Hidden Markov model-based Assamese vowel phoneme recognition using cepstral features

Abstract

We use linear prediction cepstrum coefficients LPCC-based features, namely, the weighted LPCC and delta weighted LPCC, to recognise Assamese vowel phonemes employing a discrete hidden Markov model HMM. We create a small database for the Assamese vowels, spoken in isolation by 20 speakers with equal numbers of male and female speaker. Each spoken phoneme is repeated ten times by each speaker. Thus, our database consists of 1,600 phonemes out of which 1,000 phonemes are used for the training stage and the remaining 600 phonemes are used for the recognition stage in our experiment. The overall recognition rate of our experiment is nearly about 81.5%.