摘要
介绍一个与文本无关、与语种无关、与性别无关的声纹识别系统 ,采用的技术包括MFCC特征提取、VQ矢量量化、LBG聚类建模、数据有效性距离测度和计算机深层编程等 ,在字表大小为 5 0人以内时识别率接近 10 0 % .
A voiceprint recognition system without any reference to text,language and sex is described.The adopted technology includes MFCC feature extraction,vector quantization,LBG clustering,rangefinding using data validation,deepseated computer programming and so on.The recognitoin rate of this system approaches nearly 100% while therere no more than 50 peoples phoneme modulo in the experiment.
出处
《深圳大学学报（理工版）》
EI
CAS
2002年第2期78-80,共3页
Journal of Shenzhen University(Science and Engineering)
关键词
声纹识别
说话人辨认
矢量量化
聚类算法
距离测度
计算机识别
voiceprint recognition
speaker identification
Vector Quantification
clustering algorithm
rangefinding
