My favorites | Sign in
Project Home Downloads Wiki Issues Source
Project Information
Members
Links

The Jave speech toolkit (JSTK) provides a native implementation of both library/API and applications for speech recognition, speaker verification, speech visualization (including transcription tools), and evaluation of related human rater tasks.

It is maintained by the speech group at the Univ. Erlangen-Nuremberg (www5.cs.fau.de)


JSTK - The Java Speech Toolkit

The Java Speech Tooklit (JSTK) is developed and maintained by the Speech Group at the University of Erlangen-Nuremberg. It is designed to provide both API and applications for the most popular speech tasks such as speech recognition, speaker verification, speech transcription and annotation and evaluation of human rater tasks. The current maintainers of the toolkit can be reached at

jstk@speech.informatik.uni-erlangen.de

The JSTK is licensed under GPLv3 and welcomes any contributions in terms of extensions, bugfixes, feature requests or comments. Note that code maintenance is (for now) only done at the speech group of Univ. Erlangen. Though this toolkit is intended to allow full-scale speech applications, special APIs such as the Google Android Speech API or Java Speech API are not implemented (but may however be implemented in the future).

The toolkit makes use of the following third party libraries:

JTransforms - http://sites.google.com/site/piotrwendykier/software/jtransforms

JTransforms provides (discrete) signal processing routines such as FFT, DHT, and DCT. It relies on state-of-the-art algorithms and even allows multi- threading.
JAMA - http://math.nist.gov/javanumerics/jama/
The JAMA package provides linear algebra routines such as decompositions (e.g. Eigen, LR, QR) and matrix operations (calculus, inversions). A single precision copy is maintained at the Speech Group named FJAMA, and a snapshot can be found in the download section of this project.

log4j - http://logging.apache.org/log4j

log4j provides scalable logging capabilities, however, JSTK only uses the very basic features of it -- for now. Currently, we use v 1.2.16

jspeex - http://sourceforge.net/projects/jspeex/files/jspeex

jspeex is used for decoding speex-compressed (*.spx) files. speex is a lossy codec comparable to MP3 or Ogg Vorbis, but optimized for speech, see http://www.speex.org. Currently using version 0.9.7.

junit - http://www.junit.org

Testing framework for java. currently used only sporadically. Currently using version 4.

sikoried, 19May2011

Powered by Google Project Hosting