The research of speaker recognition based on GMM and SVM

GMM and SVM models with different advantages and disadvantages were often used in speaker recognition system. The Principle and the characteristics of SVM and GMM were analyzed in paper. The method of combining the advantages of GMM and SVM as the speaker recognition model, mean parameters of GMM as SVM input, which not only can realize an accurate description of the speaker voice characteristics with fewer parameters, but also can achieve the purpose of compressed data for SVM. Experimental results shown the speaker recognition system based on GMM and SVM can effectively improve recognition rate.