speech server. Festival may be started in server mode using the option .BI --server festival-clinet allows waveforms ot be generated faster as not time is required for start up. See the festival manual for more details on recommended use. .SH OPTIONS festival/doc/festival_client.tail0000644000175000017500000000055407262605651015607 0ustar awbawb.SH BUGS This still isn't as general as it should be. It seems to require lots of special little options which suggests the general options aren't right. .SH AUTHOR Alan W Black and Paul Taylor .br (C) Centre for Speech Technology Research .br University of Edinburgh .br 80 South Bridge .br Edinburgh EH1 1HN .br http://www.cstr.ed.ac.uk/projects/festival.html festival/doc/festival.texi0000644000175000017500000127274010054646141014272 0ustar awbawb\input texinfo @c -*-texinfo-*- @c %**start of header @setfilename festival.info @settitle Festival Speech Synthesis System @finalout @setchapternewpage odd @c %**end of header @c This document was modelled on the numerous examples of texinfo @c documentation available with GNU software, primarily the hello @c world example, but many others too. I happily acknowledge their @c aid in producing this document -- awb @set EDITION 1.4 @set VERSION 1.4.3 @set UPDATED 27th December 2002 @ifinfo This file documents the @code{Festival} Speech Synthesis System a general text to speech system for making your computer talk and developing new synthesis techniques. Copyright (C) 1996-2004 University of Edinburgh Permission is granted to make and distribute verbatim copies of this manual provided the copyright notice and this permission notice are preserved on all copies. @ignore Permission is granted to process this file through TeX, or otherwise and print the results, provided the printed document carries copying permission notice identical to this one except for the removal of this paragraph (this paragraph not being relevant to the printed manual). @end ignore Permission is granted to copy and distribute modified versions of this manual under the conditions for verbatim copying, provided that the entire resulting derived work is distributed under the terms of a permission notice identical to this one. Permission is granted to copy and distribute translations of this manual into another language, under the above conditions for modified versions, except that this permission notice may be stated in a translation approved by the authors. @end ifinfo @titlepage @title The Festival Speech Synthesis System @subtitle System documentation @subtitle Edition @value{EDITION}, for Festival Version @value{VERSION} @subtitle @value{UPDATED} @author by Alan W Black, Paul Taylor and Richard Caley. @page @vskip 0pt plus 1filll Copyright @copyright{} 1996-2004 University of Edinburgh, all rights reserved. Permission is granted to make and distribute verbatim copies of this manual provided the copyright notice and this permission notice are preserved on all copies. Permission is granted to copy and distribute modified versions of this manual under the conditions for verbatim copying, provided that the entire resulting derived work is distributed under the terms of a permission notice identical to this one. Permission is granted to copy and distribute translations of this manual into another language, under the above conditions for modified versions, except that this permission notice may be stated in a translation approved by the University of Edinburgh @end titlepage @node Top, , , (dir) @ifinfo This file documents the @emph{Festival Speech Synthesis System} @value{VERSION}. This document contains many gaps and is still in the process of being written. @end ifinfo @menu * Abstract:: initial comments * Copying:: How you can copy and share the code * Acknowledgements:: List of contributors * What is new:: Enhancements since last public release * Overview:: Generalities and Philosophy * Installation:: Compilation and Installation * Quick start:: Just tell me what to type * Scheme:: A quick introduction to Festival's scripting language Text methods for interfacing to Festival * TTS:: Text to speech modes * XML/SGML mark-up:: XML/SGML mark-up Language * Emacs interface:: Using Festival within Emacs Internal functions * Phonesets:: Defining and using phonesets * Lexicons:: Building and compiling Lexicons * Utterances:: Existing and defining new utterance types Modules * Text analysis:: Tokenizing text * POS tagging:: Part of speech tagging * Phrase breaks:: Finding phrase breaks * Intonation:: Intonations modules * Duration:: Duration modules * UniSyn synthesizer:: The UniSyn waveform synthesizer * Diphone synthesizer:: Building and using diphone synthesizers * Other synthesis methods:: other waveform synthesis methods * Audio output:: Getting sound from Festival * Voices:: Adding new voices (and languages) * Tools:: CART, Ngrams etc * Building models from databases:: Adding new modules and writing C++ code * Programming:: Programming in Festival (Lisp/C/C++) * API:: Using Festival in other programs * Examples:: Some simple (and not so simple) examples * Problems:: Reporting bugs. * References:: Other sources of information * Feature functions:: List of builtin feature functions. * Variable list:: Short descriptions of all variables * Function list:: Short descriptions of all functions * Index:: Index of concepts. @end menu @node Abstract, Copying, , Top @chapter Abstract This document provides a user manual for the Festival Speech Synthesis System, version @value{VERSION}. Festival offers a general framework for building speech synthesis systems as well as including examples of various modules. As a whole it offers full text to speech through a number APIs: from shell level, though a Scheme command interpreter, as a C++ library, and an Emacs interface. Festival is multi-lingual, we have develeoped voices in many languages including English (UK and US), Spanish and Welsh, though English is the most advanced. The system is written in C++ and uses the Edinburgh Speech Tools for low level architecture and has a Scheme (SIOD) based command interpreter for control. Documentation is given in the FSF texinfo format which can generate a printed manual, info files and HTML. The latest details and a full software distribution of the Festival Speech Synthesis System are available through its home page which may be found at @example @url{http://www.cstr.ed.ac.uk/projects/festival.html} @end example @node Copying, Acknowledgements, Abstract, Top @chapter Copying @cindex restrictions @cindex redistribution As we feeel the core system has reached an acceptable level of maturity from 1.4.0 the basic system is released under a free lience, without the commercial restrictions we imposed on early versions. The basic system has been placed under an X11 type licence which as free licences go is pretty free. No GPL code is included in festival or the speech tools themselves (though some auxiliary files are GPL'd e.g. the Emacs mode for Festival). We have deliberately choosen a licence that should be compatible with our commercial partners and our free software users. However although the code is free, we still offer no warranties and no maintenance. We will continue to endeavor to fix bugs and answer queries when can, but are not in a position to guarantee it. We will consider maintenance contracts and consultancy if desired, please contacts us for details. Also note that not all the voices and lexicons we distribute with festival are free. Particularly the British English lexicon derived from Oxford Advanced Learners' Dictionary is free only for non-commercial use (we will release an alternative soon). Also the Spanish diphone voice we relase is only free for non-commercial use. If you are using Festival or the speech tools in commercial environment, even though no licence is required, we would be grateful if you let us know as it helps justify ourselves to our various sponsors. The current copyright on the core system is @example The Festival Speech Synthesis System: version 1.4.3 Centre for Speech Technology Research University of Edinburgh, UK Copyright (c) 1996-2004 All Rights Reserved. Permission is hereby granted, free of charge, to use and distribute this software and its documentation without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of this work, and to permit persons to whom this work is furnished to do so, subject to the following conditions: 1. The code must retain the above copyright notice, this list of conditions and the following disclaimer. 2. Any modifications must be clearly marked as such. 3. Original authors' names are not deleted. 4. The authors' names are not used to endorse or promote products derived from this software without specific prior written permission. THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. @end example @node Acknowledgements, What is new, Copying, Top @chapter Acknowledgements @cindex acknowledgements @cindex thanks The code in this system was primarily written by Alan W Black, Paul Taylor and Richard Caley. Festival sits on top of the Edinburgh Speech Tools Library, and uses much of its functionality. Amy Isard wrote a synthesizer for her MSc project in 1995, which first used the Edinburgh Speech Tools Library. Although Festival doesn't contain any code from that system, her system was used as a basic model. Much of the design and philosophy of Festival has been built on the experience both Paul and Alan gained from the development of various previous synthesizers and software systems, especially CSTR's Osprey and Polyglot systems @cite{taylor91} and ATR's CHATR system @cite{black94}. However, it should be stated that Festival is fully developed at CSTR and contains neither proprietary code or ideas. Festival contains a number of subsystems integrated from other sources and we acknowledge those systems here. @section SIOD @cindex SIOD @cindex Scheme @cindex Paradigm Associates The Scheme interpreter (SIOD -- Scheme In One Defun 3.0) was written by George Carrett (gjc@@mitech.com, gjc@@paradigm.com) and offers a basic small Scheme (Lisp) interpreter suitable for embedding in applications such as Festival as a scripting language. A number of changes and improvements have been added in our development but it still remains that basic system. We are grateful to George and Paradigm Associates Incorporated for providing such a useful and well-written sub-system. @example Scheme In One Defun (SIOD) COPYRIGHT (c) 1988-1994 BY PARADIGM ASSOCIATES INCORPORATED, CAMBRIDGE, MASSACHUSETTS. ALL RIGHTS RESERVED Permission to use, copy, modify, distribute and sell this software and its documentation for any purpose and without fee is hereby granted, provided that the above copyright notice appear in all copies and that both that copyright notice and this permission notice appear in supporting documentation, and that the name of Paradigm Associates Inc not be used in advertising or publicity pertaining to distribution of the software without specific, written prior permission. PARADIGM DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL PARADIGM BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. @end example @section editline Because of conflicts between the copyright for GNU readline, for which an optional interface was included in earlier versions, we have replace the interface with a complete command line editing system based on @file{editline}. @file{Editline} was posted to the USENET newsgroup @file{comp.sources.misc} in 1992. A number of modifications have been made to make it more useful to us but the original code (contained within the standard speech tools distribution) and our modifications fall under the following licence. @example Copyright 1992 Simmule Turner and Rich Salz. All rights reserved. This software is not subject to any license of the American Telephone and Telegraph Company or of the Regents of the University of California. Permission is granted to anyone to use this software for any purpose on any computer system, and to alter it and redistribute it freely, subject to the following restrictions: 1. The authors are not responsible for the consequences of use of this software, no matter how awful, even if they arise from flaws in it. 2. The origin of this software must not be misrepresented, either by explicit claim or by omission. Since few users ever read sources, credits must appear in the documentation. 3. Altered versions must be plainly marked as such, and must not be misrepresented as being the original software. Since few users ever read sources, credits must appear in the documentation. 4. This notice may not be removed or altered. @end example @section Edinburgh Speech Tools Library @cindex Edinburgh Speech Tools Library The Edinburgh Speech Tools lies at the core of Festival. Although developed separately, much of the development of certain parts of the Edinburgh Speech Tools has been directed by Festival's needs. In turn those who have contributed to the Speech Tools make Festival a more usable system. @xref{Acknowledgements, , Acknowledgements, speechtools, Edinburgh Speech Tools Library Manual}. Online information about the Edinburgh Speech Tools library is available through @example @url{http://www.cstr.ed.ac.uk/projects/speech_tools.html} @end example @section Others Many others have provided actual code and support for Festival, for which we are grateful. Specifically, @itemize @bullet @item Alistair Conkie: various low level code points and some design work, Spanish synthesis, the old diphone synthesis code. @item Steve Isard: directorship and LPC diphone code, design of diphone schema. @item EPSRC: who fund Alan Black and Paul Taylor. @item Sun Microsystems Laboratories: for supporting the project and funding Richard. @item AT&T Labs - Research: for supporting the project. @item Paradigm Associates and George Carrett: for Scheme in one defun. @item Mike Macon: Improving the quality of the diphone synthesizer and LPC analysis. @item Kurt Dusterhoff: Tilt intonation training and modelling. @item Amy Isard: for her SSML project and related synthesizer. @item Richard Tobin: for answering all those difficult questions, the socket code, and the XML parser. @item Simmule Turner and Rich Salz: command line editor (editline) @item Borja Etxebarria: Help with the Spanish synthesis @item Briony Williams: Welsh synthesis @item Jacques H. de Villiers: @file{jacques@@cse.ogi.edu} from CSLU at OGI, for the TCL interface, and other usability issues @item Kevin Lenzo: @file{lenzo@@cs.cmu.edu} from CMU for the PERL interface. @item Rob Clarke: for support under Linux. @item Samuel Audet @file{guardia@@cam.org}: OS/2 support @item Mari Ostendorf: For providing access to the BU FM Radio corpus from which some modules were trained. @item Melvin Hunt: from whose work we based our residual LPC synthesis model on @item Oxford Text Archive: For the computer users version of Oxford Advanced Learners' Dictionary (redistributed with permission). @item Reading University: for access to MARSEC from which the phrase break model was trained. @item LDC & Penn Tree Bank: from which the POS tagger was trained, redistribution of the models is with permission from the LDC. @item Roger Burroughes and Kurt Dusterhoff: For letting us capture their voices. @item ATR and Nick Campbell: for first getting Paul and Alan to work together and for the experience we gained. @item FSF: for G++, make, .... @item Center for Spoken Language Understanding: CSLU at OGI, particularly Ron Cole and Mike Macon, have acted as significant users for the system giving significant feedback and allowing us to teach courses on Festival offering valuable real-use feedback. @item Our beta testers: Thanks to all the people who put up with previous versions of the system and reported bugs, both big and small. These comments are very important to the constant improvements in the system. And thanks for your quick responses when I had specific requests. @item And our users ... Many people have downloaded earlier versions of the system. Many have found problems with installation and use and have reported it to us. Many of you have put up with multiple compilations trying to fix bugs remotely. We thank you for putting up with us and are pleased you've taken the time to help us improve our system. Many of you have come up with uses we hadn't thought of, which is always rewarding. Even if you haven't actively responded, the fact that you use the system at all makes it worthwhile. @end itemize @node What is new, Overview, Acknowledgements , Top @chapter What is new Compared to the the previous major release (1.3.0 release Aug 1998) 1.4.0 is not functionally so different from its previous versions. This release is primarily a consolidation release fixing and tidying up some of the lower level aspects of the system to allow better modularity for some of our future planned modules. @itemize @bullet @item Copyright change: The system is now free and has no commercial restriction. Note that currently on the US voices (ked and kal) are also now unrestricted. The UK English voices depend on the Oxford Advanced Learners' Dictionary of Current English which cannot be used for commercial use without permission from Oxford University Press. @item Architecture tidy up: the interfaces to lower level part parts of the system have been tidied up deleting some of the older code that was supported for compatibility reasons. This is a much higher dependence of features and easier (and safer) ways to register new objects as feature values and Scheme objects. Scheme has been