This is Sphinx 3.5 (s3.5), one of Carnegie Mellon University's open source large vocabulary, speaker-independent continuous speech recognition engine. Please see the LICENSE file for terms of use.
Over the course of development, Sphinx-3 branched into different versions. The version available here, often times referred to as "sphinx3.5", "s3.5", or "fast decoder", is a faster version of the original code. Higher speed has been achieved at the cost of less accuracy compared to the slow decoder.
THIS IS A RESEARCH SYSTEM. This is also an early release of a research system. We know the APIs and function names are likely to change, and that several tools need to be made available to make this all complete. With your help and contributions, this can progress in response to the needs and patches provided.
If you install the distribution in the default location, you can access the documentation here:
file://localhost/usr/local/share/sphinx3/doc/
There are couple of nice documents you could find in the ./doc directory. That includes
s3_codework.html : Code review document prepared by Ravi Mosur. s3_description.html : Code review document prepared by Ravi Mosur. s3_overview.html : Code review document prepared by Ravi Mosur. cmdhelp.txt : (Obsolete) Sphinx 3.3 command line arguments. s3.2.ppt : The powerpoint presentation prepared by Ravi Mosur. FAQ.html : Frequently asked question page prepared by Rita Singh sphinxman_manual.html : Manual for training prepared by Rita Singh. sphinxman_FAQ.html : FAQ for training prepared by Rita Singh. sphinxman_misc.html : Miscellaneous information for training prepared by Rita Singh. models.html : Description of the default broad cast news model
For up-to-date information, please see the web site at
A brief note of version: s3.5 decoder is an extension of s3.3 with several bug fixes and speed-up algorithm implmentation. s3.3 is a live-mode capable version of the s3.2 decoder. The decoder routine is stored in a library called libs3decoder.a. The decoder function takes blocks of speech samples and returns partial hypotheses for the blocks decoded during runtime. Examples of its use are given in main_live_example.c and main_live_pretend.c
This directory contains the scripts and instructions necessary for decoding speech input using the CMU Sphinx Recognizer.
This distribution is free software, see LICENSE for license.
This work was built over a large number of years at CMU by most of the people in the Sphinx Group. Some code goes back to 1986. The most recent work in tidying this up for release includes the following, listed alphabetically (at least these are the people who are most likely able to help you).
Installation Guide:
^^^^^^^^^^^^^^^^^^^
This sections contain installation guide for various platforms.
Linux/Unix Installation:
^^^^^^^^^^^^^^^^^^^^^^^^
This distribution now uses GNU autoconf to find out basic information
about your system, and should compile on most Unix and Unix-like
systems, and certainly on Linux. To build, simply run
./configure
make
(Or whatever you call GNU Make). This should configure everything automatically. After a successful compilation, you may test the system by running
make test
and then install it with
make install
This defaults to installing Sphinx-3 under /usr/local. You may customize it by running ./configure with an argument, as in
./configure --prefix=/my/own/installation/directory
Windows Installation:
^^^^^^^^^^^^^^^^^^^^^
To compile the sphinx 3.5 in Visual C++ 6.0,
1, unzip the file.
2, click programs.dsw
3, build the projects in the following order, libaudio, libdecoder, libutil.
4, build the live-mode decoder project, livedecode.
5, build the live-mode simulator project, livepretend.
6, build the batch-mode recognizer project, decode.
Steps 4-7 can be done easily by in the "File View" of VC6.0 by left-clicking the project and choose "Build". To run the executable, please remember to set the corresponding active project.
If you are using cygwin, the installation procedure is very similar to the Unix installation. However, there is no audio driver support in cygwin currently so one can only use the batch mode recognzier.
Alan W Black (awb@cs.cmu.edu)
Evandro Gouvea (egouvea+@cs.cmu.edu)
Ricky Houghton (ricky.houghton@cs.cmu.edu)
Kevin Lenzo (lenzo@cs.cmu.edu)
Ravi Mosur
Rita Singh (rsingh+@cs.cmu.edu)
Eric Thayer
Arthur Chan (archan@cs.cmu.edu)
Yitao Sun (yitao@cs.cmu.edu)
David Huggin-Dains (dhuggins@cs.cmu.edu)