A fuzzy approach to statistical models in speech and speakerrecognition

摘要

A unified fuzzy approach to statistical models for speech andspeaker recognition is presented. Since the expectation-maximisation(EM) algorithm is a powerful learning method for maximising thelikelihood of the observed data in the presence of hidden variables, thefuzzy EM algorithm based on the fuzzy c-means algorithm is therebyestablished. From this fuzzy EM algorithm, the fuzzy algorithms forhidden Markov models, Gaussian mixture models, and vector quantisationare developed. The experimental results on T146 and ANDOSL speech datacorpora for speech and speaker recognition show that the fuzzy approachis capable of achieving higher recognition accuracy