Monosyllable Isolated Word Recognition for Tamil language using Continuous Density Hidden Markov Model

Abstract

Developments in the communication technology have influenced the research trends to focus on speech technology. An Automatic Speech Recognition (ASR) may be classified into two based on the basic acoustic unit it can handle and they are word based or sub word based like phoneme, syllable, morpheme, senone. Hidden Markov Model (HMM) is the popular statistical model used in ASR and it assigns probability to the sequence of acoustic features extracted from the speech signal. Continuous Density Hidden Markov Models (CD-HMM) in which the observations are continuous and they are the important component of modern ASR Systems. Mel Frequency Cepstral Coefficients (MFCC) of the spoken words are used as training speech vectors to create CD-HMM which aid in recognition. This paper outlines the word based Isolated Word Recognition (IWR) for monosyllable words of Tamil language using CD-HMM and compares the result.