Tools and Applications
Sphinx-2
The speech recognition program we are
using is Sphinx-2. Developed at Carnegie Mellon University, this
software allows for real-time speech recognition without periods of
lengthy voice-training;
that is, the system is user-independant and will work just as well for
one person as it will for the next. Our project uses
Sphinx-2, a C-based distribution of the Sphinx project. More
information about Sphinx can be found at:
http://cmusphinx.sourceforge.net/html/cmusphinx.php
QuickLM is a tool for creating language models from
a given corpus file. Instructions for using QuickLM, and the tool
itself, may be found at:
http://www.speech.cs.cmu.edu/tools/factory.html
Phoenix
Phoenix is an open
source semantic parser available from the University of Colorado. It provides the
ability to extract the important information from
users' speech. The grammar can be tuned so that the dialogue manager
receives only the important information about the user's utterance in
the form of a series of [Name].VALUE fields. Each of these fields is
prefixed with the name of the frame that Phoenix matched--in our case,
this is the requested action. A comprehensive
guide to writing grammar files, in addition to downloadable source
code, may be found at:
http://cslr.colorado.edu/~whw/phoenix/download.html
Back
to the Main Page