基于混合进化计算的GMM优化方法及其在说话人辨认-应用科学学报.PDFVIP

20　 2 Vol. 20, No. 2 　 2002 6 JOU RN AL O F APPLIED SCIENCES June 2002 　　: 0255 82 7( 2002) 02 0141 04 GMM 崔玉红, 胡光锐, 何旭明 ( , 　 200030) 　 : ( EGMM ). EGMM, GMM ., . GMM GA EP ., EGMM ., GMM , EGMM 3% ,. : ; ; : TN 12. 3　　　: A Optimization of GMM Based on Hybrid EvolutionaryAlgorithm and Its Application in Speaker Identification - , - , - CU I Yu hong HU Guang rui HE Xu ming (Department of Electronic Engineering , Shanghai J iaotong University 200030, China) Abstract: When training data is limited, the performances of speaker identification ( SI) system deteriorate ow ing to the w eak generalization of GM M trained by EM algorithm. This paper pre- ( ) sents a solution of this problem by using a evolutionary Gaussian mixture model EGM M as a modeling method of SI system. Based on evolutionary algorithm, EGMM optimizes both the structure and the parameters of a GMM to better its generalization ability. Also, other speakers discriminative information is integrated into objective function to increase the accuracy of classifi- cation According to the characteristics of GM M. , we design two special evolutionary operators and present a new hybrid evolutionary algorithm using genetic algorithms and evolutionary pro- gramming. The preliminary experimental results show that the speaker models based on EGM M have more generalization ability. Compared with traditional GMM , the correct recognition rate of SI system based on EGMM increases by approximate