Software

From SpeechWiki

(Difference between revisions)
Jump to: navigation, search
m (Data)
Line 99: Line 99:
: The library can read/write complete TrainableParameters files, as well as decision trees and most other objects allowed in GMTK parameter files.
: The library can read/write complete TrainableParameters files, as well as decision trees and most other objects allowed in GMTK parameter files.
: This library is based on the code from [http://code.google.com/p/extended-hidden-vector-state-parser/ EHVS parser] project, and so is available under the [http://www.gnu.org/licenses/gpl-3.0.html GPL3] license.
: This library is based on the code from [http://code.google.com/p/extended-hidden-vector-state-parser/ EHVS parser] project, and so is available under the [http://www.gnu.org/licenses/gpl-3.0.html GPL3] license.
 +
 +
 +
==LaTeX tools==
 +
 +
We also wrote some [[LaTeX]] scripts and templates. For now they deal with importing figures into latex and a properly formatted UIUC PhD thesis template in latex. (Arthur Kantor 2010)
==Miscellaneous==
==Miscellaneous==

Revision as of 23:33, 24 July 2010

Contents

Statistical Speech Technology Group Software

Our policy: everything we write is free on the web. This wiki is intended to be definitive, because anybody in the group can edit it to add their own software. A spider-indexable backup is at http://www.isle.uiuc.edu/software .

You can access each project by browsing an SVN snapshot online or downloading at tgz file by using one of the links below.

You can also check it out of our subversion server using login name "anon" with no password (hit "enter" when a password is requested).


Learning

Pronounce (Arthur Kantor, 2007)
Description, Demo, source, tgz
An orthographic string to phonetic string mapping tool.
This tool computes American English phonetic transcriptions from plaintext. Its HMM either generates a most likely phonetic transcription, or forces alignment if a phonetic transcription is provided. So, it gives a reasonable pronunciation for both out-of-dictionary words and partially pronounced words.
HTK-based Explicit-duration HMM (Ken Chen, 2003)
Description, source, tgz
HTKtrain (Sarah Borys and Mark Hasegawa-Johnson, 2008)
source, tgz
Scripts for training HMMs using HTK


Signal Processing

PVTK (Sarah Borys and MH 2005-8)
source, tgz
Extract HTK features as training vectors for libSVM, apply trained SVMs directly to feature files
VAD (Bowon Lee, 2007)
Description, source, tgz
Voice activity detector with improved noise model
Nested STFTs (Dave Cohen, Camille Goudeseune, Mark Hasegawa-Johnson 2009)
Description, stft.c
Efficient Simultaneous Multi-Scale Computation of FFTs
Improved Mistral (Qingsong Liu 2009)
State of the Art Text-Independent Speaker Verification System,especially for NIST SRE
Based on Mistral Open Source package
Improved and New Features:
  • add full factor analysis(eigenchannel and eigenvoice), instead of simple factor analysis(eigenchannel)
  • add multi-threads for Windows as well as Linux
  • support read HTK format feature/model
  • add an effective Algorithm for fast implementation of FA.
  • code optimization(for FA)
  • fixed some bugs
Source: /ws/ifp-32-2/hasegawa/pineking/programs/Improved_Mistral


Computation

GMTK Parallel (Arthur Kantor, 2008)
Description
The scripts are written and working but not packaged yet. Email Arthur for code.
Run GMTK commands in parallel on a compute cluster
HTK Parallel (Bowon Lee, 2006)
description, source, tgz
These Perl scripts Split an HTK command for parallel excution on a SGE cluster.
Matlab GMM (Arthur Kantor 2010)
source, tgz
A somewhat optimized Matlab toolbox for calculating the likelihood of many observations against many gaussian mixtures, each with many diagonal-covariance componenents.

Data

dtmfseg (Bowon Lee, 2006)
source, tgz
Segment audio files at DTMF tones
transcription tools (Mark Hasegawa-Johnson, 2005)
source, tgz
Convert transcription formats
speechfileformats (Mark Hasegawa-Johnson, 2004)
source, tgz
Read and write HTK files in matlab
CTMRedit (Jul Cha and MH 1999)
Description, source, tgz
Manually and automatically segment CT and MR image stacks
improved MVA (Arthur Kantor 2008)
Linux binary
Perform mean and variance normalization and ARMA filtering
It's essentially this version but with
  • better error reporting (e.g. failing to open file tells you so instead of core dumping)
  • more accurate mean and variance estimation (doubles instead of floats in strategic places)
  • faster computation in the case of MV (ARMA order 0)
svn location is svn://mickey.ifp.uiuc.edu/corporaNormalizationScripts/fisher/MVA.cc
Matlab pfile I/O toolbox (Arthur Kantor 2010)
source, tgz
A Matlab toolbox for reading and writing ICSI pfile data format used by GMTK and QuickNet.
The toolbox is designed to work with large pfiles (hundreds of GB).
It is based on the pfread.m and pfinfo.m scripts by Dan Ellis.
Python library for reading/writing GMTK parameter files (Arthur Kantor 2010)
source, tgz
The library can read/write complete TrainableParameters files, as well as decision trees and most other objects allowed in GMTK parameter files.
This library is based on the code from EHVS parser project, and so is available under the GPL3 license.


LaTeX tools

We also wrote some LaTeX scripts and templates. For now they deal with importing figures into latex and a properly formatted UIUC PhD thesis template in latex. (Arthur Kantor 2010)

Miscellaneous

Other scripts written in perl, python, bash, and ruby can be found in SVN archive.

There is also auto-generated documentation for them.

Personal tools