tidied up. It is no longer "in one defun" but "in one directory". @item New documentation system for speech tools: A new docbook based documentation system has been added to the speech tools. Festival's documentation will will move over to this sometime soon too. @item initial JSAPI support: both JSAPI and JSML (somewhat similar to Sable) now have initial impelementations. They of course depend on Java support which so far we have only (successfully) investgated under Solaris and Linux. @item Generalization of statistical models: CART, ngrams, and WFSTs are now fully supported from Lisp and can be used with a generalized viterbi function. This makes adding quite complex statistical models easy without adding new C++. @item Tilt Intonation modelling: Full support is now included for the Tilt intomation models, both training and use. @item Documentation on Bulding New Voices in Festival: documentation, scripts etc. for building new voices and languages in the system, see @example @url{http://www.cstr.ed.ac.uk/projects/festival/docs/festvox/} @end example @end itemize @node Overview, Installation , What is new, Top @chapter Overview Festival is designed as a speech synthesis system for at least three levels of user. First, those who simply want high quality speech from arbitrary text with the minimum of effort. Second, those who are developing language systems and wish to include synthesis output. In this case, a certain amount of customization is desired, such as different voices, specific phrasing, dialog types etc. The third level is in developing and testing new synthesis methods. This manual is not designed as a tutorial on converting text to speech but for documenting the processes and use of our system. We do not discuss the detailed algorithms involved in converting text to speech or the relative merits of multiple methods, though we will often give references to relevant papers when describing the use of each module. For more general information about text to speech we recommend Dutoit's @file{An introduction to Text-to-Speech Synthesis} @cite{dutoit97}. For more detailed research issues in TTS see @cite{sproat98} or @cite{vansanten96}. @menu * Philosophy:: Why we did it like it is * Future:: How much better its going to get @end menu @node Philosophy, Future, , Overview @section Philosophy One of the biggest problems in the development of speech synthesis, and other areas of speech and language processing systems, is that there are a lot of simple well-known techniques lying around which can help you realise your goal. But in order to improve some part of the whole system it is necessary to have a whole system in which you can test and improve your part. Festival is intended as that whole system in which you may simply work on your small part to improve the whole. Without a system like Festival, before you could even start to test your new module you would need to spend significant effort to build a whole system, or adapt an existing one before you could start working on your improvements. Festival is specifically designed to allow the addition of new modules, easily and efficiently, so that development need not get bogged down in re-implementing the wheel. But there is another aspect of Festival which makes it more useful than simply an environment for researching into new synthesis techniques. It is a fully usable text-to-speech system suitable for embedding in other projects that require speech output. The provision of a fully working easy-to-use speech synthesizer in addition to just a testing environment is good for two specific reasons. First, it offers a conduit for our research, in that our experiments can quickly and directly benefit users of our synthesis system. And secondly, in ensuring we have a fully working usable system we can immediately see what problems exist and where our research should be directed rather where our whims take us. These concepts are not unique to Festival. ATR's CHATR system (@cite{black94}) follows very much the same philosophy and Festival benefits from the experiences gained in the development of that system. Festival benefits from various pieces of previous work. As well as CHATR, CSTR's previous synthesizers, Osprey and the Polyglot projects influenced many design decisions. Also we are influenced by more general programs in considering software engineering issues, especially GNU Octave and Emacs on which the basic script model was based. Unlike in some other speech and language systems, software engineering is considered very important to the development of Festival. Too often research systems consist of random collections of hacky little scripts and code. No one person can confidently describe the algorithms it performs, as parameters are scattered throughout the system, with tricks and hacks making it impossible to really evaluate why the system is good (or bad). Such systems do not help the advancement of speech technology, except perhaps in pointing at ideas that should be further investigated. If the algorithms and techniques cannot be described externally from the program @emph{such that} they can reimplemented by others, what is the point of doing the work? Festival offers a common framework where multiple techniques may be implemented (by the same or different researchers) so that they may be tested more fairly in the same environment. As a final word, we'd like to make two short statements which both achieve the same end but unfortunately perhaps not for the same reasons: @quotation Good software engineering makes good research easier @end quotation But the following seems to be true also @quotation If you spend enough effort on something it can be shown to be better than its competitors. @end quotation @node Future, , Philosophy , Overview @section Future Festival is still very much in development. Hopefully this state will continue for a long time. It is never possible to complete software, there are always new things that can make it better. However as time goes on Festival's core architecture will stabilise and little or no changes will be made. Other aspects of the system will gain greater attention such as waveform synthesis modules, intonation techniques, text type dependent analysers etc. Festival will improve, so don't expected it to be the same six months from now. A number of new modules and enhancements are already under consideration at various stages of implementation. The following is a non-exhaustive list of what we may (or may not) add to Festival over the next six months or so. @itemize @bullet @item Selection-based synthesis: Moving away from diphone technology to more generalized selection of units for speech database. @item New structure for linguistic content of utterances: Using techniques for Metrical Phonology we are building more structure representations of utterances reflecting there linguistic significance better. This will allow improvements in prosody and unit selection. @item Non-prosodic prosodic control: For language generation systems and custom tasks where the speech to be synthesized is being generated by some program, more information about text structure will probably exist, such as phrasing, contrast, key items etc. We are investigating the relationship of high-level tags to prosodic information through the Sole project @url{http://www.cstr.ed.ac.uk/projects/sole.html} @item Dialect independent lexicons: Currently for each new dialect we need a new lexicon, we are currently investigating a form of lexical specification that is dialect independent that allows the core form to be mapped to different dialects. This will make the generation of voices in different dialects much easier. @end itemize @node Installation, Quick start, Overview, Top @chapter Installation This section describes how to install Festival from source in a new location and customize that installation. @menu * Requirements:: Software/Hardware requirements for Festival * Configuration:: Setting up compilation * Site initialization:: Settings for your particular site * Checking an installation:: But does it work ... @end menu @node Requirements, Configuration, , Installation @section Requirements @cindex requirements In order to compile Festival you first need the following source packages @table @code @item festival-1.4.3-release.tar.gz Festival Speech Synthesis System source @item speech_tools-1.2.3-release.tar.gz The Edinburgh Speech Tools Library @item festlex_NAME.tar.gz @cindex lexicon The lexicon distribution, where possible, includes the lexicon input file as well as the compiled form, for your convenience. The lexicons have varying distribution policies, but are all free except OALD, which is only free for non-commercial use (we are working on a free replacement). In some cases only a pointer to an ftp'able file plus a program to convert that file to the Festival format is included. @item festvox_NAME.tar.gz You'll need a speech database. A number are available (with varying distribution policies). Each voice may have other dependencies such as requiring particular lexicons @item festdoc_1.4.3.tar.gz Full postscript, info and html documentation for Festival and the Speech Tools. The source of the documentation is available in the standard distributions but for your conveniences it has been pre-generated. @end table In addition to Festival specific sources you will also need @table @emph @item A UNIX machine Currently we have compiled and tested the system under Solaris (2.5(.1), 2.6, 2.7 and 2.8), SunOS (4.1.3), FreeBSD (3.x, 4.x), Linux (Redhat 4.1, 5.0, 5.1, 5.2, 6.[012], 7.[01], 8.0 and other Linux distributions), and it should work under OSF (Dec Alphas), SGI (Irix), HPs (HPUX). But any standard UNIX machine should be acceptable. We have now successfully ported this version to Windows NT and Windows 95 (using the Cygnus GNU win32 environment). This is still a young port but seems to work. @item A C++ compiler @cindex GNU g++ @cindex g++ @cindex C++ Note that C++ is not very portable even between different versions of the compiler from the same vendor. Although we've tried very hard to make the system portable, we know it is very unlikely to compile without change except with compilers that have already been tested. The currently tested systems are @itemize @bullet @item Sun Sparc Solaris 2.5, 2.5.1, 2.6, 2.7, 2.9: GCC 2.95.1, GCC 3.2 @item FreeBSD for Intel 3.x and 4.x: GCC 2.95.1, GCC 3.0 @item Linux for Intel (RedHat 4.1/5.0/5.1/5.2/6.0/7.x/8.0): GCC 2.7.2, GCC 2.7.2/egcs-1.0.2, egcs 1.1.1, egcs-1.1.2, GCC 2.95.[123], GCC "2.96", GCC 3.0, GCC 3.0.1 GCC 3.2 GCC 3.2.1 @item Windows NT 4.0: GCC 2.7.2 plus egcs (from Cygnus GNU win32 b19), Visual C++ PRO v5.0, Visual C++ v6.0 @end itemize Note if GCC works on one version of Unix it usually works on others. @cindex Windows NT/95 We have compiled both the speech tools and Festival under Windows NT 4.0 and Windows 95 using the GNU tools available from Cygnus. @example @url{ftp://ftp.cygnus.com/pub/gnu-win32/}. @end example @item GNU make Due to there being too many different @code{make} programs out there we have tested the system using GNU make on all systems we use. Others may work but we know GNU make does. @item Audio hardware @cindex audio hardware You can use Festival without audio output hardware but it doesn't sound very good (though admittedly you can hear less problems with it). A number of audio systems are supported (directly inherited from the audio support in the Edinburgh Speech Tools Library): NCD's NAS (formerly called netaudio) a network transparent audio system (which can be found at @url{ftp://ftp.x.org/contrib/audio/nas/}); @file{/dev/audio} (at 8k ulaw and 8/16bit linear), found on Suns, Linux machines and FreeBSD; and a method allowing arbitrary UNIX commands. @xref{Audio output}. @end table @cindex readline @cindex editline @cindex GNU readline Earlier versions of Festival mistakenly offered a command line editor interface to the GNU package readline, but due to conflicts with the GNU Public Licence and Festival's licence this interface was removed in version 1.3.1. Even Festival's new free licence would cause problems as readline support would restrict Festival linking with non-free code. A new command line interface based on editline was provided that offers similar functionality. Editline remains a compilation option as it is probably not yet as portable as we would like it to be. @cindex @file{texi2html} In addition to the above, in order to process the documentation you will need @file{TeX}, @file{dvips} (or similar), GNU's @file{makeinfo} (part of the texinfo package) and @file{texi2html} which is available from @url{http://wwwcn.cern.ch/dci/texi2html/}. @cindex documentation However the document files are also available pre-processed into, postscript, DVI, info and html as part of the distribution in @file{festdoc-1.4.X.tar.gz}. Ensure you have a fully installed and working version of your C++ compiler. Most of the problems people have had in installing Festival have been due to incomplete or bad compiler installation. It might be worth checking if the following program works if you don't know if anyone has used your C++ installation before. @example #include int main (int argc, char **argv) @{ cout << "Hello world\n"; @} @end example Unpack all the source files in a new directory. The directory will then contain two subdirectories @example speech_tools/ festival/ @end example @node Configuration, Site initialization, Requirements , Installation @section Configuration First ensure you have a compiled version of the Edinburgh Speech Tools Library. See @file{speech_tools/INSTALL} for instructions. @cindex configuration The system now supports the standard GNU @file{configure} method for set up. In most cases this will automatically configure festival for your particular system. In most cases you need only type @example gmake @end example and the system will configure itself and compile, (note you need to have compiled the Edinburgh Speech Tools @file{speech_tools-1.2.2} first. @cindex @file{config/config} In some case hand configuration is required. All of the configuration choices are kept in the file @file{config/config}. @cindex OTHER_DIRS For the most part Festival configuration inherits the configuration from your speech tools config file (@file{../speech_tools/config/config}). Additional optional modules may be added by adding them to the end of your config file e.g. @example ALSO_INCLUDE += clunits @end example Adding and new module here will treat is as a new directory in the @file{src/modules/} and compile it into the system in the same way the @code{OTHER_DIRS} feature was used in previous versions. @cindex NFS @cindex automounter If the compilation directory being accessed by NFS or if you use an automounter (e.g. amd) it is recommend to explicitly set the variable @code{FESTIVAL_HOME} in @file{config/config}. The command @code{pwd} is not reliable when a directory may have multiple names. There is a simple test suite with Festival but it requires the three basic voices and their respective lexicons installed before it will work. Thus you need to install @example festlex_CMU.tar.gz festlex_OALD.tar.gz festlex_POSLEX.tar.gz festvox_don.tar.gz festvox_kedlpc16k.tar.gz festvox_rablpc16k.tar.gz @end example If these are installed you can test the installation with @example gmake test @end example To simply make it run with a male US English voice it is sufficient to install just @example festlex_CMU.tar.gz festlex_POSLEX.tar.gz festvox_kallpc16k.tar.gz @end example Note that the single most common reason for problems in compilation and linking found amongst the beta testers was a bad installation of GNU C++. If you get many strange errors in G++ library header files or link errors it is worth checking that your system has the compiler, header files and runtime libraries properly installed. This may be checked by compiling a simple program under C++ and also finding out if anyone at your site has ever used the installation. Most of these installation problems are caused by upgrading to a newer version of libg++ without removing the older version so a mixed version of the @file{.h} files exist. Although we have tried very hard to ensure that Festival compiles with no warnings this is not possible under some systems. @cindex SunOS Under SunOS the system include files do not declare a number of system provided functions. This a bug in Sun's include files. This will causes warnings like "implicit definition of fprintf". These are harmless. @cindex Linux Under Linux a warning at link time about reducing the size of some symbols often is produced. This is harmless. There is often occasional warnings about some socket system function having an incorrect argument type, this is also harmless. @cindex Visual C++ The speech tools and festival compile under Windows95 or Windows NT with Visual C++ v5.0 using the Microsoft @file{nmake} make program. We've only done this with the Professonal edition, but have no reason to believe that it relies on anything not in the standard edition. In accordance to VC++ conventions, object files are created with extension .obj, executables with extension .exe and libraries with extension .lib. This may mean that both unix and Win32 versions can be built in the same directory tree, but I wouldn't rely on it. To do this you require nmake Makefiles for the system. These can be generated from the gnumake Makefiles, using the command @example gnumake VCMakefile @end example in the speech_tools and festival directories. I have only done