说话人识别系统中MFCC参数的改进算法  

Improved MFCC Algorithm in Speaker Recognition System

作　　者：兰胜坤[1,2] 石翼波 

机构地区：[1]福州海峡职业技术学院信息工程系,福建福州350102 [2]福建工程学院国脉信息学院,福建福州350102 [3]郑州新智信息技术有限公司,河南郑州450000

出　　处：《洛阳理工学院学报（自然科学版）》2013年第4期51-55,63,共6页Journal of Luoyang Institute of Science and Technology：Natural Science Edition

摘　　要：在说话人识别系统中,如何在语音信号中提取出能够表征说话人个性的特征参数是系统的关键问题之一。目前使用最多的M FCC参数主要描述了表征声道特性的谱包络特征,一般忽略了基音频率对M FCC的影响。由于基音频率能够影响M FCC参数对声道特性的准确描述,进而影响说话人识别系统的性能,因此本文提出了一种基于平滑短时幅度谱包络的S MFCC参数(smoothing MFCC)。实验表明,改进后的M FCC参数能够很好地减少基因频率对M FCC的影响,尤其对于基音频率较高的女性说话者,效果更为显著。In speaker recognition system,one of the key problems is how to extract the feature parameters which are characterizing the speaker.The currently most widely used MFCC parameter primarily describes the spectrum envelope of the sound tract characteristics and ignores the impacts of fundamental frequency theoretically.Given that the fundamental frequency is able to influence the description accuracy of MFCC parameters about the sound track characteristics,thus influence the performance of the speaker recognition system,this paper puts forward the smoothing MFCC（SMFCC）,which is based on smoothing short-term spectral amplitude envelope.Experimental results show that the improved MFCC parameters can degrade the bad influences of fundamental frequency effectively and upgrade the performances of speaker recognition system,especially for female speakers,who have higher fundamental frequency.

关 键 词：说话人识别 MFCC 谱包络 五点三次算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]