Automatic speech recognition system for isolated words by using hidden Markov models

An automatic speech recognition system is presented in this thesis for speaker-independent, isolated words, in which the vector quantization and hidden Markov modeling are combined with a linear predictive coding analysis front end. Both the vector quantizer and the hidden Markov models need to be trained for the vocabulary being recognized. Such training results in a distinct hidden Markov model for each word of the vocabulary. The next step, classification, consists of computing the probability of generating the test word with each word model and choosing the word model that gives the highest probability. The entire recognizer has been evaluated on a 11-word vocabulary, which contains ZERO to NINE, and OH. A set of eleven spoken words from sixty-four speakers is used to train the automatic speech recognition system. Finally, an independent set of eleven spoken words from sixteen speakers is used to test the system. The success of this experiment will provide us an additional approach to a more effective automatic speech recognition system.