The software contained on this page represents some of our original
contributions in signal processing. These packages are modular,
self-contained applications written in C++.
The packages below are listed alphabetically.
-
Decision Trees:
download a comprehensive general purpose decision tree package.
-
Discrete HMM Toolkit:
a simple set of tools to train discrete Hidden Markov models
using the viterbi and Baum-Welch algorithms.
-
Display Signal: tool that provides spectrograms and energy plots,
signal selection, and audio playback capabilities. Also is a single
channel version which is specifically designed for quick access to
multiple files from a single speaker (mono).
-
Echo Canceller:
a standard LMS FIR echo canceller that has been used in a wide range
of telephony applications including talk-through for speech
recognition.
-
Lattice Tools:
This set of perl scripts can be used to postprocess lattices
to add useful information such as posteriors and confidence measuress,
or to convert lattices to sausages.
-
NIST SPHERE:
a modified version of the NIST distribution that builds on Linux.
-
Parallel DSP and Fast Fourier Transforms (FFT):
a collaborative project that explores various FFT
algorithms in great detail in terms of computational efficiency.
-
Random Number Generator:
a portable random number generator with a uniform probability
distribution function.
-
Segmenter:
a graphical tool that allows you to segment and transcribe
speech. This tool, which was used for the SWB retranscription
project, supports several languages.
-
Signal and Endpoint Detection:
software to detect endpoints i.e. word, phrase or sentence
boundaries in a speech signal using energy-based criteria.
-
Signal-To-Noise Ratio:
software to implement an industry-standard approach
to estimating the signal-to-noise (SNR) ratio of an audio file.
-
Spiker:
software that detects flaws or artifacts in audio data.
-
Transcriber: is a graphical user interface tool for speech
segmentation and transcription that also provides spectrograms,
energy plots, speech selection, and audio playback capabilities.
Eventually, these packages will be merged into the IFC environment and
obsoleted. Hence, we refer to them as legacy code because we are no
longer actively developing them.
|