this under unix, it's possible it would work under the cygnus gnuwin32 system. If @file{make.depend} files exist (i.e. if you have done @file{gnumake depend} in unix) equivalent @file{vc_make.depend} files will be created, if not the VCMakefiles will not contain dependency information for the @file{.cc} files. The result will be that you can compile the system once, but changes will not cause the correct things to be rebuilt. In order to compile from the DOS command line using Visual C++ you need to have a collection of environment variables set. In Windows NT there is an instalation option for Visual C++ which sets these globally. Under Windows95 or if you don't ask for them to be set globally under NT you need to run @example vcvars32.bat @end example See the VC++ documentation for more details. Once you have the source trees with VCMakefiles somewhere visible from Windows, you need to copy @file{peech_tools\config\vc_config-dist} to @file{speech_tools\config\vc_config} and edit it to suit your local situation. Then do the same with @file{festival\config\vc_config-dist}. The thing most likely to need changing is the definition of @code{FESTIVAL_HOME} in @file{festival\config\vc_config_make_rules} which needs to point to where you have put festival. Now you can compile. cd to the speech_tools directory and do @example nmake /nologo /fVCMakefile @end example @exdent and the library, the programs in main and the test programs should be compiled. The tests can't be run automatically under Windows. A simple test to check that things are probably OK is: @example main\na_play testsuite\data\ch_wave.wav @end example @exdent which reads and plays a waveform. Next go into the festival directory and do @example nmake /nologo /fVCMakefile @end example @exdent to build festival. When it's finished, and assuming you have the voices and lexicons unpacked in the right place, festival should run just as under unix. We should remind you that the NT/95 ports are still young and there may yet be problems that we've not found yet. We only recommend the use the speech tools and Festival under Windows if you have significant experience in C++ under those platforms. @cindex smaller system @cindex minimal system Most of the modules @file{src/modules} are actually optional and the system could be compiled without them. The basic set could be reduced further if certain facilities are not desired. Particularly: @file{donovan} which is only required if the donovan voice is used; @file{rxp} if no XML parsing is required (e.g. Sable); and @file{parser} if no stochastic paring is required (this parser isn't used for any of our currently released voices). Actually even @file{UniSyn} and @file{UniSyn_diphone} could be removed if some external waveform synthesizer is being used (e.g. MBROLA) or some alternative one like @file{OGIresLPC}. Removing unused modules will make the festival binary smaller and (potentially) start up faster but don't expect too much. You can delete these by changing the @code{BASE_DIRS} variable in @file{src/modules/Makefile}. @node Site initialization, Checking an installation, Configuration, Installation @section Site initialization @cindex run-time configuration @cindex initialization @cindex installation initialization @cindex @file{init.scm} @cindex @file{siteinit.scm} Once compiled Festival may be further customized for particular sites. At start up time Festival loads the file @file{init.scm} from its library directory. This file further loads other necessary files such as phoneset descriptions, duration parameters, intonation parameters, definitions of voices etc. It will also load the files @file{sitevars.scm} and @file{siteinit.scm} if they exist. @file{sitevars.scm} is loaded after the basic Scheme library functions are loaded but before any of the festival related functions are loaded. This file is intended to set various path names before various subsystems are loaded. Typically variables such as @code{lexdir} (the directory where the lexicons are held), and @code{voices_dir} (pointing to voice directories) should be reset here if necessary. @cindex change libdir at run-time @cindex run-time configuration @cindex @code{load-path} The default installation will try to find its lexicons and voices automatically based on the value of @code{load-path} (this is derived from @code{FESTIVAL_HOME} at compilation time or by using the @code{--libdir} at run-time). If the voices and lexicons have been unpacked into subdirectories of the library directory (the default) then no site specific initialization of the above pathnames will be necessary. The second site specific file is @file{siteinit.scm}. Typical examples of local initialization are as follows. The default audio output method is NCD's NAS system if that is supported as that's what we use normally in CSTR. If it is not supported, any hardware specific mode is the default (e.g. sun16audio, freebas16audio, linux16audio or mplayeraudio). But that default is just a setting in @file{init.scm}. If for example in your environment you may wish the default audio output method to be 8k mulaw through @file{/dev/audio} you should add the following line to your @file{siteinit.scm} file @lisp (Parameter.set 'Audio_Method 'sunaudio) @end lisp Note the use of @code{Parameter.set} rather than @code{Parameter.def} the second function will not reset the value if it is already set. Remember that you may use the audio methods @code{sun16audio}. @code{linux16audio} or @code{freebsd16audio} only if @code{NATIVE_AUDIO} was selected in @file{speech_tools/config/config} and your are on such machines. The Festival variable @code{*modules*} contains a list of all supported functions/modules in a particular installation including audio support. Check the value of that variable if things aren't what you expect. If you are installing on a machine whose audio is not directly supported by the speech tools library, an external command may be executed to play a waveform. The following example is for an imaginary machine that can play audio files through a program called @file{adplay} with arguments for sample rate and file type. When playing waveforms, Festival, by default, outputs as unheadered waveform in native byte order. In this example you would set up the default audio playing mechanism in @file{siteinit.scm} as follows @lisp (Parameter.set 'Audio_Method 'Audio_Command) (Parameter.set 'Audio_Command "adplay -raw -r $SR $FILE") @end lisp @cindex output sample rate @cindex output file type @cindex audio command output @cindex audio output rate @cindex audio output filetype For @code{Audio_Command} method of playing waveforms Festival supports two additional audio parameters. @code{Audio_Required_Rate} allows you to use Festivals internal sample rate conversion function to any desired rate. Note this may not be as good as playing the waveform at the sample rate it is originally created in, but as some hardware devices are restrictive in what sample rates they support, or have naive resample functions this could be optimal. The second addition audio parameter is @code{Audio_Required_Format} which can be used to specify the desired output forms of the file. The default is unheadered raw, but this may be any of the values supported by the speech tools (including nist, esps, snd, riff, aiff, audlab, raw and, if you really want it, ascii). For example suppose you run Festival on a remote machine and are not running any network audio system and want Festival to copy files back to your local machine and simply cat them to @file{/dev/audio}. The following would do that (assuming permissions for rsh are allowed). @lisp (Parameter.set 'Audio_Method 'Audio_Command) ;; Make output file ulaw 8k (format ulaw implies 8k) (Parameter.set 'Audio_Required_Format 'ulaw) (Parameter.set 'Audio_Command "userhost=`echo $DISPLAY | sed 's/:.*$//'`; rcp $FILE $userhost:$FILE; \ rsh $userhost \"cat $FILE >/dev/audio\" ; rsh $userhost \"rm $FILE\"") @end lisp Note there are limits on how complex a command you want to put in the @code{Audio_Command} string directly. It can get very confusing with respect to quoting. It is therefore recommended that once you get past a certain complexity consider writing a simple shell script and calling it from the @code{Audio_Command} string. @cindex default voice A second typical customization is setting the default speaker. Speakers depend on many things but due to various licence (and resource) restrictions you may only have some diphone/nphone databases available in your installation. The function name that is the value of @code{voice_default} is called immediately after @file{siteinit.scm} is loaded offering the opportunity for you to change it. In the standard distribution no change should be required. If you download all the distributed voices @code{voice_rab_diphone} is the default voice. You may change this for a site by adding the following to @file{siteinit.scm} or per person by changing your @file{.festivalrc}. For example if you wish to change the default voice to the American one @code{voice_ked_diphone} @lisp (set! voice_default 'voice_ked_diphone) @end lisp Note the single quote, and note that unlike in early versions @code{voice_default} is not a function you can call directly. @cindex @file{.festivalrc} @cindex user initialization A second level of customization is on a per user basis. After loading @file{init.scm}, which includes @file{sitevars.scm} and @file{siteinit.scm} for local installation, Festival loads the file @file{.festivalrc} from the user's home directory (if it exists). This file may contain arbitrary Festival commands. @node Checking an installation, , Site initialization, Installation @section Checking an installation Once compiled and site initialization is set up you should test to see if Festival can speak or not. Start the system @example $ bin/festival Festival Speech Synthesis System 1.4.3:release Jan 2003 Copyright (C) University of Edinburgh, 1996-2003. All rights reserved. For details type `(festival_warranty)' festival> ^D @end example If errors occur at this stage they are most likely to do with pathname problems. If any error messages are printed about non-existent files check that those pathnames point to where you intended them to be. Most of the (default) pathnames are dependent on the basic library path. Ensure that is correct. To find out what it has been set to, start the system without loading the init files. @example $ bin/festival -q Festival Speech Synthesis System 1.4.3:release Jan 2003 Copyright (C) University of Edinburgh, 1996-2003. All rights reserved. For details type `(festival_warranty)' festival> libdir "/projects/festival/lib/" festival> ^D @end example This should show the pathname you set in your @file{config/config}. If the system starts with no errors try to synthesize something @example festival> (SayText "hello world") @end example Some files are only accessed at synthesis time so this may show up other problem pathnames. If it talks, you're in business, if it doesn't, here are some possible problems. @cindex audio problems If you get the error message @example Can't access NAS server @end example You have selected NAS as the audio output but have no server running on that machine or your @code{DISPLAY} or @code{AUDIOSERVER} environment variable is not set properly for your output device. Either set these properly or change the audio output device in @file{lib/siteinit.scm} as described above. Ensure your audio device actually works the way you think it does. On Suns, the audio output device can be switched into a number of different output modes, speaker, jack, headphones. If this is set to the wrong one you may not hear the output. Use one of Sun's tools to change this (try @file{/usr/demo/SOUND/bin/soundtool}). Try to find an audio file independent of Festival and get it to play on your audio. Once you have done that ensure that the audio output method set in Festival matches that. Once you have got it talking, test the audio spooling device. @example festival> (intro) @end example This plays a short introduction of two sentences, spooling the audio output. Finally exit from Festival (by end of file or @code{(quit)}) and test the script mode with. @example $ examples/saytime @end example A test suite is included with Festival but it makes certain assumptions about which voices are installed. It assumes that @code{voice_rab_diphone} (@file{festvox_rabxxxx.tar.gz}) is the default voice and that @code{voice_ked_diphone} and @code{voice_don_diphone} (@file{festvox_kedxxxx.tar.gz} and @file{festvox_don.tar.gz}) are installed. Also local settings in your @file{festival/lib/siteinit.scm} may affect these tests. However, after installation it may be worth trying @example gnumake test @end example from the @file{festival/} directory. This will do various tests including basic utterance tests and tokenization tests. It also checks that voices are installed and that they don't interfere with each other. These tests are primarily regression tests for the developers of Festival, to ensure new enhancements don't mess up existing supported features. They are not designed to test an installation is successful, though if they run correctly it is most probable the installation has worked. @node Quick start, Scheme, Installation, Top @chapter Quick start This section is for those who just want to know the absolute basics to run the system. @cindex command mode @cindex text-to-speech mode @cindex tts mode Festival works in two fundamental modes, @emph{command mode} and @emph{text-to-speech mode} (tts-mode). In command mode, information (in files or through standard input) is treated as commands and is interpreted by a Scheme interpreter. In tts-mode, information (in files or through standard input) is treated as text to be rendered as speech. The default mode is command mode, though this may change in later versions. @menu * Basic command line options:: * Simple command driven session:: * Getting some help:: @end menu @node Basic command line options, Simple command driven session, , Quick start @section Basic command line options @cindex command line options Festival's basic calling method is as @lisp festival [options] file1 file2 ... @end lisp Options may be any of the following @table @code @item -q start Festival without loading @file{init.scm} or user's @file{.festivalrc} @item -b @itemx --batch @cindex batch mode After processing any file arguments do not become interactive @item -i @itemx --interactive @cindex interactive mode After processing file arguments become interactive. This option overrides any batch argument. @item --tts @cindex tts mode Treat file arguments in text-to-speech mode, causing them to be rendered as speech rather than interpreted as commands. When selected in interactive mode the command line edit functions are not available @item --command @cindex command mode Treat file arguments in command mode. This is the default. @item --language LANG @cindex language specification Set the default language to @var{LANG}. Currently @var{LANG} may be one of @code{english}, @code{spanish} or @code{welsh} (depending on what voices are actually available in your installation). @item --server After loading any specified files go into server mode. This is a mode where Festival waits for clients on a known port (the value of @code{server_port}, default is 1314). Connected clients may send commands (or text) to the server and expect waveforms back. @xref{Server/client API}. Note server mode may be unsafe and allow unauthorised access to your machine, be sure to read the security recommendations in @ref{Server/client API} @item --script scriptfile @cindex script files @cindex Festival script files Run scriptfile as a Festival script file. This is similar to to @code{--batch} but it encapsulates the command line arguments into the Scheme variables @code{argv} and @code{argc}, so that Festival scripts may process their command line arguments just like any other program. It also does not load the the basic initialisation files as sometimes you may not want to do this. If you wish them, you should copy the loading sequence from an example Festival script like @file{festival/examples/saytext}. @item --heap NUMBER @cindex heap size @cindex Scheme heap size The Scheme heap (basic number of Lisp cells) is of a fixed size and cannot be dynamically increased at run time (this would complicate garbage collection). The default size is 210000 which seems to be more than adequate for most work. In some of our training experiments where very large list structures are required it is necessary to increase this. Note there is a trade off between size of the heap and time it takes to garbage collect so making this unnecessarily big is not a good idea. If you don't understand the above explanation you almost certainly don't need to use the option. @end table In command mode, if the file name starts with a left parenthesis, the name itself is read and evaluated as a Lisp command. This is often convenient when running in batch mode and a simple command is necessary to start the whole thing off after loading in some other specific files. @node Simple command driven session, Getting some help, Basic command line options, Quick start @section Sample command driven session Here is a short session using Festival's command interpreter. Start Festival with no arguments @lisp $ festival Festival Speech Synthesis System 1.4.3:release Dec 2002 Copyright (C) University of Edinburgh, 1996-2002. All rights reserved. For details type `(festival_warranty)' festival> @end lisp Festival uses the a command line editor based on editline for terminal input so command line editing may be done with Emacs commands. Festival also supports history as well as function, variable name, and file name completion via the @key{TAB} key. Typing @code{help} will give you more information, that is @code{help} without any parenthesis. (It is actually a variable name whose value is a string containing help.) @cindex Scheme @cindex read-eval-print loop Festival offers what is called a read-eval-print loop, because it reads an s-expression (atom or list), evaluates it and prints the result. As Festival includes the SIOD Scheme interpreter most standard Scheme commands work @lisp festival> (car '(a d)) a festival> (+ 34 52) 86 @end lisp In addition to standard Scheme commands a number of commands specific to speech synthesis are included. Although, as we will see, there are simpler methods for getting Festival to speak, here are the basic underlying explicit functions used in synthesizing an utterance. @cindex utterance @cindex hello world Utterances can consist of various types @xref{Utterance types}, but the simplest form is plain text. We can create an utterance and save it in a variable @lisp festival> (set! utt1 (Utterance Text "Hello world")) # festival> @end lisp The (hex) number in the return value may be different for your installation. That is the print form for utterances. Their internal structure can be very large so only a token form is printed. @cindex synthesizing an utterance Although this creates an utterance it doesn't do anything else. To get a waveform you must synthesize it. @lisp festival> (utt.synth utt1) # festival> @end lisp @cindex playing an utterance This calls various modules, including tokenizing, duration,. intonation etc. Which modules are called are defined with respect to the type of the utterance, in this case @code{Text}. It is possible to individually call the modules by hand but you just wanted it to talk didn't you. So @lisp festival> (utt.play utt1) # festival> @end lisp @exdent will send the synthesized waveform to your audio device. You should hear "Hello world" from your machine. @cindex @code{SayText} To make this all easier a small function doing these three steps exists. @code{SayText} simply takes a string of text, synthesizes it and sends it to the audio device. @lisp festival> (SayText "Good morning, welcome to Festival") # festival> @end lisp Of course as history and command line editing are supported @key{c-p} or up-arrow will allow you to edit the above to whatever you wish. Festival may also synthesize from files rather than simply text. @lisp festival> (tts "myfile" nil) nil festival> @end lisp @cindex exiting Festival @cindex @code{quit} The end of file character @key{c-d} will exit from Festival and return you to the shell, alternatively the command @code{quit} may be called (don't forget the parentheses). @cindex TTS @cindex text to speech Rather than starting the command interpreter, Festival may synthesize files specified on the command line @lisp unix$ festival --tts myfile unix$ @end lisp @cindex text to wave @cindex offline TTS Sometimes a simple waveform is required from text that is to be kept and played at some later time. The simplest way to do this with festival is by using the @file{text2wave} program. This is a festival script that will take a file (or text from standard input) and produce a single waveform. @cindex text2wave An example use is @example text2wave myfile.txt -o myfile.wav @end example Options exist to specify the waveform file type, for example if Sun audio format is required @example text2wave myfile.txt -otype snd -o myfile.wav @end example Use @file{-h} on @file{text2wave} to see all options. @node Getting some help, , Simple command driven session, Quick start @section Getting some help @cindex help If no audio is generated then you must check to see if audio is properly initialized on your machine. @xref{Audio output}. In the command interpreter @key{m-h} (meta-h) will give you help on the current symbol before the cursor. This will be a short description of the function or variable, how to use it and what its arguments are. A listing of all such help strings appears at the end of this document. @key{m-s} will synthesize and say the same information, but this extra function is really just for show. @cindex @code{manual} The lisp function @code{manual} will send the appropriate command to an already running Netscape browser process. If @code{nil} is given as an argument the browser will be directed to the tables of contents of the manual. If a non-nil value is given it is assumed to be a section title and that section is searched and if found displayed. For example @example festival> (manual "Accessing an utterance") @end example Another related function is @code{manual-sym} which given a symbol will check its documentation string for a cross reference to a manual section and request Netscape to display it. This function is bound to @key{m-m} and will display the appropriate section for the given symbol. Note also that the @key{TAB} key can be used to find out the name of commands available as can the function @code{Help} (remember the parentheses). For more up to date information on Festival regularly check the Festival Home Page at @example @url{http://www.cstr.ed.ac.uk/projects/festival.html} @end example Further help is available by mailing questions to @example festival-help@@cstr.ed.ac.uk @end example Although we cannot guarantee the time required to answer you, we will do our best to offer help. @cindex bug reports Bug reports should be submitted to @example festival-bug@@cstr.ed.ac.uk @end example If there is enough user traffic a general mailing list will be created so all users may share comments and receive announcements. In the mean time watch the Festival Home Page for news. @node Scheme, TTS, Quick start, Top @chapter Scheme @cindex Scheme introduction Many people seem daunted by the fact that Festival uses Scheme as its scripting language and feel they can't use Festival because they don't know Scheme. However most of those same people use Emacs everyday which also has (a much more complex) Lisp system underneath. The number of Scheme commands you actually need to know in Festival is really very small and you can easily just find out as you go along. Also people use the Unix shell often but only know a small fraction of actual commands available in the shell (or in fact that there even is a distinction between shell builtin commands and user definable ones). So take it easy, you'll learn the commands you need fairly quickly. @menu * Scheme references:: Places to learn more about Scheme * Scheme fundamentals:: Syntax and semantics * Scheme Festival specifics:: * Scheme I/O:: @end menu @node Scheme references, Scheme fundamentals, , Scheme @section Scheme references If you wish to learn about Scheme in more detail I recommend the book @cite{abelson85}. The Emacs Lisp documentation is reasonable as it is comprehensive and many of the underlying uses of Scheme in Festival were influenced by Emacs. Emacs Lisp however is not Scheme so there are some differences. @cindex Scheme references Other Scheme tutorials and resources available on the Web are @itemize @bullet @item The Revised Revised Revised Revised Scheme Report, the document defining the language is available from @example @url{http://tinuviel.cs.wcu.edu/res/ldp/r4rs-html/r4rs_toc.html} @end example @item a Scheme tutorials from the net: @itemize @bullet @item @url{http://www.cs.uoregon.edu/classes/cis425/schemeTutorial.html} @end itemize @item the Scheme FAQ @itemize @bullet @item @url{http://www.landfield.com/faqs/scheme-faq/part1/} @end itemize @end itemize @node Scheme fundamentals, Scheme Festival specifics, Scheme references, Scheme @section Scheme fundamentals But you want more now, don't you, not just be referred to some other book. OK here goes. @emph{Syntax}: an expression is an @emph{atom} or a @emph{list}. A list consists of a left paren, a number of expressions and right paren. Atoms can be symbols, numbers, strings or other special types like functions, hash tables, arrays, etc. @emph{Semantics}: All expressions can be evaluated. Lists are evaluated as function calls. When evaluating a list all the members of the list are evaluated first then the first item (a function) is called with the remaining items in the list as arguments. Atoms are evaluated depending on their type: symbols are evaluated as variables returning their values. Numbers, strings, functions, etc. evaluate to themselves. Comments are started by a semicolon and run until end of line. And that's it. There is nothing more to the language that. But just in case you can't follow the consequences of that, here are some key examples. @lisp festival> (+ 2 3) 5 festival> (set! a 4) 4 festival> (* 3 a) 12 festival> (define (add a b) (+ a b)) # festival> (add 3 4) 7 festival> (set! alist '(apples pears bananas)) (apples pears bananas) festival> (car alist) apples festival> (cdr alist) (pears bananas) festival> (set! blist (cons 'oranges alist)) (oranges apples pears bananas) festival> (append alist blist) (apples pears bananas oranges apples pears bananas) festival> (cons alist blist) ((apples pears bananas) oranges apples pears bananas) festival> (length alist) 3 festival> (length (append alist blist)) 7 @end lisp @node Scheme Festival specifics, Scheme I/O, Scheme fundamentals, Scheme @section Scheme Festival specifics There a number of additions to SIOD that are Festival specific though still part of the Lisp system rather than the synthesis functions per se. By convention if the first statement of a function is a string, it is treated as a documentation string. The string will be printed when help is requested for that function symbol. @cindex debugging Scheme errors @cindex debugging scripts @cindex backtrace In interactive mode if the function @code{:backtrace} is called (within parenthesis) the previous stack trace is displayed. Calling @code{:backtrace} with a numeric argument will display that particular stack frame in full. Note that any command other than @code{:backtrace} will reset the trace. You may optionally call @lisp (set_backtrace t) @end lisp Which will cause a backtrace to be displayed whenever a Scheme error occurs. This can be put in your @file{.festivalrc} if you wish. This is especially useful when running Festival in non-interactive mode (batch or script mode) so that more information is printed when an error occurs. @cindex hooks A @emph{hook} in Lisp terms is a position within some piece of code where a user may specify their own customization. The notion is used heavily in Emacs. In Festival there a number of places where hooks are used. A hook variable contains either a function or list of functions that are to be applied at some point in the processing. For example the @code{after_synth_hooks} are applied after synthesis has been applied to allow specific customization such as resampling or modification of the gain of the synthesized waveform. The Scheme function @code{apply_hooks} takes a hook variable as argument and an object and applies the function/list of functions in turn to the object. @cindex catching errors in Scheme @cindex @code{unwind-protect} @cindex errors in Scheme When an error occurs in either Scheme or within the C++ part of Festival by default the system jumps to the top level, resets itself and continues. Note that errors are usually serious things, pointing to bugs in parameters or code. Every effort has been made to ensure that the processing of text never causes errors in Festival. However when using Festival as a development system it is often that errors occur in code. Sometimes in writing Scheme code you know there is a potential for an error but you wish to ignore that and continue on to the next thing without exiting or stopping and returning to the top level. For example you are processing a number of utterances from a database and some files containing the descriptions have errors in them but you want your processing to continue through every utterance that can be processed rather than stopping 5 minutes after you gone home after setting a big batch job for overnight. @cindex @code{unwind-protect} @cindex catching errors Festival's Scheme provides the function @code{unwind-protect} which allows the catching of errors and then continuing normally. For example suppose you have the function @code{process_utt} which takes a filename and does things which you know might cause an error. You can write the following to ensure you continue processing even in an error occurs. @lisp (unwind-protect (process_utt filename) (begin (format t "Error found in processing %s\n" filename) (format t "continuing\n"))) @end lisp The @code{unwind-protect} function takes two arguments. The first is evaluated and if no error occurs the value returned from that expression is returned. If an error does occur while evaluating the first expression, the second expression is evaluated. @code{unwind-protect} may be used recursively. Note that all files opened while evaluating the first expression are closed if an error occurs. All global variables outside the scope of the @code{unwind-protect} will be left as they were set up until the error. Care should be taken in using this function but its power is necessary to be able to write robust Scheme code. @node Scheme I/O, , Scheme Festival specifics, Scheme @section Scheme I/O @cindex file i/o in Scheme @cindex i/o in Scheme Different Scheme's may have quite different implementations of file i/o functions so in this section we will describe the basic functions in Festival SIOD regarding i/o. Simple printing to the screen may be achieved with the function @code{print} which prints the given s-expression to the screen. The printed form is preceded by a new line. This is often useful for debugging but isn't really powerful enough for much else. @cindex @code{fopen} @cindex @code{fclose} Files may be opened and closed and referred to file descriptors in a direct analogy to C's stdio library. The SIOD functions @code{fopen} and @code{fclose} work in the exactly the same way as their equivalently named partners in C. @cindex @code{format} @cindex formatted output The @code{format} command follows the command of the same name in Emacs and a number of other Lisps. C programmers can think of it as @code{fprintf}. @code{format} takes a file descriptor, format string and arguments to print. The file description may be a file descriptor as returned by the Scheme function @code{fopen}, it may also be @code{t} which means the output will be directed as standard out (cf. @code{printf}). A third possibility is @code{nil} which will cause the output to printed to a string which is returned (cf. @code{sprintf}). The format string closely follows the format strings in ANSI C, but it is not the same. Specifically the directives currently supported are, @code{%%}, @code{%d}, @code{%x}, @code{%s}, @code{%f}, @code{%g} and @code{%c}. All modifiers for these are also supported. In addition @code{%l} is provided for printing of Scheme objects as objects. For example @lisp (format t "%03d %3.4f %s %l %l %l\n" 23 23 "abc" "abc" '(a b d) utt1) @end lisp will produce @lisp 023 23.0000 abc "abc" (a b d) # @end lisp on standard output. @cindex pretty printing When large lisp expressions are printed they are difficult to read because of the parentheses. The function @code{pprintf} prints an expression to a file description (or @code{t} for standard out). It prints so the s-expression is nicely lined up and indented. This is often called pretty printing in Lisps. @cindex reading from files @cindex loading data from files For reading input from terminal or file, there is currently no equivalent to @code{scanf}. Items may only be read as Scheme expressions. The command @lisp (load FILENAME t) @end lisp @exdent will load all s-expressions in @code{FILENAME} and return them, unevaluated as a list. Without the third argument the @code{load} function will load and evaluate each s-expression in the file. To read individual s-expressions use @code{readfp}. For example @lisp (let ((fd (fopen trainfile "r")) (entry) (count 0)) (while (not (equal? (set! entry (readfp fd)) (eof-val))) (if (string-equal (car entry) "home") (set! count (+ 1 count)))) (fclose fd)) @end lisp @cindex @code{parse-number} @cindex @code{atof} @cindex string to number @cindex convert string to number To convert a symbol whose print name is a number to a number use @code{parse-number}. This is the equivalent to @code{atof} in C. Note that, all i/o from Scheme input files is assumed to be basically some form of Scheme data (though can be just numbers, tokens). For more elaborate analysis of incoming data it is possible to use the text tokenization functions which offer a fully programmable method of reading data. @node TTS, XML/SGML mark-up, Scheme, Top @chapter TTS Festival supports text to speech for raw text files. If you are not interested in using Festival in any other way except as black box for rendering text as speech, the following method is probably what you want. @example festival --tts myfile @end example This will say the contents of @file{myfile}. Alternatively text may be submitted on standard input @example echo hello world | festival --tts cat myfile | festival --tts @end example @cindex text modes Festival supports the notion of @emph{text modes} where the text file type may be identified, allowing Festival to process the file in an appropriate way. Currently only two types are considered stable: @code{STML} and @code{raw}, but other types such as @code{email}, @code{HTML}, @code{Latex}, etc. are being developed and discussed below. This follows the idea of buffer modes in Emacs where a file's type can be utilized to best display the text. Text mode may also be selected based on a filename's extension. Within the command interpreter the function @code{tts} is used to render files as text; it takes a filename and the text mode as arguments. @menu * Utterance chunking:: From text to utterances * Text modes:: Mode specific text analysis * Example text mode:: An example mode for reading email @end menu @node Utterance chunking, Text modes, , TTS @section Utterance chunking @cindex utterance chunking @cindex @code{eou_tree} Text to speech works by first tokenizing the file and chunking the tokens into utterances. The definition of utterance breaks is determined by the utterance tree in variable @code{eou_tree}. A default version is given in @file{lib/tts.scm}. This uses a decision tree to determine what signifies an utterance break. Obviously blank lines are probably the most reliable, followed by certain punctuation. The confusion of the use of periods for both sentence breaks and abbreviations requires some more heuristics to best guess their different use. The following tree is currently used which works better than simply using punctuation. @lisp (defvar eou_tree '((n.whitespace matches ".