?? readme
字號:
==========================================================================Users of version 0.31 and below, please note: As of 0.32, thedefault feature extractor really is Rasta-PLP, it was vanilla PLPbefore. You *have* to "rm ~/.ears/MRASTA-V/*" and train_ears again!The old behaviour can then be gotten by configuring FEATURE PLP and retraining.Sorry for the inconvenience. ralf========================================================================== --------------------------------------------------------- EARS (something like 'Easy Automatic Recognition of Speech...) ---------------------------------------------------------The EARS package is intended as a limited ready-to-use single word recognizer for Linux systems. However, its design already aims atbeing a host for all kinds of methods used in speech recognition (SR).The ALPHA versions of the package consist of the following programs: train_ears - to speak a list of words and train the recognizer listen - listens to your mic and, when a word is seen, shows the corresponding keyboard action (in beta versions, there will also be the 'ears' program itself which acts like 'listen' but additionally drives a shell. As of now, you can get similar functionality with either Karl J. Runge's launch.pl or Dan Chaney's act program. Both are in contrib/)The code for EARS is available under the GNU General Public License.See the file COPYING for details.You can get source code from http://home.pages.de/~ears/ ftp://sunsite.mirror/...apps/sound/speech/ ftp://svr-ftp.eng.cam.ac.uk/pub/comp.speech/recognitionThere is now also a mailinglist where you can tune in and get thelatest news about the package (thanks to Chris Blum). Send mail to majordomo@phil.uni-sb.de with subscribe ears-list in the body.To just send a mail to the list, use ears-list@phil.uni-sb.de .What you need to run binaries-----------------------------First, a medium fast machine (i486/66 and up) with FPU. Then, a soundcardwith the capability to record from a microphone. Third, a sound driver:1. Possibility: a Linux system with kernel sound support. It may be possible to use EARS on other machines running the Linux driver calledVoxware from Hannu Savolainen (mail hannu@voxware.pp.fi).2. Possibility: a Linux/Un*x system having the Open Sound System (OSS)driver which is a nice product BTW. Just change two lines in the Makefile. To get the driver, see http://www.4front-tech.com3. Possibility: a system having the Sun audio system. Read README.sunon that and change the Makefile accordingly.4. Possibility: another Un*x system having the AF-Tools. Readthe file README.af on that one. Support is incomplete since somethings changed and I don't have an AF machine to adapt the source.What you need to compile the package------------------------------------You need gcc-2.7.2 and libstdc++.a from libg++-2.7.1.4 or higher, and for the non-commandline version libncurses.a and libpanel.a (from ncurses-1.9.9e).Unfortunately, due to the flowing nature of gcc/libstdc++, I can only guarantee successful compiles for my own configuration which is, as ofnow, gcc-2.7.2-elf together with libg++-2.7.1.4.What you get------------A SR program you can play with and add your own source code to it.Isn't that nice? ;)What is implemented now-----------------------As of this alpha release, the feature extractor is: PLP and Rasta-PLP from the Mrasta package (version 2.2.1). Soon to come: Jah-Rasta.The implemented recognizers are- Dynamic Time Warp (DTW), thanks to Dr. Robinson's Cookbook.- an improved version of DTW called NDTW (this is the default).- non-recurrent neural nets on fixed-size sound patterns, like vanilla backpropagation (BP) and backprop with momentum term (BPMT), also known as conjugate gradient descent. Unusable.Install-------EARS will untar into a directory 'ears-<version>', for example $ tar xvfz ears-0.32.tar.gzuntars into ./ears-0.32/*Then cd to that dir and do a 'make'. Then, depending on your habits, get the usual quantity of stimulant. Now start train_ears, record some digits, run listen and see themrecognized. Then change the basename to alphabet and after recordingthose, start 'listen alphabet'. In ideal situations, recognition can be as accurate as 95%.New words are easily added by either editing a given .words file or creating a new one. Don't forget to train_ears after you increased the word list.ears creates a directory .ears in your home path, so if you want toget rid of ears completely, you have to delete this too.Understanding EARS------------------I have tried to outline the source in doc/implementation.txt.After grokking that, it should be easier for you to read the code, which, unfortunately, isn't as much commented as I would like.Also, you might read CHANGES to see how coding/designing went along.Improving EARS--------------Things that I think are to do are in doc/TODO --- YMMV.Apart from improvements of the data handling, EARS needs to have thefollowing things to make it useful for method comparison:- more feature extractors- a better endpoint detection- more, better and faster recognizers. I guess there are dozens of methods that could be implemented, for example HMMs, TDNNs, all kind of non-recurrent and recurrent neural nets, wavelets, zero- crossings, hybrid methods, etc.- a program for cross-validation that gives error rates for specific feature extractor/recognizer combinations.And, of course, we need the 'ears' program to drive a shell. Thislater has to have the ability to take back erroneous inputs, and allthings one needs to work with it, like showing of alternativeinterpretations, learning while using, and so on.Other packages--------------Israel Diamand & Nir Arzooni have built what appears to be a module forsending voice commands to fvwm using EARS. Check <http://tochna.technion.ac.il/projects/winter96/FvwmVoice/FvwmVoice.html>Acknowledgements----------------Without the following people, EARS wouldn't be possible:- Linus Torvalds- all people who made Linux to what it is- Hrynek Hermansky for his Rasta feature extraction- Tilo Schuerer for librecog and his patience- Tony Robinson for his cookbook and the archive in CambridgeThanks to all people who helped with the project:- Tilman Enss adapted Makefile to GNU standard; fixed bugs; wrote usage; made German message catalog.- Niels Thorwirth ported ears to AF platforms; fixed bugs; wrote documents; introduced config.sh.- Karl J. Runge contributed a perl script to launch programs.- Dan Chaney wrote a nice C-frontend for 'listen' that does even more.- Michael Ward added the sun sound interface, many suggestions.And last but not least:- the 4front-tech people for kindly giving me a free OSS license.Contributing------------Changes to the source code should be mailed as diffs to ralf@ark.franken.deI prefer unified diffs (use diff -u) as they are well readable.Please include the version number the diffs are made against!Many thanks in advance for your contribution,ralf@ark.franken.de (Ralf W. Stephan)Security--------You can mail me for MD5 checksums of the distributed packagesor check yours against the one on the webpage.-----BEGIN PGP PUBLIC KEY BLOCK-----Version: 2.6.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=q/s6-----END PGP PUBLIC KEY BLOCK----------BEGIN PGP PUBLIC KEY BLOCK-----Version: 2.6.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4tcg-----END PGP PUBLIC KEY BLOCK-----
?? 快捷鍵說明
復制代碼
Ctrl + C
搜索代碼
Ctrl + F
全屏模式
F11
切換主題
Ctrl + Shift + D
顯示快捷鍵
?
增大字號
Ctrl + =
減小字號
Ctrl + -