Text Dependent Speaker Independent Isolated Word Speech Recognition Using HMM

This paper aims to develop text dependent spea ker independent isolated word recognition system using HMM. In this paper MFCC is used fo r extracting acoustic features and then used to train HMM parameters using forward backward algorithm with EM principle . GMM is used to model the distribution of speech features for each state of HMM. F in ally the computed HMM parameters for words are stored to the database in respective HMM models. To recognize the spoken w ord the likelihood of generation of test speech features from stored HMM model and the most likely path sequence through the HMM models are found. Finally the one with maximum likelihood is selected as recognized word.