*\n.*\n\\(.\\|\n\\)*") ;; 2 or more newlines ((1)) ((punc in ("?" ":" "!")) ((1)) ((punc is ".") ;; This is to distinguish abbreviations vs periods ;; These are heuristics ((name matches "\\(.*\\..*\\|[A-Z][A-Za-z]?[A-Za-z]?\\|etc\\)") ((n.whitespace is " ") ((0)) ;; if abbrev single space isn't enough for break ((n.name matches "[A-Z].*") ((1)) ((0)))) ((n.whitespace is " ") ;; if it doesn't look like an abbreviation ((n.name matches "[A-Z].*") ;; single space and non-cap is no break ((1)) ((0))) ((1)))) ((0))))) @end lisp The token items this is applied to will always (except in the end of file case) include one following token, so look ahead is possible. The "n." and "p." and "p.p." prefixes allow access to the surrounding token context. The features @code{name}, @code{whitespace} and @code{punc} allow access to the contents of the token itself. At present there is no way to access the lexicon form this tree which unfortunately might be useful if certain abbreviations were identified as such there. Note these are heuristics and written by hand not trained from data, though problems have been fixed as they have been observed in data. The above rules may make mistakes where abbreviations appear at end of lines, and when improper spacing and capitalization is used. This is probably worth changing, for modes where more casual text appears, such as email messages and USENET news messages. A possible improvement could be made by analysing a text to find out its basic threshold of utterance break (i.e. if no full stop, two spaces, followed by a capitalized word sequences appear and the text is of a reasonable length then look for other criteria for utterance breaks). Ultimately what we are trying to do is to chunk the text into utterances that can be synthesized quickly and start to play them quickly to minimise the time someone has to wait for the first sound when starting synthesis. Thus it would be better if this chunking were done on @emph{prosodic phrases} rather than chunks more similar to linguistic sentences. Prosodic phrases are bounded in size, while sentences are not. @node Text modes, Example text mode, Utterance chunking, TTS @section Text modes @cindex text modes We do not believe that all texts are of the same type. Often information about the general contents of file will aid synthesis greatly. For example in Latex files we do not want to here "left brace, backslash e m" before each emphasized word, nor do we want to necessarily hear formating commands. Festival offers a basic method for specifying customization rules depending on the @emph{mode} of the text. By type we are following the notion of modes in Emacs and eventually will allow customization at a similar level. Modes are specified as the third argument to the function @code{tts}. When using the Emacs interface to Festival the buffer mode is automatically passed as the text mode. If the mode is not supported a warning message is printed and the raw text mode is used. Our initial text mode implementation allows configuration both in C++ and in Scheme. Obviously in C++ almost anything can be done but it is not as easy to reconfigure without recompilation. Here we will discuss those modes which can be fully configured at run time. A text mode may contain the following @table @emph @item filter A Unix shell program filter that processes the text file in some appropriate way. For example for email it might remove uninteresting headers and just output the subject, from line and the message body. If not specified, an identity filter is used. @item init_function This (Scheme) function will be called before any processing will be done. It allows further set up of tokenization rules and voices etc. @item exit_function This (Scheme) function will be called at the end of any processing allowing reseting of tokenization rules etc. @item analysis_mode If analysis mode is @code{xml} the file is read through the built in XML parser @code{rxp}. Alternatively if analysis mode is @code{xxml} the filter should an SGML normalising parser and the output is processed in a way suitable for it. Any other value is ignored. @end table These mode specific parameters are specified in the a-list held in @code{tts_text_modes}. When using Festival in Emacs the emacs buffer mode is passed to Festival as the text mode. Note that above mechanism is not really designed to be re-entrant, this should be addressed in later versions. @cindex @code{auto-text-mode-alist} @cindex automatic selection of text mode Following the use of auto-selection of mode in Emacs, Festival can auto-select the text mode based on the filename given when no explicit mode is given. The Lisp variable @code{auto-text-mode-alist} is a list of dotted pairs of regular expression and mode name. For example to specify that the @code{email} mode is to be used for files ending in @file{.email} we would add to the current @code{auto-text-mode-alist} as follows @lisp (set! auto-text-mode-alist (cons (cons "\\.email$" 'email) auto-text-mode-alist)) @end lisp If the function @code{tts} is called with a mode other than @code{nil} that mode overrides any specified by the @code{auto-text-mode-alist}. The mode @code{fundamental} is the explicit "null" mode, it is used when no mode is specified in the function @code{tts}, and match is found in @code{auto-text-mode-alist} or the specified mode is not found. By convention if a requested text model is not found in @code{tts_text_modes} the file @file{MODENAME-mode} will be @code{required}. Therefore if you have the file @file{MODENAME-mode.scm} in your library then it will be automatically loaded on reference. Modes may be quite large and it is not necessary have Festival load them all at start up time. Because of the @code{auto-text-mode-alist} and the auto loading of currently undefined text modes you can use Festival like @example festival --tts example.email @end example Festival with automatically synthesize @file{example.email} in text mode @code{email}. @cindex personal text modes If you add your own personal text modes you should do the following. Suppose you've written an HTML mode. You have named it @file{html-mode.scm} and put it in @file{/home/awb/lib/festival/}. In your @file{.festivalrc} first identify you're personal Festival library directory by adding it to @code{lib-path}. @example (set! lib-path (cons "/home/awb/lib/festival/" lib-path)) @end example Then add the definition to the @code{auto-text-mode-alist} that file names ending @file{.html} or @file{.htm} should be read in HTML mode. @example (set! auto-text-mode-alist (cons (cons "\\.html?$" 'html) auto-text-mode-alist)) @end example Then you may synthesize an HTML file either from Scheme @example (tts "example.html" nil) @end example @exdent Or from the shell command line @example festival --tts example.html @end example Anyone familiar with modes in Emacs should recognise that the process of adding a new text mode to Festival is very similar to adding a new buffer mode to Emacs. @node Example text mode, , Text modes, TTS @section Example text mode @cindex email mode Here is a short example of a tts mode for reading email messages. It is by no means complete but is a start at showing how you can customize tts modes without writing new C++ code. The first task is to define a filter that will take a saved mail message and remove extraneous headers and just leave the from line, subject and body of the message. The filter program is given a file name as its first argument and should output the result on standard out. For our purposes we will do this as a shell script. @example #!/bin/sh # Email filter for Festival tts mode # usage: email_filter mail_message >tidied_mail_message grep "^From: " $1 echo grep "^Subject: " $1 echo # delete up to first blank line (i.e. the header) sed '1,/^$/ d' $1 @end example Next we define the email init function, which will be called when we start this mode. What we will do is save the current token to words function and slot in our own new one. We can then restore the previous one when we exit. @lisp (define (email_init_func) "Called on starting email text mode." (set! email_previous_t2w_func token_to_words) (set! english_token_to_words email_token_to_words) (set! token_to_words email_token_to_words)) @end lisp Note that @emph{both} @code{english_token_to_words} and @code{token_to_words} should be set to ensure that our new token to word function is still used when we change voices. The corresponding end function puts the token to words function back. @lisp (define (email_exit_func) "Called on exit email text mode." (set! english_token_to_words email_previous_t2w_func) (set! token_to_words email_previous_t2w_func)) @end lisp Now we can define the email specific token to words function. In this example we deal with two specific cases. First we deal with the common form of email addresses so that the angle brackets are not pronounced. The second points are to recognise quoted text and immediately change the the speaker to the alternative speaker. @lisp (define (email_token_to_words token name) "Email specific token to word rules." (cond @end lisp This first condition identifies the token as a bracketed email address and removes the brackets and splits the token into name and IP address. Note that we recursively call the function @code{email_previous_t2w_func} on the email name and IP address so that they will be pronounced properly. Note that because that function returns a @emph{list} of words we need to append them together. @lisp ((string-matches name "<.*@.*>") (append (email_previous_t2w_func token (string-after (string-before name "@@") "<")) (cons "at" (email_previous_t2w_func token (string-before (string-after name "@@") ">"))))) @end lisp Our next condition deals with identifying a greater than sign being used as a quote marker. When we detect this we select the alternative speaker, even though it may already be selected. We then return no words so the quote marker is not spoken. The following condition finds greater than signs which are the first token on a line. @lisp ((and (string-matches name ">") (string-matches (item.feat token "whitespace") "[ \t\n]*\n *")) (voice_don_diphone) nil ;; return nothing to say ) @end lisp If it doesn't match any of these we can go ahead and use the builtin token to words function Actually, we call the function that was set before we entered this mode to ensure any other specific rules still remain. But before that we need to check if we've had a newline with doesn't start with a greater than sign. In that case we switch back to the primary speaker. @lisp (t ;; for all other cases (if (string-matches (item.feat token "whitespace") ".*\n[ \t\n]*") (voice_rab_diphone)) (email_previous_t2w_func token name)))) @end lisp @cindex declaring text modes In addition to these we have to actually declare the text mode. This we do by adding to any existing modes as follows. @lisp (set! tts_text_modes (cons (list 'email ;; mode name (list ;; email mode params (list 'init_func email_init_func) (list 'exit_func email_exit_func) '(filter "email_filter"))) tts_text_modes)) @end lisp This will now allow simple email messages to be dealt with in a mode specific way. An example mail message is included in @file{examples/ex1.email}. To hear the result of the above text mode start Festival, load in the email mode descriptions, and call TTS on the example file. @example (tts ".../examples/ex1.email" 'email) @end example The above is very short of a real email mode but does illustrate how one might go about building one. It should be reiterated that text modes are new in Festival and their most effective form has not been discovered yet. This will improve with time and experience. @node XML/SGML mark-up, Emacs interface, TTS, Top @chapter XML/SGML mark-up @cindex STML @cindex SGML @cindex SSML @cindex Sable @cindex XML @cindex Spoken Text Mark-up Language The ideas of a general, synthesizer system nonspecific, mark-up language for labelling text has been under discussion for some time. Festival has supported an SGML based markup language through multiple versions most recently STML (@cite{sproat97}). This is based on the earlier SSML (Speech Synthesis Markup Language) which was supported by previous versions of Festival (@cite{taylor96}). With this version of Festival we support @emph{Sable} a similar mark-up language devised by a consortium from Bell Labls, Sub Microsystems, AT&T and Edinburgh, @cite{sable98}. Unlike the previous versions which were SGML based, the implementation of Sable in Festival is now XML based. To the user they different is negligable but using XML makes processing of files easier and more standardized. Also Festival now includes an XML parser thus reducing the dependencies in processing Sable text. Raw text has the problem that it cannot always easily be rendered as speech in the way the author wishes. Sable offers a well-defined way of marking up text so that the synthesizer may render it appropriately. @cindex CSS @cindex Cascading style sheets @cindex DSSSL The definition of Sable is by no means settled and is still in development. In this release Festival offers people working on Sable and other XML (and SGML) based markup languages a chance to quickly experiment with prototypes by providing a DTD (document type descriptions) and the mapping of the elements in the DTD to Festival functions. Although we have not yet (personally) investigated facilities like cascading style sheets and generalized SGML specification languages like DSSSL we believe the facilities offer by Festival allow rapid prototyping of speech output markup languages. Primarily we see Sable markup text as a language that will be generated by other programs, e.g. text generation systems, dialog managers etc. therefore a standard, easy to parse, format is required, even if it seems overly verbose for human writers. For more information of Sable and access to the mailing list see @example @url{http://www.cstr.ed.ac.uk/projects/sable.html} @end example @menu * Sable example:: an example of Sable with descriptions * Supported Sable tags:: Currently supported Sable tags * Adding Sable tags:: Adding new Sable tags * XML/SGML requirements:: Software environment requirements for use * Using Sable:: Rendering Sable files as speech @end menu @node Sable example, Supported Sable tags, , XML/SGML mark-up @section Sable example Here is a simple example of Sable marked up text @example The boy saw the girl in the park with the telescope. The boy saw the girl in the park with the telescope. Good morning My name is Stuart, which is spelled stuart though some people pronounce it stuart. My telephone number is 2787. I used to work in Buccleuch Place, but no one can pronounce that. By the way, my telephone number is actually @end example @cindex SABLE DTD @cindex @file{Sable.v0_2.dtd} After the initial definition of the SABLE tags, through the file @file{Sable.v0_2.dtd}, which is distributed as part of Festival, the body is given. There are tags for identifying the language and the voice. Explicit boundary markers may be given in text. Also duration and intonation control can be explicit specified as can new pronunciations of words. The last sentence specifies some external filenames to play at that point. @node Supported Sable tags, Adding Sable tags, Sable example, XML/SGML mark-up @section Supported Sable tags @cindex Sable tags There is not yet a definitive set of tags but hopefully such a list will form over the next few months. As adding support for new tags is often trivial the problem lies much more in defining what tags there should be than in actually implementing them. The following are based on version 0.2 of Sable as described in @url{http://www.cstr.ed.ac.uk/projects/sable_spec2.html}, though some aspects are not currently supported in this implementation. Further updates will be announces through the Sable mailing list. @table @code @item LANGUAGE Allows the specification of the language through the @code{ID} attribute. Valid values in Festival are, @code{english}, @code{en1}, @code{spanish}, @code{en}, and others depending on your particular installation. For example @example ... @end example If the language isn't supported by the particualr installation of Festival "Some text in .." is said instead and the section is ommitted. @item SPEAKER Select a voice. Accepts a parameter @code{NAME} which takes values @code{male1}, @code{male2}, @code{female1}, etc. There is currently no definition about what happens when a voice is selected which the synthesizer doesn't support. An example is @example ... @end example @item AUDIO This allows the specification of an external waveform that is to be included. There are attributes for specifying volume and whether the waveform is to be played in the background of the following text or not. Festival as yet only supports insertion. @example My telephone number is

