基于MFCC和VQ码书的说话人识别系统研究  

Research on a Speaker Recognition System using MFCC Features and Vector Quantization

作　　者：张旭博[1] 周渊平[1] 

机构地区：[1]四川大学电子信息学院,四川成都610065

出　　处：《通信技术》2009年第9期162-164,共3页Communications Technology

基　　金：国家自然科学基金(编号:60672020)资助

摘　　要：随着信息技术和通信技术的迅速发展,说话人识别越来越受到重视,目前在说话人识别中,要提高识别率需要解决如何选取合适的有效的表征说话人特征的可靠参数和选取恰当的识别模型,文中在matlab实现了一个说话人识别系统,并在此基础上对系统进行了改进。with the development of communication and information technology, the speaker recognition attracts more and more attention from the people. Currently in the field of speaker recognition, there still exist two important unsolved problems for the enhancement of recognition rate, that is, how to select the best feature coefficient for the speaker system and how to select more effective and more reliable mode of speaker recognition. This article gives discussion of those two problems. And based on this some research and improvement is als made.

关 键 词：说话人识别 特征参数 MFCC(梅尔频标倒频系数) VQ(矢量量化模型) 

分 类 号：TP391.42[自动化与计算机技术—计算机应用技术]