ANNOUNCING THE RELEASE OF THE OGI SPEECH TOOLS (Version 1.0) ============================================================= A. What are the OGI Speech Tools ================================ The OGI Speech Tools are set of speech data manipulation tools developed at the Center for Spoken Language Understanding (CSLU) at the Oregon Graduate Institute of Science and Technology (Portland Oregon). The tools can be used to compute and display signal representations, label speech at different levels (e.g., phonetic, phonemic and word), train neural network classifiers, and display the output of classification or recognition algorithms time-aligned with the speech. A Software Capitalization grant from NSF has allowed to enhance, document and distribute the tools. Continued maintenance and improvement of the tools is supported by CSLU. These include 1. An X windows display tool (LYRE) for displaying data in a time synchronous fashion. These mainly constitute the following data representations: a. the speech signal b. the spectrogram c. the phoneme labels The data displayed may, however, also include any other type of time synchronous information, provided the data is stored in the required format. 2. A Neural Network (NOPT) training package, which uses Conjugate Gradient descent optimization. 3. An extensive set of C library routines (LIBNSPEECH) for the manipulation of speech data. These also include the following signal processing routines. a. PLP Analysis b. Rasta PLP Analysis c. Linear Predictive Coding d. Mel Cepstrum Coding e. Fast Fourier Transform 4. An extensive set of utilities for converting from files of different format. These include routines such as: a. Converting from ADC to NIST type headers b. Converting to mu-law encoded files c. Converting from mu-law encoded files back to NIST headered files d. Filtering the speech waveform. e. Converting to from binary to an ascii data stream. 5. A database utility (find_phone) to efficiently automate speech data related queries. This utility creates a database based on a listing of speech files and there corresponding location and label files (lola). It allows the user to specify a particular label or set of labels in a given context, display all occurrences of the label, and relabel the occurrences if desired. 6. A Vector-Quantizer based on the Linde Buzo and Gray (LBG) algorithm. 7. An set of PEARL Scripts which have been used mainly to automate the use of the OGI Speech Tools. 8. MAN Pages for all routines and programs developed, as well as a user manual in both in postscript and {\bf tex} format. B. How do I get the OGI Speech Tools ==================================== The OGI speech tools are available via anonymous ftp from speech.cse.ogi.edu and are located in /pub/tools Step by Step instructions ========================= First you need to create a directory (on your home machine for example) , where the speech tools are going to be located. Assume it is named --> speechtools % mkdir speechtools [enter] ~~~~~~~~~~~~~~~~~ % cd speechtools [enter] ~~~~~~~~~~~~~~~ % mkdir src [enter] ~~~~~~~~~~ % cd src [enter] ~~~~~~ % ftp speech.cse.ogi.edu [enter] ~~~~~~~~~~~~~~~~~~~~~~ Connected to lydia.cse.ogi.edu 220 lydia FTP server (SunOS 4.1) ready. % Name: (speech.cse.ogi.edu:loginname) anonymous [enter] ~~~~~~~~~ 331 Guest login ok, send ident password. Password: (your email address) [enter] ~~~~~~~~~~~~~~~~~~~~ 230 Guest login ok, access restrictions apply. ftp> cd pub/tools [enter] ~~~~~~~~~~~~ 250 CWD command successful. ftp> ls [enter] ~~ The following files should be listed INSTALL CHANGES ogitools.v1.0.tar.Z sun4.binaries.tar.Z sun4.libraries.tar.Z decmips.binaries.tar.Z decmips.libraries.tar.Z alpha.binaries.tar.Z alpha.libraries.tar.Z ftp> bin [enter] ~~~~~ 200 Type set to I. ftp> prompt [enter] ~~~~~~~ Interactive mode off. ftp> get ogitools.v1.0.tar.Z [enter] ~~~~~~~~~~~~~~~~~~~~~~~~ ftp> quit You now have the speech tools. The INSTALL file contains instructions for installing the speechtools. The CHANGES file contains a list of all changes made since the BETA release of the OGI Speech Tools. To get started quickly, ftp the related binaries files for your system. The OGI Speech Tools were written using ANSI style C. If you current compiler does not support ANSI C style, try getting hold of the GNU C (gcc) compiler. The configuration file can be changed to use the GCC compiler by changing the .conf files in the ~/speechtools/src/conf/ directory. There the variable CC='acc' for sun4, must be changed to CC='gcc'. C. WHAT NEXT ============ We at CSLU (OGI) would like to compile a mailing list of all speech tools users, so that we can inform our users community of any additions and changes to the speech tools. We would therefore like you, to drop your email address by simply mailing a message to tools@cse.ogi.edu This email address might also be used for requesting more information on installing the speech tools and for reporting any bugs detected. This would greatly be appreciated, since we are aiming to keep the speech tools both portable and robust. Any further suggestions and additions which you could add will also be appreciated. Johan Schalkwyk (tools@cse.ogi.edu) Center for Spoken Language Understanding Oregon Graduate Institute