    festival/doc/hierHeader.inc0000644000175000017500000000104007262605651014303 0ustar awbawb Festival Class Hierachy

    Festival Documentation
    V1.2.5 December 1997


    festival/doc/indexHeader.inc0000644000175000017500000000104607262605651014471 0ustar awbawb Festival Documentation Index

    Festival Documentation
    V1.2.5 December 1997


    festival/doc/banner.inc0000644000175000017500000000066307625474505013527 0ustar awbawb

    This page is part of the Festival Text to Speech System documentation
    Copyright University of Edinburgh 1997
    Contact: festival@cstr.ed.ac.uk


    festival/doc/edcrest.gif0000644000175000017500000000711307262605651013677 0ustar awbawbGIF89aqs¥U¯¯¯ª>U,LQQ§©§ixj&Ñq›¶v8ŽŒ‡{Z/’5éåÜÖ×Æ¡ÿ???)'&ÿ^ S9ÿ¨B97.UH/kklžžž”””5PkªÿÙQVQ>!SƒÿÿÿzQ Ò³v²‹P...žn2obS¿¿¿|ÆÆ fðè( í(@!ù/,qsþÀ—pH,ÈdÑ 0~@ ¥tJ­Z¯I……¬„%ŠK.›«ŠqQAAzÚ¯RóL¯Ÿ%$÷hÐ÷Esv†uoB zIQG!%‡™fŒ jE€FyG C¦š¬U _ŒF¢E np ­½/…E¹,pD³F˜zxŠF À¾e©H³ ÑvZBÝk¹Õw¼DÊL† Æä¡ «åSÊÊDø™¯Ô^¸Øƒ‹yØ^ø3Xåà ¯B0¬â]3,h,@¡ã>ç¨Pð  Ø»‰J@Jƒ†B0cÊœYLAÀ%%ü¼Aie¤þ‹H”`ó‚‰ @À”)„¤&L\ˆÉÂÅI#˜Ôâ™D€%B.âóh hÓª]›™û²l]$‰ëdE!œeË·¯… LPû⪗ñì*ù@6¿#'€`*Ä*Ú\V\/€‘C‹¶ð,a…ŒÚ½Ê9Œž y@i9G’½ØÖšHƒLȦMܯ[ê‘·@¯>§×À^¼zßSYÿb‚ß!Ü"‘[ï7ÁT(›1S¢O<ù÷|-[QŽÅŸ?2"ø²,ЇD_!€Û¸0K)Ò}6ÀRK5eá…f¨á†þ VP±d†<Èk(¦¨âŠ,¶èâ‹/^@Øoóœ¡ û áÕ¤Á@)äDiä‘A~v ÔxP †Ñ±Ùœ €ˆ å–\fV—_~å!¯E&<€@ioQ&˜p)§dSHD„!ÜDÇUœüwe"L`衈Ž@ Œ2Á ,`‚¤‘.€@¤—.@¤4Úh # *êØ)Û€è¡CÏ6& 5h¡£&JAžJB¥ 0€i¤º&µÀ¬ž¢à ±"Zê®vr—NL‚{¼Zl¢ÐÚè —f餕Šð«µŸ;­¡ÇR'˜wþ{²!…FRðyZÒŽk¨¢Á @€¥•BÚ«¥à2 ª¼äššÖ¯ù9ÄzIP AQ‚ÀciÅ ð %{¦“^zo®û/À¤ œh‡Nz"ÑDI8‚Îкê#¬ Ÿl£ ô|iÁ rÀȪ…ÀÔðAðÌÆ!„ÓP²/AñÌCÏ[­§œpÀ%ˆàœÇâ]îZרAy`òoY Ô-­®%ñÐô2ú Ði@ àâéÇ]¯½Ö[¼$¸OÁÕ@èT ÁØ‹¢PBêº+¾–`&ca Öô5(¡ê®¨¢WP<Ðñ4o>B"³§Àls;îvöæ— ºÔÿ@eë¾;”¿¼²™ ŠÖôæ#î¸O“‡‚¸yªAž8à?µ@ 3šx_U‚§zÔ€ÞV±ãéO‚ÕrAÕr¾”DM°Ù *ö‚×¼¯k#Øáð>æ‰V*,_tÀm.0X &à œî!xÀ G8À(À `0‚± u#8·8±’À× ð54®…Ø“"p  ë`Á¡ã =UC<ÆJ-p!ù)ž £$Á *BQ¡°_i„Œ`N€8*þe@”©Œ(S©J h V‹dA¿PKã «BØ €‹(À,x'"SÇ`Eà+­œ€1[ BªÀ@%)]Ï``Õ"é©jö%-T™ø €œ›  îØJ8à˜*NžÞI‚xúñ%Sù#0OÎ1ršŽb_j9&“št¢BýþVË´ Ø#Wà‚ 2ž*Ѐ 9S t £û HCªÆµÔvÎú BUºR@ñ€¸L‘iÈ´sò4¤=9:/36•š„^Hõ-U˜+e¨:§ÊÌŠfÕ Øh"_ùU°Š”/$k˜˜R´®´“0xiLUp€P”å£ °‚Æb üa]íZÔ5–LJw ædEÁº@ÎäYyâóÐ( ‚3NÖŸ× f+›9¿6õŽ0•©[ÀTÀª0líùS~Ö¶|!˜^ÃC[™1u³5dë(}»¤r«Æ=nX&[*Ü„—;ëfåþÙ¶þ–™[•æx)åa3B•Ä;^|æöpÀaIp^År€qMÁ²ÛTä²¥mX¹¬*ð67Z ìñ¤zÌà“‘ïTA/‚Ø3$°B l>íÀb¹TÊ,ŸûÕü V·4Ý€*`Øüfu«…ê›vUlÙMm¯ÀÔìq‘WÞé®àš<ÕêFaÕ¹?ؽwð}›jBC ÿýïN ¯¿pM}J3Þ •¨ä ÌÚׇOª)€iZ€et-XÁVЂÜ9|:ñv“EV* ´‰¶E#­¼Ñ"¥>"R .ÐÃ& šÇi™Jt¨‚þwµY¡UFT-ãÜ‚R›:Ϩ>u©SÀÑîŽÀ$f¡sê0ÎOÎÕ»®wÍëþ¥à¼Ö5±`0ÅݯŸƒv]dG l®†¯Ò‹ðB)°è®u 8áÍÐlMl¦’[bˆh°¥sh‹¬å¹AfæFÅ-‚á8ùÁŽ5aŽåFÝ )\`Ýk7{ýP!¸à ™ñÀ6ÙE„„¡Ðrþ¬'YÀj›ÛÉí®ŽÁ†Š¼Àq»ysm•>^õ[pyuîÝ Ga„ ‰ Âj~0B-jø}ԜĴžN·ËHqÄ0 ØxŠÀ‰ ;7Nûq'[>®þôÇ9I¬v} ˜°Ís§ºhž™³žTòþwIÔP azˆÞØ5WöŸ×õœiOÜ# v5$l†b_1®«^ËZ†ý]^Rø¼ƒlíiyKÛ¥^Áû¼ð{@ò¤§6^Ý›ƒü_0Ø ˜‘^d˜Ù4 }~-|€N½g–q™¸JÖ,_³nãóÀ—}Ôúø "ø;}–§37xWþü÷Á~ lâyÇ òGãOà…mâëƒ0ÝI káO#èO¿úÙ0Lëjí$s6÷!ól(  [5:úñkŸg|Åp[µ7.ÊÃwÀ08BÀHµþ÷¡ Îfp{±7£BW+eÑ— ·t×W>o(ª@ nôw¬â\0-f÷UŒ|  _e+î·:ð–FsC€V€Ta÷äƒÿgw+%ÑW2òUxp)pd§rMGà;Ø€WP%ÐS„Á"_{X `h}0PjÑC2Gϲ _GSa g•&Y%€ÁW _eqˆrY=EˆVSvðSq¾–6‹ÔZ{|DñyÏPWP$E‡>ÈÑ€JS ÷7jà¡€–|ñ ŸÇ “¥V ”UK…6|RK¾þ¶zgÇRŸ‡(@…W‰“ÕI[µ$ˆ½“i«vE$¡ ®U}Ç‹¾XK¬¸Pl¨à¤S?Q ½Ñz‰Xq$¦Æj‹ÑX1ÓØe°Q¤VjAÂg>UKç…œ0yñ{KQ—Ž§fŽ)&€Pà$ÐX«–ŽéØSáR¡f`7(.ˆˆÃÑ°‘Ga&v†g}öbÁÕRG¡OñMÇQ†Ø€ PP€Kå Gà?uNÇ6$sÀÖB>“xqAv†b[àÅøwtp¡2dqz‘1(Ð1Uy‘þ·•[°qÕ3jHÀµ&‡¥Fhø“3 bˆINúˆ>-cp°Ã=öhKqB™+Ð?9!tI yÉ!0‹gÀ >¡p àÀ–Ò?~­ÃSQŒÛ“ë²*³¶d×xGQ!äoMÁP 9™«ñ<œ!ðcü‘—ú±ŸÙx¿Ùxáävy ?A…á~rM‹¹ƒé7èG›ÁpK xŒSp QP© I· ÍÙ÷±>f^Ñ ßiÜ·ƒ¶© m—tª` óá„n2"ÇsñŒÐMð¹“:ø,†ñ}g7€ó a@ }0uái¿qDý¹› ƒ´ƒ‚ë9³³ƒ;‰ ŸZ1˜Ó—FÔ è‚ NÛÙ¡WÀ$I¡Ü`šGPšº¡¢šÀ¡3—ž—% ± 4j £’0—TÉàÒÙ£WðE$@`¤HÊß4ŸØ¡A;festival/doc/cstr.gif0000644000175000017500000000132407262605651013217 0ustar awbawbGIF87a|?ñÿÿÿÿ,|?þ”©Ëí£œ Ø‹³Þ¼û†âx …¦êÊBgÇòL×öçúÎÃÑùê ‡Ä¢1P)ULǦó)<™6 ŠÍj‘Ui×܊ǽ®ÂbJ“×lÐ˶çí)|¦ë·fG~øôÖ`høÔ‡€VÅxuhôWR¸÷ˆItÀ¡xy虩³™±©6÷ˆ:4!—*ªªs·P©•Šé“s;s–¶ä*ôÚk3ŠQúgt™z zÅ[ìó%Ù À—C,£ Y=ʨ\»[­[Nþ™È˜îô|Ã-ÍEú‹ˆŽéÙÌ¥ëœH‘p°¼âvn\¬;öáSw@4ÓúQô—ož­þ8hj‰£‘c.ŠçÔÕ8XGÉ7‡øBN|‰‹$L^ûÒi\fG •VˆHZüérZI}"óm‘©N¢%ar™hêϘÛLKç"wXDÀ4*Ôˆdû¥ªÐWD‰Ù~hlºm,[° žJ›0ïYbyÈsáÎ×^½>é…`W.U¦C³J¨”+0–æð–y·Áݧ¢Î»"n‚•Å&n¡Y&Ù—x颾Yäp¬Å®~ÍYµkÌrC?H kÈ̯#$6mõñÉ$x ®ê›»8á‘Uµ6ƒá,SˈG‹`6«Ö[늣YkÄ®©{eÃü¼°ê]².ÞRüdί›üƒM9Ú wÔ~ˆ1TPôÑãØJü So 1¨ÖÁ‘Á*i1Üy׌áÌÊ°Ý…§PG`(n½ƒ¡{#úBE0½hÝh\}H£*ƒ˜GYŽ*N‘ÄN)úˆáVì°H¤ÔسS’*&¥”“ô¡WRvä1^iÈ‘‹4Ée@FfÙc˜t°““Gfžù€Ò½ gœ€QBg5u¦ÙQž@êygŸyÚù'ž€J2ˆl;P;festival/doc/est.jpg0000644000175000017500000001045607262605651013060 0ustar awbawbÿØÿàJFIFÿþICREATOR: XV Version 3.10a Rev: 12/29/94 Quality = 60, Smoothing = 15 ÿÛC    ' .)10.)-,3:J>36F7,-@WAFLNRSR2>ZaZP`JQROÿÛC&&O5-5OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOÿÀÉ"ÿÄ ÿĵ}!1AQa"q2‘¡#B±ÁRÑð$3br‚ %&'()*456789:CDEFGHIJSTUVWXYZcdefghijstuvwxyzƒ„…†‡ˆ‰Š’“”•–—˜™š¢£¤¥¦§¨©ª²³´µ¶·¸¹ºÂÃÄÅÆÇÈÉÊÒÓÔÕÖ×ØÙÚáâãäåæçèéêñòóôõö÷øùúÿÄ ÿĵw!1AQaq"2B‘¡±Á #3RðbrÑ $4á%ñ&'()*56789:CDEFGHIJSTUVWXYZcdefghijstuvwxyz‚ƒ„…†‡ˆ‰Š’“”•–—˜™š¢£¤¥¦§¨©ª²³´µ¶·¸¹ºÂÃÄÅÆÇÈÉÊÒÓÔÕÖ×ØÙÚâãäåæçèéêòóôõö÷øùúÿÚ ?ôê(¢€ (¢€ (¢€ )2(Ü(h¤Ü(Ü(h¤Ü(È ¢Š(¢‘›hÍyÕŸÄÛc©ÞC¨G¶Ùd"ÞX”œ¨'ïsôé@Erâ‡;êÛ?øš_øX>ÿ þA“ÿ‰ ºŠäáaxoþ‚ùOþ&øX>ÿ þA“ÿ‰ ºŠä¿á`økþ‚?ùOþ&²5_‰¶‹—£B÷,@S"aë“@‰EgØÜÊöÑ}¤§Ÿ°y›v9ǵ_S‘@ EPEPEPX>-Ôµ]7NI4[µÜ4€(X*àäàí[Õ ãå4ä¯ñ3]†gŽêÂÑ]x)±Ôƒï“QÉñ?X8òí,××pcýEQøÑŸ€XT>;žOòÅsºt?hÔ ˆŒ†q‘íÔÐn|yâµÆt¨FFFmäÿâª)~"ø’õÖ±ÿ½ ýšºM?| V£Ùù‘t ¤rÎh‚µÌÿǽýðÿüUz–Ÿzó[ÄÓ`HÊ ÀÎ9¯ñF‡‰"‚ ÆÖ ØpGùõ¯PÓ² "PÁ÷M@ïŽ5?ì¿ ^N­‰<¨ýw7}9?…y>‡áYõVêéÌí÷p>g÷ƒÞ½7ÆÞºñ±Š9ãŽÖwÌ­œ·AǾ3ùÕ°ª…*(  tPkàí!q¾eÇ÷ä#?–+Æšf™¥Em•°Ši’w±ùGÔûþ•êÑX…\⼋Ç×bãÄóD‡)l¢!õêR*ç#C$Š‹Éb¯D³ðÖ‹"({=ì¼Ç÷ë\^ŸªÅÆD9ü:~¸¯TÑ­·‘@—žÒ¦ˆýe·~Å\°ür>Ó›þ…·gì²3>=TÿŽ+Ú¾Æ }+‘Ñ<+q¦j7×WrG+Ü>P¦x’sŸÂ€:-5Ù±šÞ‹îŠË°·ÙŽ+Y:Š( Š( Š( ¢œ€‡5-sÞ1ÕSOðÝôÊàIå”N‰¸΀òüëÊt¿ Üj0‹½Bå£óFå_¼íîIéZ’xÑâýÅÜèÿíá‡èz‚ÜÃ:ŽEu= œƒNERkÀî£Õ<;¨4 q,CDä¾â»ïêºÝƒÏvò‚ûc Àøóü¨¼¿¸ŽÎÂ{™8Hcgo ¯®§{«¹®%9y\»rs^µãíM­ü'${þ{§X‡Ó©ý?ò ê|kŸ2r>ó‡ùý+Ô´h0ƒŠâ<-n±ÚÁ>ldýO5èÚl[c}Så¨ÞOJ°:Q@GZ–Š(¢Š(¢Šiu†hY‚Œ“Šƒ ƒš‚XZNCþE2pháé^añní#‚ÎÆ3óHÆGÐp?™ü«Ðî/#·‹·=x‡Žµ&Ô¼M;òÂJ=1×õ&ØçÑKº¢õc[0è—oÈBjžžýxáã]î›dXM&ÄœVìäî|9ªK –`9Ù¼çõâ©øz&³ #ýQÞ}±ÓõÅzy°`Ÿýjå|5¢^}ªòêæÝá.Ø@ã‚I?Ò†š.ðoBω5)íô9vÈCLDyúõý® ¾|~gÝÜ3ô®ãÆVr¦Š SˆåV?NGõÁÒ µ}M³f‘‡Ìkn_g k‰ð¶ª&E‚FĨ1ÏñZïl™¦PfƒhÊÔâ~ Á¶ÞÖF9`åGÐþµjx|Ig¦ÚĪ0õ<Ÿç].£ Ã}ý®(ä ÊïÅT6^QÂãv¡+™·N÷8ˆZ™º¼µ´³vÿëÖ¹‹>Ñ{ =™†~êmjëíº½ÍÀ9Vrÿ²8 «>€ÉvÒvAõ4»\ôm Ü.®®+†TAòN{W)£´°°ä{×_ew ¨·½åv¹v#!¼ëRT2¼¿òÉA½5c”Ï&  ’Å™m-ÍI@Q@ù+…85NKiI-¸ùR²Ìóg¦p@Âãë@Á€åÉ{Õ{Ûì_{¹«óù6ä½Xè8ÜÜ“A­(s2¶¡:ÚYÏu)È ŸÀf¼BiZižW9gbÄú“^—ñTKm'ìHãι `u :Ÿé^cAUÚ¿*:/ ÁÃÊGÞm£ð¯EÒ¡r^HÚÆ!Á럮k½´´0Ñ6¥i Ûs’pobÇ–6Õ+ÆŽÒ ndm©–cì*ÍÅÊZÆ^å–4^¬ÇWœøËÅ+ªcMÓšß¼p?ÖÀ{Ÿ®²”lcJçGi«Ùø›H¸ŒÛÈ-÷„ŽÆ¼ÇQ²—O½’ÚaÊžfzŽ…bšv• ¢¨f2ÝZƒÄ>U´8_.då¡ö®^du$Ï.ŠW†U–&*êrí^áOGwn#*åξ¾ãÚ¼ÎæÞ[K‡‚t+"im.e³¹Iàm®‡?_j¤'{ì·uù›5•â{ÕÓô+©ó†ØU?Þ<ç\ô>.·´°†æUwóz*c ŽµÌx—Ä×ë$b?&ÙBg$ŸSZóÅGCI9j`WWá˜qj­Œ—b¥r•ÐèZô60ˆn vÇÝdÁ?•csf®z-’„9=…Qµ{íZ}ö÷¶‰˜ˆcGe‰êx­ G’1ÇCڹǹ4¿Øò^›RÄE9çi8ïõ¤¶:ãn]NÓGÔ. …!¾”L㬀c?…n•3a•ÈÕÄèÌcI&prwM÷º÷®§I¹.¾[E3*¶¨ºöªÃïúÔˆUMàŸ­1âv'µD-YpmÇÔÐd[¢œsÖŠd¦@?v>ô±îÛóš¯,² 1÷G®3H&mÀ+ü(¾²ÅôÍg“û¾=+rî<OZÂÁRcq‚(:0òIØòÛÿ k’ÌÓ]˲¬Òk=¼;~­´ˆóþõzÔöûêÓƒL8¡è:”’W<úËAÖã;­&òû2­a¥øÅ“jꎣÚá‡ò®Þ 8ãCÆM9 và ¢ª1lä”Ò<ÒûÃÞ ™ÿÒæ7Ï»ùÒiZ4Ö„w:—Q|À—ÝÚ½¬³Ú¸ÿí¶±‚÷¥|þÿ®ET©Ù—v:}2Hî#Dé"pÊrã[ky'#W)áwXô[0­ÿ,ÇCß½uVñËp *œzšÉÓfªjçâ5 ?ÏÒb©ïJócÁÁ¯y¼²’49\¯µxŽ°¨šÅâÅ÷ÎçM-úš¾häi!š5‘T†PÃ=zÿ!]å­­£Ç„·‰ ôA\/„à%¤”Ž +Ñ´øAA‘ZFFrŸ+*K¡Å/XPƒþÈ¡4 xÔ·ÙbÈÿ`WCúsìjÁxÚ »[2ƒCE#Dx“îm;¾•ËÄEk7_1@™Är%=Žy®Úæ&-"ÆÁ\© Øέqš­•ôZ•´RÏi#Ünùͪäm¨[©{¦—‡®§¹´ßq‚ÁʇÀèØ÷®¯Hæbk—Óc¸6O(–LðUvñô®ÏIµòbÜÝM2jKK'8ã­G”±Þ —®vŽµ vì@òÇçAÎ[¢’9¢€€z€jš(Ûyö=!õ€+ ¼Ÿ¸v÷¦]Z,éæF0Õp*Ž€RÐó#FÛe>´"8}+vhãt"@1Y²iÎt íIêýµãf@ˆ9ã½;h¦¿™ýädbÈ°ä…ö­ãQ(œ¥'->Ð=ý+Ëþ$´ŸÛ–ñ¶B‹pÀvoð¯ZŠkX×å>¦²5Ë +Tùï-"•Àگ̦zÔJ§1¬(òêy—†uĵÛkrû£žƒØ× [ë¨a.b ýàâ¹-GÂ6~q6¾lkØÈýjœ~ iþ? úçŸëP]Ž“^ñÄV–¯œËqréʧ¹?Ò¼îÊÆëS¹+ –$åÜôä×V¾¶·“÷¯,Øí÷Aü¹­½:É AqˆÐº£ŠºN’ÖѤq.å_ֺˡX`ú[! x­3-«¦UÆ|¤Ê—1éHßtÔNè‡÷M‘èiê^O¸„ÖŽªhÅQ’d2&bNk™Õ綟XÓå†âH¼ÍäH¼d`w®²M:f†G”€6ž? àl®,¬ü7o5χ#˜ÁiÊFK{úÖ c»ÚÚ<¨ít›Á;Ï Ž•¼“ aT•Çx9ž×MLãìè²#RxP}«­Šaº6#4^û’$±HÛp3èEH"@rf› »©©hQEQER6qòžih  RÃ3“æGlT°ÂÁT³ŽÂ¬Q@O"F™`µUe¶p ‘í&­In²6âNiŸeV »ÇJ®tÈÈÊ; Õit·í"Ÿc[#Ú©Æ Ý’Oµ)ÉhóvƒP&ulªWSU/,¡X©>†¹·¹„ÚeöY*h´y± V”NÍ:£y õ­Ô›[ɦ8ûÒ*Õ„Ó€L„juÊ”œm< ÒÛ ]»þ_JÎO¨ï²Án»¶§¬ùŒ4j0:Š³Œõ¨¼{‰ŒõÁ €|Oq¸ô¬С·Ó#ÓÆdŠ4Ø u"º5EUÂŒ B€Ð/£è+¦Â° DSòïê¥t° D S¼µ§Š(¢Š(¢ŠÿÙfestival/doc/est_small.jpg0000644000175000017500000000371707262605651014252 0ustar awbawbÿØÿàJFIFÿþICREATOR: XV Version 3.10a Rev: 12/29/94 Quality = 60, Smoothing = 15 ÿÛC    ' .)10.)-,3:J>36F7,-@WAFLNRSR2>ZaZP`JQROÿÛC&&O5-5OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOÿÀ?d"ÿÄ ÿĵ}!1AQa"q2‘¡#B±ÁRÑð$3br‚ %&'()*456789:CDEFGHIJSTUVWXYZcdefghijstuvwxyzƒ„…†‡ˆ‰Š’“”•–—˜™š¢£¤¥¦§¨©ª²³´µ¶·¸¹ºÂÃÄÅÆÇÈÉÊÒÓÔÕÖ×ØÙÚáâãäåæçèéêñòóôõö÷øùúÿÄ ÿĵw!1AQaq"2B‘¡±Á #3RðbrÑ $4á%ñ&'()*56789:CDEFGHIJSTUVWXYZcdefghijstuvwxyz‚ƒ„…†‡ˆ‰Š’“”•–—˜™š¢£¤¥¦§¨©ª²³´µ¶·¸¹ºÂÃÄÅÆÇÈÉÊÒÓÔÕÖ×ØÙÚâãäåæçèéêòóôõö÷øùúÿÚ ?ôê(¢€ LZ­©C%Å„°Å<–îãXþòý+Åoµÿ[ëiÑë7˜Æ<žq@ç¸zÑ‘ë^m›®\A¾ßŽ7EIàCVûuý¾©w-ÃC Œn9Î?JôjËÕuý7H¸‚BåakŒì-ÓSÛ­i!Ê\&¹ I­ø–k­B6Öê±[¡´é7/Ð`@F£ÌH ¸vcè­¨¥Y+^s©øEc×ôë‹;5Ž$Üò²F6ñõ®óL’¿­^¢Š(¢:œTM$ÿÕå}A ukÈì4Ë‹©H lÿ\ ð} Mö½çÌrwžäŸþ½z_Å Z8<8m#`e¸p‡€äÿ/Ö¼ïÚ:ê ZD$ÀÁ"€õ=2mvËEÑüÆ $Ä[¹‘APjšÎ™áôk¹ QqpwyQu‘±Ï?Ö¹Ÿ ›}bÄÚÃ)RK>r@Ç#šÇñcϼ¾vïÝ¢• éÿëͲ;ÔñGŠnMo¥Z$ddFòäTºÿµ5²ºÓž „Í1úÕ_ Ï ýª¼d±aÐA¥þšÃX»Ô  ªç4à’½Îw↥öÝvXÏÉoÈÿi¹þX®»À–8†>êŒý{טù¯ªø‘§|¿™)cô+Ø|36F7,-@WAFLNRSR2>ZaZP`JQROÿÛC&&O5-5OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOÿÀ•ë"ÿÄ ÿĵ}!1AQa"q2‘¡#B±ÁRÑð$3br‚ %&'()*456789:CDEFGHIJSTUVWXYZcdefghijstuvwxyzƒ„…†‡ˆ‰Š’“”•–—˜™š¢£¤¥¦§¨©ª²³´µ¶·¸¹ºÂÃÄÅÆÇÈÉÊÒÓÔÕÖ×ØÙÚáâãäåæçèéêñòóôõö÷øùúÿÄ ÿĵw!1AQaq"2B‘¡±Á #3RðbrÑ $4á%ñ&'()*56789:CDEFGHIJSTUVWXYZcdefghijstuvwxyz‚ƒ„…†‡ˆ‰Š’“”•–—˜™š¢£¤¥¦§¨©ª²³´µ¶·¸¹ºÂÃÄÅÆÇÈÉÊÒÓÔÕÖ×ØÙÚâãäåæçèéêòóôõö÷øùúÿÚ ?ôê(¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ ÇñN¥&—¡Íq„œ‘3É?á“øV»º¢v ª2I8W—xÏ\MZýa·$ÛÛäÙsô­)ÚFU§ÉoÁþ)žêv±Ô]¤•ÉhŸoÔ§Ò­ø³Å­¢IoªÇ$ò|î®tíÜŸå^so4–÷Ï mt`Ëõ»©Ëªjfyr6¢ \csú’X˜ò+£L½ª²å—CÛm/"»µŠâ&%@Ê}X yo„|HlâoˆÌnOÝö>ÕßÙÞ¬ À沌”‘½j2§+3RŠj0aNª1 (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¢€ (¤WWVgZ(¢€ (¢€ (éY÷ÚÞbJÜ\¦ñÆÅù›?AÓñ¦“{)(«¶hUkëû]>ÜÜ^L±Æ2yçé\¥ÿŽ•Óí=0óý”r×WW³´·s¼ŽIÆãœg°ôÕ¼0ò{œµ1ZGScÄž*ŸTV¶µS ®HnreçŒñÇN•ÌR•ü©„éPQVGª9»²,f¨Þ®Ùû G¥W¼}»`r9hsA£« [ÙÕL†âºÄ7:{|ËìO#ñ®V«Ñµx²¼]Ñö4Ô+C–g®húõ¦ ¸‚\²õR0En$‡Zñœ«V Ž„VîŸâ½JÁ2Vx×þzd·çW JÚG-l¦KZNçªç4Weãû&ÂÞ[Ë =Jüê?¯é] ˆt«í¢Úúf襶·äy®…8½™æÏR5h¦‰ô4슣¢ŒŠ‚êòÚÎ5ÌñÅõflP 6OEy¿‰> JÒy:ÕO3²ƒ»èé]w†õYuÒêä9ÓF9 ãÚͤ#ŠzÙ²LNLúôö«ét Œ¬1üÃøjóeN¤å©ïBµPVe‹co(\õ*1ÐÔòg,Ä–Ï'Þ£#¡ö®¸C–68*Uö’æ355â9$RÙÆÒ{ýj]IGÙ·ѪƆaiYÔØšá®—>§­ƒsöiÄY,˜!uSJ¤éé]»>›¯™%ÌXÆ@ ?@+••åvQµK ¯>¬T^‡ÑàêÔ©ª-º•ˆ4•60O$êj¾‡DÒŠ»z挊œB‡¹ëéHö·)“튿g+^Ç*ÄÓnÉ•ÈÂ)ô†’4–£­—3g 3þèt˜ËH+Ñ8-Ž§ÿŸÂº­ ºñ]4Ö‡‘Š•ê[±Úhñíkux§G¶1ZC¥hsQ@Q@Q@Q@Q@Q@Q@Q@Q@Q@2Yc†&–gXãA–f8z“O¯'ñ>±sâÿÃáÍ>cˆ›cüeyf> `à{gè±­üP±µ‘¡ÒmšñÔãÍcµ?çô¬»/‹+&/ôÈx\©ÎJï4Ï èše²Co§ÀÅG2H©&¹OÃà«VßÛ´W¤d Àõn‹ùó@šGßkðqÐ1¿zõ­2Ê=;M¶±‰™’Þ5Œ3u8ÍY¢Š(¢Š(¢Š(¢Š(¤#"–ŠæüObe·óÓ†ˆ~•Ëä+ÑnbFU€ ŒkƒÕ-~Åv뱶1ÊßÿÕ^Ž¯Øg‹™áìý¢ù•]˜DÅ[qšæBI4ÌeÉryÏZé‡(8ïŽþÕÛ@Òy…n¤Ž3[bhʪ\¬çÀâaBüËr…®˜ÒÅ]]=`l’Wc•¢S È?xžyÉïXÑÁ¸¾i8œÅJ<”ÖäEå댅À‘άm^ùùºÓ$°qߊíhòS+0—·Lžj½OAž\e99 IíŽùâ¡p7y Îó¬Ú5ŒŠÅyÎî õ¨Š¯fQÎXóÓ¥BW=k6¢ÌÝPdxþ/ëY¶çšÓÖ0–$z°ŸÆ²mÛ¼¼oÄ}>NýÏ™¦‹•¤e"–â¤l^Sm3ëSŽ…}ÄT‘ë‚ÀcœLqQÖ³º8ñý¤y»eeæÅjÉa µ¹šwTEÉ®9&’<ìv]Ã¥yžC™˜ú“šßÛyrÀ+ë!& esÜ,vý*#RfŸ—9Æß›ëXÇVvÔj1¿bÝ´@2 íÁ>µÚhÿtâ¹:ò+¾Ñ-¶¢ñ]‰XðdÜÙ¿j›PUšdK…§ÓQEQEQEr^ñíߌu}ël‘@ZHdÀ*6ð97^½zöëh¢Š(¢Š(¢Š(¢Š(¢Š(¢Š(¢Š«©´‰¥Ý¼9ó)Ž¹ÁÅxg¯#²ñ–<§ e1“é¸νö¸xþéI­ ï´Kä <Ám€ïŒúgô ÖiR^Y*"–b{_;ß\\ëþ ’lžò|"“êp£ðàW°|GÕÆ—ái¢FÄן¹Ažp~ñü¸ükƒøY¤›ïÙ1Y&üöÞxQüÏá@¹¤ØǦivÖ0ýÈ# œc$Oây«tQ@Q@Q@KYÔbÒt›B~R-ïÃñ8å^×¼MâÛÛJhâݾQV0rz~\ç­{5äŽSæ§mƒª‡Ô ³§j–¤m&Ÿw ©Ãym¿_Jñ½KÒ´Ãk6f×nvÉ.[?g^¸Ç¯o^µ§ðÐ&‹4ú¾­r¶v³ÇäÂ%8óNA$àc¯½zۺƌîÁUFI' Í·ñ‹ut-­õKI&'Q’}­y×ÅmbòK›K(d+§Ë™Jž%$ž¿N8­ x/Ãz®‹g«J“¾èñ,FbxcÆä׿¥Tø©âX.#E±™dÚþeÃ!Èt_SøU‡~ Ñ´¦û\w/yrás[€AÐuÏ\ô•o¦Á®xÃì\F+Y§*€ÛcNO°Í{Å…¶e¥¤+1.ÕP?Ï4û‹ˆmmÞâæTŠ$gs€ÖªÚkZ]í¼·×ÐI ?ë$6§Ôö¯0ø§â¼ÔÆnÄAjA—ïÉèçšÕ—@k„3E´­Äª·Röþ p~Šâ(Ðlï¬ï£i,® ¸E8-À>œS/õ; 5¯îá·W8S+…ÏÓ5æ?®Ýu[û,’B%Ç`U€ÿÙ©ßôûÖÕ­oßjДEî0Ë7Nsíì(Õ!š;ˆRhdŠE Ž§!èE6ææÞÖ-ÔñÃêò0P?\ÏŠƒ¼%§é ,Ú¯ ºCž~o¦qŠ­áŸ ]ø¤®¹â«™¦ŽNa€±]Ã׺¾Ã¯_¨Myâ_LL_Ú¶¤ž>÷ŸJÀ¾¶XYf·pöïÊ4|Œ}EfüA³ð¶ŒñÁg§«_•D„Eõ`SéÇò­MY?á“PÔmà‚Ú5f†8“hØ;óܜִjºr¹Í‰ÃƼ-ÔáüU}æζˆ~Xþfç©?ýjÖð­¨·Ó Œ0óüÄ÷ þ¦¹(RMCQU'æ•ù>•èqŽ ˆpp1Ûá]˜kÔ¨ê3ÏƵBŒhÄI¯­­ÿÖJ3ýÑÉü»UÖT¶#Šö%°jÔöV×w© Œ8=WJØß|0õÆ+JÒ®¾Œ°ÑÂ5ïïæ1µ9]Nر œàúôªpÙ¼–É$ä“Þ·¡±…@ÞÀ~!XÓˆ‡|gÖ¹}jÎÒ;\6-À£ ²ÄÏXóõ§8S°dzƒciîqÍT¾¸ŽÒÜÍ&pªp¹êzç]üªœ<ÎUêy³—×e3jEA$ ÇãÔÿŸjŠØFÞ¾µ_q’Vv9f$“W­òŠðjÕ\ÎLû\6^ÍB.Ö4-íäp3–†¥}8cî¿çÿÖªñÝÜBr’ÇŸçZVºÚ•eº…Kcä+À'ßÒ³S„ž¨è–F>ëº1®-žòCPjþ¥©É4;[N’†$O¦qY/s&7ŽäTÊ›¾†Ô±*P¼÷p›gŒàûU?,ŸïëMns€pÝ>µ5´`‘çV¯)î\Šã˹‘•ŽÍÜ{Ö½­ëY­Û£çÉNÜö%²}ÇJçN@ïS!v” œ¸ÚÏáZÍÜ÷}{[­=. •=¹,8Åo(ãŠùÿK:ÐÓ®—N󤳅#€=²®:Öµ‡üCcb¶j#œ*ü’H¥ŸÏ\óŽE{]à'ÅÞ û9Š[ùœØg'rä`àÔÖÞ:ñ ½¼°%éu‘‚Ã%01•={ÅẌüGmrð[H×;ïòKÙÉÛŽ™ç5Üê[W†. noþ"€ñ V¹Ò5-kf#-.@8ÏÜÔ×ccwýœwPcdd`ŽÄ~uã>4×/5+Èå¹·’’¤B G€TAÛšÊOk‚¶÷Ò¸ÀT8ÀçüM{/‹µÓ4Ønc—Ëap£=ƒ^i¨ÌÐx×ÄFA"zÖ}ÇŠõWOk=VS(_ž7#œ…Æ?k'R¿{Fk¨‹F%P¤=6€Aõé@5¥úêzý­ÐB„é²FAõX]Oò­o†ž"KKuÒž6c%Álú ~9üëÏ¢»¸¡–hÌ@¢²ñ×9ˆ$}*kYî-gÙ9Iíäólcë@FÑ^+¥ø¿\§î¯w™@Q'Í–ÆBÿßG«ÿ –¯u}8”Ih g¸AÄÑe¶Y€?‡^(Ðo5Si­ÛX¼yŽáu Ÿå\Æ«âÛË-yÑU¦‚ñ#b{¡$ÿ%ÅqÚ߈õ‰¤Ó¦¸ ¡§ºTé"¥[Çöé\åÍýÌ—%˜ªÎå䌔œ’8öÉ qÓ¼D·wö4{VòÑ."#± ’åRKâ.-"ßTi[ìÓȱn; þ÷¦+ÃàÔ/mÖÎæÚyÚ’¨s÷R>‡'õ­9ožëB0°uö«g'9Á–ç@ꎲ"¼lXdrõ§W‘i¾.Õíõoa#í³Ž m†s)¶8àã Ö¶tŸkK¦#O¦Ku1D»UbTï àûŠôJ+‡—ÆšÕ¬.|5+‚ 6>¿tðzúŽ†µ­™ ‚=ÿ•tG8ã­sV ³‹U›SÔåmJêSgQµ>‹Óéè+¥¢€3.ü=¢ÞFRçK´qÓ>P}äW™|Eð–• ÛAw§<‘´Òìò· c9óÇsÖ½‚¼oâô¾"ñ„ZU†eçÈ@: ùòýÚÆðæ“âG_í O’åˆËœã‘ƒ×ƒé]u¯Š<ˆåКvþûÚ8'ò W hZ\Z6m§Ã‚!Lþóu'ñ9«ôæž oÝøvîúýâÓà‰2m­¾üƒ NNž½ºWŸè³^ë6ðÁs ´…²&™€TÇ~{ú{×ÑL¡”«TŒ{×;qà_ \LÒÉ¥ fë²GAùrÑi^ðó‰õ;ìê,r#_ÞcþÀ8ÿ¾r8º»¼ñ\ÝZOg‘¡‚ † éÛN=ëÚ´Í JÒGüKì …¿¾,àGšv­¢éºÌ+§iê¿tž ýäP›øÆûJÕ¼£}–hå¿@ˆ±£è6áG }N*Ä©{á… û’êþR¡{Är>»Tý ®ßJ𶇤MçiútQËÙØ—aô,N? »©é–Zµ›Zj6ë<$çiÈÁõr yGÂù´Ë ËýOR¼‚ a#8î9$§ éë]eßšX§›EÒ..í`Vi.¤ýÜc¶zöãƒíZÖ¾ðÕ¤¾dZTE¿é£4ƒòbEl\XÚÝX½”ð#[:ìhñ€>~ÒŒˆâ“Vc‚IŒ·çŽY¿Óñ¯EñŠî5P‡FÒ¦}=aa-äß"ãÂ;ÿž+£´ðG†­&Ã¥DXtó¤“+fêÊÚîÆK)â o"ld¨ÛéÇJñ?‡úݦ¬]^^—*mYr]‹)~G­wúmŽ©¯Þmë–þT®m,SÜnϸÜÐ ­7Â:—p.,´ØÒUå]‹9Sê7ŠÛ ›æ¹kýhÜjNÙšmÓ7p çò¯f¿ñm…½ŸØü=²þícÄQCó$`¬GaþqÖ­ßx/Ú…ÛÝ]i¨ÓHrÌ®é“ë… fµ,4» 6ÜÛØÚCGª¢ãw××ñ ž¼Ù5]ad¿¹ÃÜÌ<Ùœð¹<“èvž5ñU¥Æ…‹¢Á*Ù#óÙp®©•}„çük¾ÁžŽ9t˜ g8l²ÿß$â¯jz—ªÚ%­ýœrÃÊ/ÝÛô#PxÂÕæ¥}4QE„S#7§ŸaúÕÉ5”{–‹L·’íí. ϹßGà? ÂâHô´Èìò;È’*[½*‡Ë‚ÀUPü+Zu¥Ofs×ÃB¶²Zœ|R3(,˜|rÏ?Z”u<~µ5ݤ–ï*¡•CΣ؜W«N¼$·>~¶¬œI† <ÿúéOœÔj„eP1žHÿ?Ö¢mJÙS!ËqÀQÏZÑՂݨT–Ñež„ð3ŸZÀÖä«äFxNþ§ü*k›Ù®7$cdgŒc’)m,ZFçbq*k–;Æ éKÚOs—X™$*ãUøÅu’ø}.`étaÔVÞ‘sfIzÆ:דV¡õX,D/inW ¨Ý(WìiÙϹlÑì¹FH‰¥“Ê>X;‚öÍF2Ä '°«Äg çùUˆ"Tû€–?Äz×D!)]|M:Z-YöžKDŽ&LÒš–>kíÆ›nìt5Ð .K§—‚Œ=Æy§[Z‹æR» AÇÔ•Ræw1lt;;É0 ”FøçžO–µAœaìáóZÎñ €3òfPGæ» ØG5þ·osåó†U‡bÎEu6ºm½­õÍÜ+µîB‰bW<ÿãÔÉ2|)áh|9%óC+2]*­ü!G|“úV¥¶‘§Ûnò­!Ÿî p:½EP¼Ñt»èš;«$W岃“œç>¹'ó5Iü%¢øo#°‚'‰ËˆlŒr?ZÜ¢€3í´=&Òóívºu´3ããŒ)ÇáZQ@GÅxco C1AæGr»[Œƒš·£xkL½Ò´-Eí£ZE’aÁAGãÖ ø«ÿ"šÿ×ÊZè<1ÿ"¾“ÿ^qè€9øwKû5ˆ‚Õ-³>ÂÑ()É#ó5À>†«¥kªù´»¤EeæRΧù^éqo ÌF)㇜òÍ0’|Z{ý¶?ý%eèˆ#Õ­%•Æ’Ó Û»æ*‘ïóUïø^æMGEÖV5–Î_0Ë“÷ –Qõè+ºÑ<;moªA­Û·–ÓY,RDÊNzpµ¿km¤>M´Ka‹Q€ $ŸÔšÌ_ hit.SN…d_ ó5~M>ÎYšY-bghÚ&%Ü„ä©õ«4P1yágñ WêÉöS Š[vŒ ¾ÞµÃj¾ –ËN%…À¹(‡.003Û$ž¾•ì×$R²"ºžÌ2(˜ÓügaâGÔ Hœ–å>ÎË÷Œí€G>¾õ¥?†ô™­"¶û",qH²¦3• ç¯\rx÷­z(¨Ó¬…ÜWKmÏ ÕpUOo¥XŽ(âFŠ˜±Ú1’zŸ­:Š(Àô¢Šòm?â¯ÿ dqjϵ—œc– ƒ÷}G-׃‚~•ë ‚2A®WÄÞÓ5ûƒu¹ínÏÞ’0¨î}ê;¾Ÿn-¢ñ]Ø·Æ6F@ô“€,x×Äë¢Ø[#æêwdQ§%3üD/SøÖgÃÏM¥—ÕõeÿN˜~í“=IÿhþŸ‰®HðÖ›¤ÈÓÅÍtßzæá·ÈÇêz~­Š(¢Š(¢Š(¢Š(¢Š(¢Š(¢Š(¢Š(¢Š(¢Š)’FSè  ‹Ë+¾Ñ²N»‚ Ô[+v 6›FlýÚŒhÏŸ»^ˆúz÷E4iÉýÚâ-ôS‘•­Û &2µ¾–H½ªÊB«Ú€(­‚„Æ+6ÿJºL cÆt 9½ÑˆbBò;ÖTšdŠpÚ½>k~ÕIô”'îÒ²+š]Ï;Kr~é­;M‰ZìSI@~í[†Áµ2L[ .2µ¯ý‘i4L“@¤9Ž0IkB8UG¥CHÒ*(v3ɦ3RQEQEQEQEQÖ´«mkM–ÆñIÇuSØŠ›O´[ :ÚͲÛı=€?¥X¢€ òÍ Úk¹<] ]ÅÒ6p$šõ:ྂ36F7,-@WAFLNRSR2>ZaZP`JQROÿÛC&&O5-5OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOÿÀK÷"ÿÄ ÿĵ}!1AQa"q2‘¡#B±ÁRÑð$3br‚ %&'()*456789:CDEFGHIJSTUVWXYZcdefghijstuvwxyzƒ„…†‡ˆ‰Š’“”•–—˜™š¢£¤¥¦§¨©ª²³´µ¶·¸¹ºÂÃÄÅÆÇÈÉÊÒÓÔÕÖ×ØÙÚáâãäåæçèéêñòóôõö÷øùúÿÄ ÿĵw!1AQaq"2B‘¡±Á #3RðbrÑ $4á%ñ&'()*56789:CDEFGHIJSTUVWXYZcdefghijstuvwxyz‚ƒ„…†‡ˆ‰Š’“”•–—˜™š¢£¤¥¦§¨©ª²³´µ¶·¸¹ºÂÃÄÅÆÇÈÉÊÒÓÔÕÖ×ØÙÚâãäåæçèéêòóôõö÷øùúÿÚ ?ôê(¢€ (¢€ (¢€ (¢€ ($¦Š(¢™,±Â…å‘QGRÇP°úÍ×õH´­*k‡`$*V5Ï,Ý«;Uñ}…š2Ú´ÍÓ Â©ÿ óý[S¼Õn|ëÉ7÷T}Õ€VÔè¶îÎZ¸˜ÅZ:³FÓÆ–z%ÜÌïpÔHÇqžyýk¬ð¿‰¢Õ4øD®>ÒŠ@z“ëø×–ÏøÙ{‘ÅC§^Ko"É ŒŽ½¬±7§+­Žì¾1ÄSpoT{ê°a‘K^¢øÜÆ‹£@ÿ–‰ýEuv^#Òop!½‹qþ;Oäk8ÔŒ¶f•pµi|HÕ¢:°È`h,É"¬çŠÆ_irkI¥C1–væA” àŸ^+aX7CEî7·Š( AEPEPEPEPEPEPEPEP?‰RAd—³+Âû²§·JÂÄ‚(ѽÙk¯»…n-¤…ÇÊêA¯=™ Þ¡#ò¯C¡8òÉlxùƒ©JjpvLЛÄZ›‰U?ÝQýkò{‹¦Ýs3ÈÚn•;à)$à{Õs$.À Q jêtá ‘ç{ZÓݶP•*¼‹“Z2GÆpjŠ[™®Îv¢ž+ ²qZ#« QêìWaY. wn¸ÀÎEn̉æ(½³Y³DPŒg…râaÍOSÓËë{*Új„I ŠG`ÜŠÕÔ´Ôµ¶Ž]êKœmÏ5’c=«Çq啱…_iNëbXµ èÝÏÈE$ú¥üë¶{ˉѤ$T~T†j£¯U¹‡¸Þ††ƒpmµT¸^¨Ôcú×®h/qlûבi—}Í{ƒ•dŸJ馭ÉÅJõ«EUœáEPEPEPEPY^ ÖáÑ,–gC,Ò¸ŽTó#ž‚µkñ΋ªj:žy¥Æ%û$Û™ *Aç·_V4·¶mNM[N´Tû9(öÜÓZ~Ö5ÍR&:¾Qì Äl6Éô?sÿæ“UÖôŸÛ1Ý#‡”Ùàgè7ô+X#µµŠÞÛHG -Q@¯x’ÇB’Ú…–Yî[lQD¹cÎ?­;Äzßö˜·d–åÂ,i×'Ö€5ë“ñMŽÉVò1Áù_úê`s,ÈÈP²‚Tõ\ö¨ï-ÒæÝãqô¥QÓ—1†"Š­MÅž_l÷q GÀ$v5$6ðÄ ¤ý\½µ{§ŠN@û§ÔW=â[¹•-ì혉'nvœW£R4Úö¯SÆ¡R´_ÕÖ†´‰òžV’>>•r|›hâ,[b€IïQ̳©­ä´»8¢ß5–¦t‰ƒÓ5‹ªü·1vàÿ:ÜYRgo-IQüDVµÍú(êú×)§I´{9b’Ä$ÉŒsÍFIFõØË*Š·W„“LûÆã(XѳÔ- OÞÛ3·×ŠÎ„Ó3ª…ÞÙÀíQ9 2N*[5ódÈä[{Ó8§A6Ž‡ÃfK´ã^¯i—¯µr>Ó¶ ‘–»@01]IXñäîîÅ¢Š(WRÔ-ôË_´Ý±U$ ã'ùÕªç|yÿ"¼ßõÖ/ý V­–©ix‹åÌ¢CÇ–Xnžß  ´QEQEQE2Y#†3$®¨‹Éf8©A­é7àÔ­$qÕVe'ùÔ—úm¦¢Ðý²?5amË”ŸqÐþ5É|CÒôK_ËqýŸ]»,p—c'Û¯  ^ ˆë¾5Õ¼C'ÍNc€Ÿ~ä£õ®×\Ö¬´+»¿“jôU³Ÿ@+”ðïƒ5]3N„ÚkóÙI*‡šº†#ßò¨|Qà[Rº³¸Q’øGþµgp‡¯UÀÀãÛó  Vž9»—Äštúdq¥Ê‡$M¹âSÈ-Æâ¯Ýø®âO>£X%ÔШyZI„c}Þ9ê++QðUíÞ‹-½šYÙHä1PÍ$’‘ýùòÆ3SÚx[S3ɪKö{mA,~Í D唶ݻØãô  m55oxÞ}V1mi­å§˜ ˆÈÀÆ2zœý+§›ÅŒ|E6“gh.~ÉKs(l` }p?§á Ä:FŒlQ¬íX»HódÊîON0èyúUO øSÄx¿{‹‹Xg¼$É0ýã°çØdž¼ý(NÛÆfçÁ—ôV Ò@å|àäsœtÁ¥Mÿ |o Œöoqª^B²}’6ÿVÎYAY¾ð–«c¦6ŸªÏ Y™Œ­'-)ÀcÙxè:Ó€ ÔÖÔªÙ¥-ŽZô9“”>&¬T¹–ûpF›?¼94Ʋ¹¹E72 £œt«Æ9-Ï—6wlf”È€|ΣêkЕ5S^m"åGÝäÕD d(À©®JêAu©É"Ÿ—8_ ®šöèL­¿Î_‚GLV;è—V½òÙã=Àé\8ÚŠÜ°è{>ºŽu´lžÑ-žB¤÷aÇçT¯Å¬mŠ`ã¾âŸò²â›åÉ#“ÀÏjòÔù´hú‰ÐT}èËO3âà±Ú:Å-¥ë[¸8$ôÇôÍj¶”]ÆìoâÇ~k£Ð¼·°™—•ª°cÈRqø]QVG‹Vnr»eÿ xöÆÒÊÞJ ¢a•y•r¼¯5¨~'h"]».¶îÆï/¶zãÒ·tŸ iÚvš–fæ„Ž]rM Áéœt÷«:‘tŸMµ|‚2bçÞ¨ÈÆÓ~ h7îêf’¤ãÌC‚=r3Xú¯Ž¬ì|V“Å;Íaö`Ž¨:¶IÈüñù×qk¦XY—6¶pB_ŠFìz×9®éöw~4Òa¸¶ŠHå·]JŽFrþ-ñŽ³¦˜ìZUŒ§ÎŽ0wR8ú\Ì~!”]-Ò礨èHã _ƒÿ} ôøkOO±]Ck q‰£„ª.Üîp9Ç^ ¬À}iå͵-ŒŠë(ûÌ¥X{ƒƒŠèl>!éÛf=$D]ÀÇ÷›ãT5¿ˆöóŸ²<…R܆çóR®¾ßBÒíÔì±·ÜÈÛË8¿•QÕ<#¥jNò#Ó.