Softwares

Grapheme to phoneme converter for Tamil [Download]

The code implements a rule-based automatic grapheme to phoneme converter for Tamil. The rules are represented in Tamil.c. The code for aligning the lexicon to generate training examples for decision-tree based grapheme to phoneme conversion can be accessed here.

Building a speech recognizer for Indian languages using HTK - A Tutorial [Download]

In many cases, research for building spoken language technologies for minority languages is hampered by the lack of resources. There are also other confusions like standardization of scripts, output format, etc. This tutorial will introduce a novice (who is familiar with perl and shell scripting) through various steps of building a digit recognizer for Tamil.

Sphinx training scripts [Download]

I have included some of the perl scripts I used for training and testing a speech recognizer on TIDIGITS data using Sphinx. Hope you may find it useful. I have no idea of its compatibility in windows (try cygwin, if it gives any errors)

Text categorization [Download]

Some of the C and Perl code I wrote to perform text categorization on Emille corpus using Weka.

Automatic Echo Cancellation [Download]

The C code contains an implementation of adaptive echo cancellation. The project was done during the ADSP course at IIT Madras. The directory also contains the assembly implementation on ADSP 2192-12 (Analog DSP). I have included necessary files to run the code using Visual DSP++.

Universal speech models converter [Download]

This tool can convert sphinx continuous models into HTK compatible models. The present version of the code can be obtained from the CVS of sourecforge. I have tested it.