Isolated-Word Speech Recognition for Tamil Language using Hidden Markov Models

K.Yogendirakumar, A. R. Weerasinghe, W. G. D. M. Wathugala


Speech recognition technology has evolved with time and enhanced human interaction with computers. Although the recognition has limited capabilities like recognising discrete words or speech within a vocabulary, users of the commercial Automatic Speech Recognisers (ASR) are surprisingly contempt with such features. Still the speech scientists and the users of ASR know that, there is a long way to go in order to cope up with dynamic natural human speech. There are many current speech recognizers used for different versions of the English language with very good results. In this research a prototype of a speech recogniser for the Tamil language has been implemented. The type of recognition is in a discrete manner with a limited set of vocabulary. The approach used for the recognition is a probabilistic one, where the Hidden Markov Model (HMM) is used to model the speech utterance. The HMM implementation and the manipulation is done using the entropic toolkit HTK.

Citation Info :

In Conference Proceedings - 6th International Information Technology Conference on From Research to Reality, Infotel Lanka Society Colombo, Sri Lanka, 29 Nov- 01 Dec 2004, pp. 30-39, ISBN 955-8974-01-3.