îí5»[B6¼0í#„dcÿ×W?±®umj OX8my¶³6Öþûž„û­tTPEPEPEPEPHÀ0Á¥¢€25-µ9Qšç¥ð€/‘Ò»ŠJdr¶Š ËšÚ—IàòöŽ•£E8mCÂJÎZ%>‚ µð‹ 2â»ò Ò•†äÞìǰЭ Œ+Ƭ=ÍkAVé²ÕŽöÇò¤¢˜‚Š( «Ëeo5ärF ð¿pÁ«PŒ†tËQë}þ†+[Oµû”vÁ·Èyª'é°äþ•þ†+b€ (¢€ (¢€ (¢€?ÿÙfestival/doc/festival_tiny.jpg0000644000175000017500000000425707262605651015147 0ustar awbawbÿØÿàJFIFÿþICREATOR: XV Version 3.10a Rev: 12/29/94 Quality = 60, Smoothing = 15 ÿÛC    ' .)10.)-,3:J>36F7,-@WAFLNRSR2>ZaZP`JQROÿÛC&&O5-5OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOÿÀ2¤"ÿÄ ÿĵ}!1AQa"q2‘¡#B±ÁRÑð$3br‚ %&'()*456789:CDEFGHIJSTUVWXYZcdefghijstuvwxyzƒ„…†‡ˆ‰Š’“”•–—˜™š¢£¤¥¦§¨©ª²³´µ¶·¸¹ºÂÃÄÅÆÇÈÉÊÒÓÔÕÖ×ØÙÚáâãäåæçèéêñòóôõö÷øùúÿÄ ÿĵw!1AQaq"2B‘¡±Á #3RðbrÑ $4á%ñ&'()*56789:CDEFGHIJSTUVWXYZcdefghijstuvwxyz‚ƒ„…†‡ˆ‰Š’“”•–—˜™š¢£¤¥¦§¨©ª²³´µ¶·¸¹ºÂÃÄÅÆÇÈÉÊÒÓÔÕÖ×ØÙÚâãäåæçèéêòóôõö÷øùúÿÚ ?ôÚ(¢€Š( Š*ž«q-¥“O †(A ÷éÅs;"e%äË”W7ÿ Rɵlû5g_ø®ñÔ­¬IûGæ5¿ÕjõG+ÇÐ]N²úúÛO¶iî¥TEõê}…yŽ¡âÝCíwóÛÈéëµ?pv#ßΫj77²ù—S<­êÇ¥fÈ~Âɧ¹’ÆsM4´G£x_Å°jPÇÓª\…äýãê+«s_?[;Dì¿Ý8­ÛOjÖ {ÇÚ?Àaú×›í\_,¤x%RÒ›ùËI^^¿µE\=µ»\ýj½§Œ5KÝzÚ{™¶Ãu Ó¿­Z¨žÇ,°³Šmž³EV°¸76©)xUšÐæ (¢€ (¢€ (¢€¨ßk:f"Ç}}ýۮɻclå°qõ¯8º{»OF—VH5½JèÄLñ†fËuç<€(Ñ`ž+ˆ„°H’FÝA©*†¥Zèö+mi à¹›š¿@GqÍÄã*ÀƒP.§fÚ›i«6ë¤O1ò¯¹éF¨Á©E$–ÛöÇ#FK)\‘×íBÐM]Xá.¡k[™!“‚‡…f­Ð¸b"…ÙAûØâ»OéÅÐ^B¹d0õÅjW†ÆÐyI*¯©5êÆ«©ÓÛsçç‡:®W¾ÄwP]ªvæªËUDÛÉõ­r„¢îv9ÅgÊc2ÁË vª8ß™™Ò­5±•fmúeºAFjUg dsÅG3©ÊW±Å:G õ¯ŸÄ?Þ3ïr襆Nú´M5ŒÑÚý “8æ¥Ð­Ìú‚:°žÒ¶B'Ðf»oi†[‘3/ UM]ÝbfãO•»³Ñ,"òlãOAViÀv¢·<Áh¬i¯E¿ŠV¦ÙÚä)Éd–E´„<Ã0A—þ´çº~¦ÚN“w⯺ž¦ßº·!yá GÖœ^&Ô}FÅn-e¹ŠÞ3nʼ4„|ï9®¯û#Mò`¶ò•·òÆ3늙,­c¸3¥¼K1]¥Â€qéšá®ü]q>‰=÷Ÿ¼q·“„÷§î®kœº"ç^WfA œjÒᲫ+@?_å^­ý“¦ìtû¶×mÌ<±‚}M6MNwi” #\ ®pw2«ITVêyÏ“4²n7L#=)²C ¼Ž£€ $õ5Ój^¸W-dñF ÞÜ̦òL =À®¹b)üI;žtpuŸ¹&¹|?´b÷LË’ÌI=ô¸Sø²Õ诂atZ ’Î;Ö$>½iÀu়©BnZŸQN½SVºhàY.ž`B1lñ]ÿ‚n#ƒPÓd´Ócó uò„ᄇÁàzWeTîìEÍÝÆý¦ÖBøÇÞÊ‘ýi’y/‰5 rç[ÿI²ò'â6~Tg Ï§Óx?Rñ8Ò"4´žÝAÈò„8ÏLVõÍŒw~-;ãF GpG^_óÅt ªŠ*Ž¨͵ _ÆÛ 6OÌp¡²Qù€®M¼ñX­æ™©#“2®81ï“]> 2)h9nµ<újŸá¸úQZ4PQE´QEQEQE%Q@ ô¢ŠZ(¢€ (¢€)¨ÛØ0Ÿø«”Q@Q@Q@ÿÙfestival/testsuite/Makefile0000644000175000017500000000547611363106376014512 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1996,1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### TOP=.. DIRNAME=testsuite ALL_DIRS = data correct TEST_SCRIPTS = fest text voice modes parse scherr TEST_MODULES = # not sure we'll need these SCRIPTS = $(TEST_SCRIPTS:%=%.sh) testinstall.sh FILES = Makefile \ $(SCRIPTS) ALL = include $(TOP)/config/test_make_rules test: tmpdir testinstall testinstall: @ sh testinstall.sh $(TOP) && $(MAKE) --no-print-directory tests tmpdir: @ rm -rf tmp @ mkdir tmp tests: test_modules test_scripts festival/testsuite/fest.sh0000644000175000017500000000556607304260360014341 0ustar awbawb#!/bin/sh ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## Basic test of synthesis FESTIVAL=$TOP/bin/festival HOME=$TOP/testsuite/data export HOME test_basic () { echo "basic " >&2 rm -f tmp/fest2.utt tmp/fest3.utt $FESTIVAL -b data/fest1.scm || exit 1 diff tmp/fest2.utt tmp/fest3.utt } test_utt () { echo "utt feats " >&2 $FESTIVAL -b data/utt1.scm || exit 1 } test_info () { echo info and help >&2 $FESTIVAL -h } echo >$OUTPUT test_basic 2>&1 >> $OUTPUT test_utt 2>&1 >> $OUTPUT test_info 2>&1 >> $OUTPUT exit 0 festival/testsuite/text.sh0000644000175000017500000000515307262605651014365 0ustar awbawb#!/bin/sh ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## Basic test of synthesis FESTIVAL=$TOP/bin/festival HOME=$TOP/testsuite/data export HOME test_text () { echo "text " >&2 $FESTIVAL -b data/fest2.scm || exit 1 } echo >$OUTPUT test_text 2>&1 >> $OUTPUT exit 0 festival/testsuite/voice.sh0000644000175000017500000000553711363106376014511 0ustar awbawb#!/bin/sh ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## Basic test voices FESTIVAL=$TOP/bin/festival HOME=$TOP/testsuite/data export HOME do_voices () { echo "multi-voices " >&2 $FESTIVAL -b data/voices.scm || exit 1 for i in rab kal slt do if cmp tmp/${i}1.wav tmp/${i}2.wav then echo $i voice: pass else echo $i voice: fail fi done # CG voices have some randomness in them so they wont be the same } echo >$OUTPUT do_voices 2>&1 >> $OUTPUT exit 0 festival/testsuite/modes.sh0000644000175000017500000000515007262605651014505 0ustar awbawb#!/bin/sh ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## Test text modes FESTIVAL=$TOP/bin/festival HOME=$TOP/testsuite/data export HOME do_modes () { echo "text modes " >&2 $FESTIVAL -b data/modes.scm || exit 1 } echo >$OUTPUT do_modes 2>&1 >> $OUTPUT exit 0 festival/testsuite/parse.sh0000644000175000017500000000523407262605651014513 0ustar awbawb#!/bin/sh ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1998 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## Test text modes FESTIVAL=$TOP/bin/festival HOME=$TOP/testsuite/data export HOME do_parse () { echo "Probabilistic grammars are easy to use, but difficult to train." | $TOP/examples/scfg_parse_text || exit 1 } echo >$OUTPUT do_parse 2>&1 >> $OUTPUT exit 0 festival/testsuite/scherr.sh0000644000175000017500000000537107262605651014671 0ustar awbawb#!/bin/sh ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## Basic test voices FESTIVAL=$TOP/bin/festival HOME=$TOP/testsuite/data export HOME do_test () { echo "checking error handling in batch mode " >&2 $FESTIVAL -b data/scherr.scm echo "checking error handling in interactive mode " >&2 echo "(load \"data/scherr.scm\")" | $FESTIVAL --interactive } echo >$OUTPUT do_test >> $OUTPUT 2>&1 exit 0 festival/testsuite/testinstall.sh0000644000175000017500000000515207304260166015741 0ustar awbawb#!/bin/sh ########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 2001 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## ## test overall installtion (and not continue if there are problems FESTIVAL=$1/bin/festival HOME=$1/testsuite/data export HOME echo "(quit)" | $FESTIVAL -b || exit 1 $FESTIVAL -b data/testinstall.scm || exit 1 exit 0 festival/testsuite/data/Makefile0000644000175000017500000000541310067612313015403 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## Data files used in test ## ########################################################################### TOP=../.. DIRNAME=testsuite/data TESTFILES = fest1.scm fest2.scm utt1.scm dump1.scm voices.scm modes.scm \ scherr.scm testinstall.scm htstest.scm FILES = Makefile .festivalrc $(TESTFILES) include $(TOP)/config/common_make_rules festival/testsuite/data/.festivalrc0000644000175000017500000000024107304555116016107 0ustar awbawb;;; ;;; A festivalrc to ensure the tests can be run in the same environment ;;; (if (boundp 'voice_rab_diphone) (set! voice_default 'voice_rab_diphone)) festival/testsuite/data/fest1.scm0000644000175000017500000000664007304260360015475 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Simple test of a single utterance (define (remove_ids utt) "(remove_ids utt) To all diffs on saved files this removed the _id feature from every item in utt so other utts will look the same as this if regenerated." (mapcar (lambda (r) (mapcar (lambda (i) (item.remove_feature i "id")) (utt.relation.items utt r))) (utt.relationnames utt)) utt) (set! utt1 (Utterance Text "On May 5 1985, around 1985 people joined Mr. Black's project.")) (utt.synth utt1) (utt.save.words utt1 "-") (print (utt.features utt1 'Word '(name R:Token.parent.name R:SylStructure.daughtern.name))) ;; tidy the utt up so it'll look the same on resynthesis (remove_ids utt1) (utt.set_feat utt1 "max_id" 0) (utt.set_feat utt1 "filename" 0) (utt.save utt1 "tmp/fest2.utt") (utt.save.segs utt1 "-") ;;; Test Utterance i/o (set! utt2 (utt.load nil "tmp/fest2.utt")) (utt.synth utt2) (remove_ids utt2) (utt.set_feat utt2 "max_id" 0) (utt.set_feat utt2 "filename" 0) (utt.save utt2 "tmp/fest3.utt") festival/testsuite/data/fest2.scm0000644000175000017500000001226307262605651015505 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Some specific tokens etc that might cause problems (require 'festtest) ;;; Test the tokenization (test_words "e.g. 12,000 pounds") (test_words "It costs $12 million") (test_words "Prussia's influence (1864-87) will be discussed at a conference May 2-10. Call (203) 450-3343, ($1.43 a minute) for details.") (test_words "23/06/97 at 03:45") (test_words "During the 1950's and 60s, 1.45% took AB123.") ;;; Money, money, money, (must be funny ...) (test_words "$10, $10,000, $1.00, $1.23, $1.03, $2.56.") (test_words "HK$100 million, \\10,000, Y1.2345, £1.23, M$1.03, C$2.56.") (test_words "A$1.25, £650.00, C$1.23 billion, Y10,000, #1.23.") ;;; Various special symbols (test_words "I think that is No 123.") (test_words "It was exactly 12:45:23.") (test_words "The date will be 3/3/04.") (test_words "Its on the 1st, and 2nd on 185th and Cornell.") ;;; Fractions (test_words "About 2/3 of the stocks increased by more than 1/16%, while the other 1/2 didn't.") ;;; Abbreviations (test_words "The U.S. government, EU and NASA are involved.") ;;; Roman numerals (test_words "Henry V: Part I Act II Scene XI: Mr X is I believe, V I Lenin, and not Charles I.") ;;; Saint Street Doctor Drive (test_words "Dr Taylor is at 12 High St. Edinburgh.") (test_words "Dr Taylor is at St Andrew's St, Edinburgh.") (test_words "Dr Taylor is at St Andrew's, St Albans.") (test_words "Dr Taylor is at Dr Johnson Dr, West Linton.") (test_words "Dr Taylor is with Dr Black Dr Caley and St Erasmus.") (test_words "Dr Taylor is at Dr Black Dr near the bus station.") ;; Test the phrase break mechanism (test_phrases "The man wanted to go for a drive in.") (test_phrases "The man wanted to go for a drive in the country.") (test_phrases "The man wanted to go for a drive-in the country.") (test_phrases "The man wanted to go for a drive--in the country.") (test_phrases "He gave the big boys' lunch in the park.") (test_phrases "He gave the `big boys' lunch in the park.") (test_phrases "That is it---unless you want more.") (test_phrases "That is it -- unless you want more.") ;; Some tests of utterance/punctuation boundary (test_segments "They called him Mr. Black though he preferred Alan.") (test_segments "They called him Mr. Black was the colour of his beard.") (test_segments "(They called him Mr.) Black was the colour of his beard.") (test_segments "The U.S. Secretary didn't arrive in time.") (test_segments "My cat who lives in Edinburgh has nine lives.") ;;; This was showed up different durations on different platforms ;;; sees ok now. Problem was lexicon was in different order ;;; (i.e. qsort did different things on different platforms) (test_segments "Prussia's influence (1864-87) will be discussed at a conference May 2-10. Call (203) 450-3343, ($1.43 a minute) for details.") festival/testsuite/data/utt1.scm0000644000175000017500000000675607262605651015371 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Walk about an utterance using features, relations etc. (set! utt1 (Utterance Text "hello, this example has three syllables.")) (utt.synth utt1) (format t "num syllables: %d\n" (length (utt.features utt1 'Syllable '(name)))) (set! segs (utt.relation.items utt1 'Segment)) (set! s1 (item.relation.parent (nth 2 segs) 'SylStructure)) (set! s2 (item.relation.parent (nth 3 segs) 'SylStructure)) (set! s3 (item.relation.parent (nth 4 segs) 'SylStructure)) (if (equal? s1 s2) (print "syls match: ok") (print "syls don't match: error")) (if (not (equal? s2 s3)) (print "syls don't match: ok") (print "syls match: error")) (print (list (item.feat s3 'name) (item.feat s3 'R:SylStructure.daughter1.name) (item.feat s3 'R:Syllable.n.R:SylStructure.daughter1.name) (item.feat s3 'R:Syllable.p.R:SylStructure.daughter1.name) (item.feat s3 'R:SylStructure.parent.name) (item.feat s3 'R:SylStructure.parent.daughter1.daughtern.name) (item.feat s3 'R:SylStructure.parent.R:Token.parent.whitespace) (item.feat s3 'R:SylStructure.parent.R:Word.n.R:Token.parent.p.punc) )) festival/testsuite/data/dump1.scm0000644000175000017500000000553307262605651015512 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Simple test of a single utterance (set! utt1 (Utterance Text "\"This test,\" she said, \"Should include punctuation; and interesting symbols like, 555-3434 ; 56&67, 4/7 etc.\"")) (utt.save utt1 "tmp/utt1.utt") (utt.load utt1 "tmp/utt1.utt") (utt.synth utt1) (utt.save.segs utt1 "tmp/savesegs1.segs") (utt.save utt1 "tmp/utt2.utt") (set! utt2 (utt.load nil "tmp/utt2.utt")) (utt.synth utt2) (utt.save.segs utt2 "tmp/savesegs2.segs") festival/testsuite/data/voices.scm0000644000175000017500000000746512437426042015755 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Test some voice switching (voice_rab_diphone) (set! utt1 (Utterance Text "The date was July 24th 1997.")) (utt.synth utt1) (utt.save.wave utt1 "tmp/rab1.wav") (voice_kal_diphone) (set! utt1 (Utterance Text "The date was July 24th 1997.")) (utt.synth utt1) (utt.save.wave utt1 "tmp/kal1.wav") (voice_cmu_us_rms_cg) (set! utt1 (Utterance Text "The date was July 24th 1997.")) (utt.synth utt1) (utt.save.wave utt1 "tmp/rms1.wav") (voice_cmu_us_awb_cg) (set! utt1 (Utterance Text "The date was July 24th 1997.")) (utt.synth utt1) (utt.save.wave utt1 "tmp/awb1.wav") (voice_cmu_us_slt_cg) (set! utt1 (Utterance Text "The date was July 24th 1997.")) (utt.synth utt1) (utt.save.wave utt1 "tmp/slt1.wav") ;; Check if voices interfere with each other or not (voice_rab_diphone) (set! utt1 (Utterance Text "The date was July 24th 1997.")) (utt.synth utt1) (utt.save.wave utt1 "tmp/rab2.wav") (voice_kal_diphone) (set! utt1 (Utterance Text "The date was July 24th 1997.")) (utt.synth utt1) (utt.save.wave utt1 "tmp/kal2.wav") (voice_cmu_us_rms_cg) (set! utt1 (Utterance Text "The date was July 24th 1997.")) (utt.synth utt1) (utt.save.wave utt1 "tmp/rms2.wav") (voice_cmu_us_awb_cg) (set! utt1 (Utterance Text "The date was July 24th 1997.")) (utt.synth utt1) (utt.save.wave utt1 "tmp/awb2.wav") (voice_cmu_us_slt_cg) (set! utt1 (Utterance Text "The date was July 24th 1997.")) (utt.synth utt1) (utt.save.wave utt1 "tmp/slt2.wav") festival/testsuite/data/modes.scm0000644000175000017500000000605210070124573015556 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Some test of various text modes (define (output_words utt) (utt.save.words utt "-") utt) (set! tts_hooks (list utt.synth output_words)) (gc-status nil) (format t "\n\nSABLE mode\n") (unwind-protect (tts (string-append libdir "/../examples/example2.sable") nil)) (format t "\n\nOGI's mark up mode\n") (unwind-protect (tts (string-append libdir "/../examples/ex1.ogi") 'ogimarkup)) (format t "\n\nAn email mode\n") (unwind-protect (tts (string-append libdir "/../examples/ex1.email") nil)) (voice_kal_diphone) (format t "\n\nA singing mode\n") (unwind-protect (tts (string-append libdir "/../examples/songs/doremi.xml") 'singing)) festival/testsuite/data/scherr.scm0000644000175000017500000000527507262605651015755 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 1996,1997 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Test catching of errors (define (subfunc) (load "Makefile") ;; shouldn't be a valid Scheme file ) (unwind-protect (subfunc) (begin (format t "caught the error\n") (printtt) ;; will give another error )) (format t "after the error\n") ;; shouldn't get here festival/testsuite/data/testinstall.scm0000644000175000017500000000602011363106377017020 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;; ;;; Centre for Speech Technology Research ;; ;;; University of Edinburgh, UK ;; ;;; Copyright (c) 2001 ;; ;;; All Rights Reserved. ;; ;;; ;; ;;; Permission is hereby granted, free of charge, to use and distribute ;; ;;; this software and its documentation without restriction, including ;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;; ;;; permit persons to whom this work is furnished to do so, subject to ;; ;;; the following conditions: ;; ;;; 1. The code must retain the above copyright notice, this list of ;; ;;; conditions and the following disclaimer. ;; ;;; 2. Any modifications must be clearly marked as such. ;; ;;; 3. Original authors' names are not deleted. ;; ;;; 4. The authors' names are not used to endorse or promote products ;; ;;; derived from this software without specific prior written ;; ;;; permission. ;; ;;; ;; ;;; THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;; ;;; SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;; ;;; THIS SOFTWARE. ;; ;;; ;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; See if the voices are installed (define (check_tests_will_work) "(check_tests_will_work) These tests *require* ked, don and rab to be installed." (unwind-protect (begin (voice_kal_diphone) (SynthText "hello world") (voice_rab_diphone) (SynthText "hello world") ) (begin (format stderr "\n") (format stderr "The festival tests require the kal and rab diphone voices to be\n") (format stderr "installed. Festival may work without that diphone set, but the results of\n") (format stderr "these tests aren't relevant.