RESEARCH ON ROBUST SPEAKER RECOGNITION SYSTEM

摘要

In order to improve robust performance,this paper studies a speaker recognition system which is made of speech enhancement module and speaker recognition module.The system adopts dynamic threshold algorithm which could improve ability of pretreatment noise canceling,lifting wavelet transform adding MFCC algorithm which enhance robust performance of feature parameter extraction and SVM-GMM algorithm which could advance recognition rate.The experiments show that this system has high robust performance and recognition rate.