Speech Recognition Study Based on MFCC and CDHMM and Its Realization Using a Mixed Program Method

This paper introduces the theory and implementation of CDHMM speech recognition. It provides an improving K-means initialization method based on character of speech signal. The technique of reliability measure is used to resolve the problem of parameters combination, which can considerably reduce the calculation burden. It can speed up the whole system development by using a mixed program mode with C++ and Matlab. ;;