利用基音和MFCC的性别分类器的设计与实现  

Design and Implementation of Gender Classifier Using Pitch and MFCC

作　　者：万洪杰[1] 杜利民[1] 邓浩江[1] 

机构地区：[1]中国科学院声学所语音交互技术研究中心,北京100080

出　　处：《微计算机应用》2007年第6期597-600,共4页Microcomputer Applications

摘　　要：在分析了基音频率参数、MFCC参数间关系以及二者与说话者性别之间联系的基础上,根据贝叶斯网络基本理论,构建了一种贝叶斯网络结构,并给出了依据该结构的性别分类器的构建和分类方法。通过多组实验,对比了不同MFCC参数对性能的影响,解析了该性别分类器的性别区分有效性。This paper analyzes the relationshi Pbetween pitch and MFCC paremeter and the relationshi Pof these two parameters with the gender of speakers. On this basis, combined with basic theory of bayesian network, we construct a bayesian network structure as gender classifier. Detailed discription are given on how to construct this classifier and how to use this classifier. Mutiple expriments analyze the different performance under different MFCC parameters. Results show that this classifier can discriminate the gender of speakers efficiently.

关 键 词：基音 MFCC 性别 贝叶斯网络 

分 类 号：TN912.3[电子电信—通信与信息系统]