摘要
从说话人的语音信号中提取出能反映其个性特征的参数是声纹识别的关键环节之一。介绍了MEL倒谱系数(MFCC)及其提取方法,在此基础上提出了一阶差分(△MFCC),并将MFCC和△MFCC结合起来作为声纹识别的参数进行了MEL参数提取实验。实验结果表明将MFCC和△MFCC结合起来作为特征参数在识别性能上优于独立的MFCC。
The extraction of parameters reflecting the characteristics of speaker's speech signal is the key to voiceprint recognition.MEL cepstrum coefficient(MFCC) and extraction method was introduced in this paper.On this basis,the first difference(△MFCC) was put forward.MFCC and △MFCC were combined to conduct an experiment of parameter extraction from voiceprint recognition.The results show that the extraction combining MFCC with △MFCC is superior to the independent MFCC.
出处
《电源技术》
CAS
CSCD
北大核心
2011年第4期433-435,共3页
Chinese Journal of Power Sources
作者简介
黄或玉(1977-),男.内蒙古自治区人,硕士。讲师,主要研究方向为电子信息技术、模式识别,计算机测控技术等。