Download Our Software
Prototype System (v5.15): In this release, we have made
a number of minor bug fixes, and made slight improvements
to the runtime efficiency of the Baum-Welch trainer.
These bug fixes will not significantly
impact the overall decoding performance. This release was
tested on Solaris x86 and Linux (RedHat v9.0). Note also that
this is the first release in which we no longer officially support
Sun Sparc hardware.
Production System: A new release (r00_n11) of our C++ speech
for more information about the contents of this release.
Prototype System (v5.14): A cross-word context-dependent
LVCSR system. Recommended for speech technologists and
application developers. This release is compatible with gcc
version 3.2.1. Support to output the zeroth cepstral coefficient
has also been added to the extract_feature utility
(useful for some interoperability experiments).
This release (r00_n10) contains many common features found in modern
speech to text (STT) systems: a front end that converts the signal to
a sequence of feature vectors, an HMM-based acoustic model
trainer, and a time-synchronous hierarchical Viterbi decoder.
The new features over last release include:
- Phonetic decision tree-based state-tying;
- Lexical tree-based N-gram decoding (used to implement
cross-word phonetic models);
- Annotation graph support;
- Significant changes to the I/O interfaces for
- Integrated parallel processing support.
A tutorial is available
to help you get started, and an off-line version
can be downloaded as well.
- 2002 Software Archives
- 2001 Software Archives
- 2000 Software Archives
- 1999 Software Archives
- 1998 Software Archives
- 1997 Software Archives