Speaker-independent Isolated Speech Recognition Research Articles

Speaker Independent Isolated Speech Recognition System for Tamil Language using HMM

The ambitious speech research for more than 50 years is having a machine to understand fluently spoken speech. This can be achieved by developing an Automatic Speech Recognition (ASR) system. Such a significant job specifically for Indian language has been focused in this paper. Different types of speech recognition systems are available for different application domains. This research work presents a speaker independent isolated speech recognition system for Tamil language. The most flexible and successful approach to speech recognition so far has been Hidden Markov Model (HMM) which is implemented in this research work. The HMM method provides a natural and highly reliable approach of recognizing speech for a broad range of applications. The experiments using HMM furnish high-quality word accuracy of 88% for trained and test utterances spoken by the speakers. The performance evaluation of the system is done based on the Word Error Rate (WER) which gives 0.88 WER for the above research work.