\n") (exit -1)))) (check_tests_will_work) festival/testsuite/data/htstest.scm0000644000175000017500000000657010067612313016152 0ustar awbawb;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; ;;; ;;; Language Technologies Institute ;;; ;;; Carnegie Mellon University ;;; ;;; Copyright (c) 2004 ;;; ;;; All Rights Reserved. ;;; ;;; ;;; ;;; Permission is hereby granted, free of charge, to use and distribute ;;; ;;; this software and its documentation without restriction, including ;;; ;;; without limitation the rights to use, copy, modify, merge, publish, ;;; ;;; distribute, sublicense, and/or sell copies of this work, and to ;;; ;;; permit persons to whom this work is furnished to do so, subject to ;;; ;;; the following conditions: ;;; ;;; 1. The code must retain the above copyright notice, this list of ;;; ;;; conditions and the following disclaimer. ;;; ;;; 2. Any modifications must be clearly marked as such. ;;; ;;; 3. Original authors' names are not deleted. ;;; ;;; 4. The authors' names are not used to endorse or promote products ;;; ;;; derived from this software without specific prior written ;;; ;;; permission. ;;; ;;; ;;; ;;; CARNEGIE MELLON UNIVERSITY AND THE CONTRIBUTORS TO THIS WORK ;;; ;;; DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ;;; ;;; ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ;;; ;;; SHALL CARNEGIE MELLON UNIVERSITY NOR THE CONTRIBUTORS BE LIABLE ;;; ;;; FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ;;; ;;; WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ;;; ;;; AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ;;; ;;; ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ;;; ;;; THIS SOFTWARE. ;;; ;;; ;;; ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; ;;; Some test the basic arctic/hts voices (format t "cmu_us_awb_arctic_hts\n") (voice_cmu_us_awb_arctic_hts) (set! utt1 (SynthText "A whole joy was reaping, but they've gone south, you should fetch azure mike.")) (utt.save.segs utt1 "-") (format t "%l\n" (wave.info (utt.wave utt1))) (format t "cmu_us_bdl_arctic_hts\n") (voice_cmu_us_bdl_arctic_hts) (set! utt1 (SynthText "A whole joy was reaping, but they've gone south, you should fetch azure mike.")) (utt.save.segs utt1 "-") (format t "%l\n" (wave.info (utt.wave utt1))) (format t "cmu_us_jmk_arctic_hts\n") (voice_cmu_us_jmk_arctic_hts) (set! utt1 (SynthText "A whole joy was reaping, but they've gone south, you should fetch azure mike.")) (utt.save.segs utt1 "-") (format t "%l\n" (wave.info (utt.wave utt1))) (format t "cmu_us_slt_arctic_hts\n") (voice_cmu_us_slt_arctic_hts) (set! utt1 (SynthText "A whole joy was reaping, but they've gone south, you should fetch azure mike.")) (utt.save.segs utt1 "-") (format t "%l\n" (wave.info (utt.wave utt1))) festival/testsuite/correct/Makefile0000644000175000017500000000542511363106376016145 0ustar awbawb########################################################################### ## ## ## Centre for Speech Technology Research ## ## University of Edinburgh, UK ## ## Copyright (c) 1997 ## ## All Rights Reserved. ## ## ## ## Permission is hereby granted, free of charge, to use and distribute ## ## this software and its documentation without restriction, including ## ## without limitation the rights to use, copy, modify, merge, publish, ## ## distribute, sublicense, and/or sell copies of this work, and to ## ## permit persons to whom this work is furnished to do so, subject to ## ## the following conditions: ## ## 1. The code must retain the above copyright notice, this list of ## ## conditions and the following disclaimer. ## ## 2. Any modifications must be clearly marked as such. ## ## 3. Original authors' names are not deleted. ## ## 4. The authors' names are not used to endorse or promote products ## ## derived from this software without specific prior written ## ## permission. ## ## ## ## THE UNIVERSITY OF EDINBURGH AND THE CONTRIBUTORS TO THIS WORK ## ## DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ## ## ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT ## ## SHALL THE UNIVERSITY OF EDINBURGH NOR THE CONTRIBUTORS BE LIABLE ## ## FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES ## ## WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN ## ## AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ## ## ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF ## ## THIS SOFTWARE. ## ## ## ########################################################################### ## Correct answers to tests ## ########################################################################### TOP=../.. DIRNAME=testsuite/correct OUTFILES = fest_script.out text_script.out voice_script.out modes_script.out \ parse_script.out scherr_script.out FILES= Makefile $(OUTFILES) include $(TOP)/config/common_make_rules list: echo $(FILES) festival/testsuite/correct/fest_script.out0000644000175000017500000000567612437425356017600 0ustar awbawb # 0.3557 100 On 0.5594 100 May 1.0018 100 fifth 1.4512 100 nineteen 1.7458 100 eighty 2.1355 100 five 2.6985 100 around 3.0873 100 nineteen 3.3696 100 eighty 3.6559 100 five 4.1902 100 people 4.7505 100 joined 5.0889 100 Mr 5.5076 100 Black 0.0000 100 's 6.2561 100 project (("On" "On" "syl") ("May" "May" "syl") ("fifth" "5" "syl") ("nineteen" "1985" "syl") ("eighty" "1985" "syl") ("five" "1985" "syl") ("around" "around" "syl") ("nineteen" "1985" "syl") ("eighty" "1985" "syl") ("five" "1985" "syl") ("people" "people" "syl") ("joined" "joined" "syl") ("Mr" "Mr" "syl") ("Black" "Black's" "syl") ("'s" "Black's" 0) ("project" "project" "syl")) # 0.2100 100 # 0.3003 100 o 0.3557 100 n 0.4170 100 m 0.5594 100 ei 0.6965 100 f 0.8121 100 i 0.9129 100 f 1.0018 100 th 1.0717 100 n 1.1847 100 ai 1.2470 100 n 1.3192 100 t 1.4005 100 ii 1.4512 100 n 1.5869 100 ei 1.6648 100 t 1.7458 100 ii 1.8676 100 f 2.0689 100 ai 2.1355 100 v 2.3455 100 # 2.3945 100 @ 2.4588 100 r 2.6183 100 au 2.6730 100 n 2.6985 100 d 2.7561 100 n 2.8594 100 ai 2.9039 100 n 2.9655 100 t 3.0359 100 ii 3.0873 100 n 3.2023 100 ei 3.2887 100 t 3.3696 100 ii 3.4757 100 f 3.6081 100 ai 3.6559 100 v 3.7813 100 p 3.8705 100 ii 4.0170 100 p 4.0970 100 @ 4.1902 100 l 4.4002 100 # 4.4947 100 jh 4.6593 100 oi 4.7115 100 n 4.7505 100 d 4.8165 100 m 4.8785 100 i 4.9613 100 s 5.0367 100 t 5.0889 100 @ 5.1635 100 b 5.2019 100 l 5.3340 100 a 5.4238 100 k 5.5076 100 s 5.6497 100 p 5.6830 100 r 5.7822 100 o 5.8880 100 jh 5.9953 100 e 6.1471 100 k 6.2561 100 t 6.6761 100 # num syllables: 9 "syls match: ok" "syls don't match: ok" ("syl" "ou" "dh" "h" "hello" "l" "" ",") Usage: festival Usage: festival ... In evaluation mode "filenames" starting with ( are evaluated inline Festival Speech Synthesis System: 2.4:beta December 2014 -q Load no default setup files --libdir Set library directory pathname -b Run in batch mode (no interaction) --batch Run in batch mode (no interaction) --tts Synthesize text in files as speech no files means read from stdin (implies no interaction by default) -i Run in interactive mode (default) --interactive Run in interactive mode (default) --pipe Run in pipe mode, reading commands from stdin, but no prompt or return values are printed (default if stdin not a tty) --language Run in named language, default is english, spanish and welsh are available --server Run in server mode waiting for clients of server_port (1314) --script Used in #! scripts, runs in batch mode on file and passes all other args to Scheme --heap {10000000} Set size of Lisp heap, should not normally need to be changed from its default -v Display version number and exit --version Display version number and exit festival/testsuite/correct/text_script.out0000644000175000017500000001164007262605651017605 0ustar awbawb Word test: e.g. 12,000 pounds e g twelve thousand pounds Word test: It costs $12 million It costs twelve million dollars Word test: Prussia's influence (1864-87) will be discussed at a conference May 2-10. Call (203) 450-3343, ($1.43 a minute) for details. Prussia 's influence eighteen sixty four to eighty seven will be discussed at a conference May second to tenth Call two zero three four five zero three three four three one dollar forty three a minute for details Word test: 23/06/97 at 03:45 twenty three zero six ninety seven at three forty five Word test: During the 1950's and 60s, 1.45% took AB123. During the nineteen fifty 's and sixty 's one point four five percent took A B one two three Word test: $10, $10,000, $1.00, $1.23, $1.03, $2.56. ten dollars ten thousand dollars one dollar one dollar twenty three one dollar three two dollars fifty six Word test: HK$100 million, \10,000, Y1.2345, £1.23, M$1.03, C$2.56. one hundred million Hong Kong dollars ten thousand yen one point two three four five yen one pound twenty three one M dollar three two Canadian dollars fifty six Word test: A$1.25, £650.00, C$1.23 billion, Y10,000, #1.23. one Australian dollar twenty five six hundred and fifty pounds one point two three billion Canadian dollars ten thousand yen one pound twenty three Word test: I think that is No 123. I think that is number one hundred and twenty three Word test: It was exactly 12:45:23. It was exactly twelve hours forty five minutes and twenty three seconds Word test: The date will be 3/3/04. The date will be three three o four Word test: Its on the 1st, and 2nd on 185th and Cornell. Its on the first and second on one hundred and eighty fifth and Cornell Word test: About 2/3 of the stocks increased by more than 1/16%, while the other 1/2 didn't. About two third 's of the stocks increased by more than one sixteenth percent while the other half didn't Word test: The U.S. government, EU and NASA are involved. The U S government E U and NASA are involved Word test: Henry V: Part I Act II Scene XI: Mr X is I believe, V I Lenin, and not Charles I. Henry the fifth Part one Act two Scene eleven Mr X is I believe V I Lenin and not Charles the first Word test: Dr Taylor is at 12 High St. Edinburgh. doctor Taylor is at twelve High street Edinburgh Word test: Dr Taylor is at St Andrew's St, Edinburgh. doctor Taylor is at saint Andrew 's street Edinburgh Word test: Dr Taylor is at St Andrew's, St Albans. doctor Taylor is at saint Andrew 's saint Albans Word test: Dr Taylor is at Dr Johnson Dr, West Linton. doctor Taylor is at doctor Johnson drive West Linton Word test: Dr Taylor is with Dr Black Dr Caley and St Erasmus. doctor Taylor is with doctor Black doctor Caley and saint Erasmus Word test: Dr Taylor is at Dr Black Dr near the bus station. doctor Taylor is at doctor Black drive near the bus station Phrase test: The man wanted to go for a drive in. The man wanted to go B for a drive in BB Phrase test: The man wanted to go for a drive in the country. The man wanted to go for a drive B in the country BB Phrase test: The man wanted to go for a drive-in the country. The man wanted to go B for a drive-in the country BB Phrase test: The man wanted to go for a drive--in the country. The man wanted to go for a drive B in the country BB Phrase test: He gave the big boys' lunch in the park. He gave the big boys lunch B in the park BB Phrase test: He gave the `big boys' lunch in the park. He gave the big boys B lunch in the park BB Phrase test: That is it---unless you want more. That is it B unless you want more BB Phrase test: That is it -- unless you want more. That is it B unless you want more BB Segment test: They called him Mr. Black though he preferred Alan. # dh ei k oo l d h i m m i s t @ b l a k # dh ou h ii p r i f @@ d a l @ n # Segment test: They called him Mr. Black was the colour of his beard. # dh ei k oo l d h i m m i s t @ # b l a k w o z dh @ k uh l @ r o v h i z b i@ d # Segment test: (They called him Mr.) Black was the colour of his beard. # dh ei k oo l d h i m m i s t @ # b l a k w o z dh @ k uh l @ r o v h i z b i@ d # Segment test: The U.S. Secretary didn't arrive in time. # dh @ y uu e s s e k r @ t r ii d i d @ n t # @ r ai v i n t ai m # Segment test: My cat who lives in Edinburgh has nine lives. # m ai k a t # h uu l i v z i n e d i n b r @ h a z n ai n l ai v z # Segment test: Prussia's influence (1864-87) will be discussed at a conference May 2-10. Call (203) 450-3343, ($1.43 a minute) for details. # p r uh s i@ z i n f l u@ n s # ei t ii n s i k s t ii f oo t uu ei t ii s e v @ n # w i l b ii d i s k uh s t a t @ k o n f @ r @ n s # m ei s e k @ n d t uu t e n th # k oo l t uu z i@ r ou th r ii # f oo f ai v z i@ r ou # th r ii th r ii f oo th r ii # w uh n d o l @ f oo t ii th r ii @ m i n i t # f oo d ii t ei l z # festival/testsuite/correct/voice_script.out0000644000175000017500000000006111363106377017717 0ustar awbawb rab voice: pass kal voice: pass slt voice: pass festival/testsuite/correct/modes_script.out0000644000175000017500000000725112437425440017727 0ustar awbawb SABLE mode # 0.9927 100 Homographs 1.1053 100 are 1.3276 100 words 1.5219 100 that 1.6625 100 are 1.9794 100 written 2.0507 100 the 2.5113 100 same 2.9623 100 but 3.1767 100 have 3.6562 100 different 4.7686 100 pronunciations 5.3067 100 such 5.5096 100 as 5.8947 100 lyves 6.0943 100 and 6.6187 100 lives 7.2384 100 You 7.5743 100 say 7.7106 100 e 7.8591 100 thur 8.3142 100 while 8.5008 100 I 8.8347 100 say 8.9573 100 i 9.1226 100 thur # 0.2456 100 We 0.3975 100 can 0.5532 100 say 0.7675 100 things 1.0376 100 fast # 0.4145 100 and 1.6015 100 slowly # 0.3967 100 And 0.5812 100 then 0.7590 100 at 1.2333 100 normal 1.7070 100 speed OGI's mark up mode # 0.3960 100 This 0.5348 100 is 0.6286 100 an 1.3301 100 example # # 0.6725 100 This 0.9124 100 is 0.9781 100 a 1.4984 100 slow 2.3918 100 example # # 1.0347 100 This 1.4041 100 is 1.5196 100 a 2.2010 100 very 2.9970 100 slow 4.4190 100 example # 0.2256 100 a 0.6647 100 normal 1.0413 100 one # 0.3334 100 and 0.3990 100 a 0.8220 100 fast 1.2018 100 talking 1.7704 100 example # 0.2693 100 Maybe 0.4129 100 this 0.5545 100 one 0.6439 100 is 0.7683 100 too 1.0700 100 fast # # 0.3977 100 My 0.7424 100 name 0.9284 100 is 1.4371 100 Mike # 0.3075 100 m 0.4917 100 i 0.7783 100 k 1.0070 100 e # 0.3807 100 My 1.0291 100 telephone 1.4310 100 number 1.7364 100 is # 0.5268 100 six 0.9183 100 five 1.4268 100 zero 1.7948 100 two 2.2058 100 seven 2.4521 100 eight 2.9536 100 seven # 0.4811 100 Mike 0.8075 100 here An email mode # 0.4325 100 From # 0.4879 100 Alan 0.8582 100 W 1.1638 100 Black 1.9989 100 awb 2.3649 100 at 3.0684 100 cstr 3.2789 100 dot 3.4261 100 ed 3.6272 100 dot 3.9380 100 ac 4.1695 100 dot 4.6229 100 uk # 0.8236 100 Subject # 0.7615 100 Example 0.9717 100 mail 1.4475 100 message # 0.4879 100 Alan 0.8582 100 W 1.1744 100 Black 1.4320 100 writes 1.5817 100 on 1.9377 100 twenty 2.3531 100 seventh 2.8842 100 November 3.5314 100 nineteen 3.9270 100 ninety 4.3941 100 six # 0.3288 100 I'm 0.6680 100 looking 0.8646 100 for 0.9079 100 a 1.2169 100 demo 1.5192 100 mail 2.0264 100 message 2.3382 100 for 2.9966 100 Festival 3.3299 100 but 3.6226 100 can't 3.8911 100 seem 3.9974 100 to 4.3597 100 find 4.5942 100 any 5.1401 100 suitable # 0.2946 100 It 0.5420 100 should 0.6891 100 at 1.0428 100 least 1.2159 100 have 1.4558 100 some 1.8934 100 quoted 2.5100 100 text 2.8172 100 and 2.9738 100 have 3.2205 100 some 3.7405 100 interesting 4.3659 100 tokens 4.7799 100 like 4.8230 100 a 5.0490 100 U 5.2006 100 R 5.4659 100 L 5.7574 100 or 6.1057 100 such 6.5152 100 like # 0.4687 100 Alan # 0.4171 100 Well 0.6177 100 I'm 0.8280 100 not 1.0981 100 sure 1.7706 100 exactly 2.1386 100 what 2.2396 100 you 2.4730 100 mean 2.6799 100 but 3.4537 100 awb 3.6200 100 at 4.0615 100 cogsci 4.2720 100 dot 4.4192 100 ed 4.6202 100 dot 4.9311 100 ac 5.1625 100 dot 5.6160 100 uk 6.0661 100 has 6.1574 100 an 6.6786 100 interesting 6.9187 100 home 7.3790 100 page 7.7348 100 at 8.0158 100 h 8.2150 100 t 8.3928 100 t 8.6405 100 p 0.0000 100 : 9.2899 100 slash 9.7375 100 slash 10.1158 100 w 10.5072 100 w 10.8986 100 w 11.1222 100 dot 11.8090 100 cstr 12.0195 100 dot 12.1667 100 ed 12.3678 100 dot 12.6924 100 ac 12.9295 100 dot 13.3850 100 uk 13.9456 100 slash 14.2057 100 ~ 14.9393 100 awb 15.3305 100 slash 15.5351 100 which 15.7482 100 might 15.9966 100 be 16.4165 100 what 16.5380 100 you're 16.8793 100 looking 17.0996 100 for # 0.5377 100 Alan # 0.3350 100 P 0.5774 100 S 0.7758 100 Will 0.9539 100 you 1.3255 100 attend 1.3857 100 the 1.9213 100 course # 0.3271 100 I 0.6470 100 hope 0.9680 100 so # 0.4741 100 bye 0.6674 100 for 1.0077 100 now A singing mode # 0.5000 100 doe 0.9600 100 ray 1.4500 100 me 1.9600 100 fah 2.4700 100 sew 2.9500 100 lah 3.4100 100 tee 4.0000 100 doe festival/testsuite/correct/parse_script.out0000644000175000017500000000076307262605651017737 0ustar awbawb (("NT00" ("NT14" ("NT06" ("NT01" ("jj" "Probabilistic")) ("NT07" ("nns" "grammars"))) ("NT17" ("NT04" ("vbp" "are")) ("NT08" ("NT16" ("jj" "easy")) ("NT09" ("NT15" ("to" "to")) ("NT08" ("NT04" ("vb" "use")) ("NT08" ("NT04" ("punc" ",")) ("NT09" ("NT15" ("cc" "but")) ("NT08" ("NT16" ("jj" "difficult")) ("NT09" ("NT15" ("to" "to")) ("NT08" ("vb" "train"))))))))))) ("NT13" ("punc" ".")))) festival/testsuite/correct/scherr_script.out0000644000175000017500000000242712437425440020106 0ustar awbawb checking error handling in batch mode SIOD ERROR: unbound variable : ########################################################################### closing a file left open: Makefile SIOD ERROR: unbound variable : printtt closing a file left open: data/scherr.scm caught the error checking error handling in interactive mode Festival Speech Synthesis System 2.4:beta December 2014 Copyright (C) University of Edinburgh, 1996-2010. All rights reserved. clunits: Copyright (C) University of Edinburgh and CMU 1997-2010 clustergen_engine: Copyright (C) Carnegie Mellon University 2005-2014 hts_engine: The HMM-Based Speech Synthesis Engine "hts_engine API" hts_engine API version 1.07 (http://hts-engine.sourceforge.net/) Copyright (C) The HMM-Based Speech Synthesis Engine "hts_engine API" Version 1.07 (http://hts-engine.sourceforge.net/) Copyright (C) 2001-2012 Nagoya Institute of Technology 2001-2008 Tokyo Institute of Technology All rights reserved. All rights reserved. For details type `(festival_warranty)' festival> SIOD ERROR: unbound variable : ########################################################################### closing a file left open: Makefile SIOD ERROR: unbound variable : printtt closing a file left open: data/scherr.scm caught the error festival> festival> festival/.time-stamp0000644000175000017500000000005212445315020013051 0ustar awbawbfestival 2.4 Sat Dec 20 10:47:28 EST 2014