计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (22): 54-55.DOI: 10.3778/j.issn.1002-8331.2008.22.015

• 理论研究 • 上一篇    下一篇

语音信号特征提取中Mel倒谱系MFCC的改进算法

张 震,王化清   

  1. 郑州大学,郑州 450001
  • 收稿日期:2007-10-09 修回日期:2007-12-24 出版日期:2008-07-11 发布日期:2008-07-11
  • 通讯作者: 张 震

Improved algorithm of Mel-Frequence Cepstral Coefficients in characteristics extraction based on voice signal

ZHANG Zhen,WANG Hua-qing   

  1. Zhengzhou University,Zhengzhou 450001,China
  • Received:2007-10-09 Revised:2007-12-24 Online:2008-07-11 Published:2008-07-11
  • Contact: ZHANG Zhen

摘要: 从说话人的语音信号中提取说话人的个性特征是声纹识别的关键。主要介绍语音信号特征提取方法中的Mel倒谱系数(Mel-Frequence Cepstral Coefficients,MFCC)的特点及其改进算法(3Q+1),分析给出了较详细的计算过程并通过实验比较了其和传统算法在语音识别系统中的差别。

关键词: 语音识别, 特征提取, MFCC

Abstract: Extracting individual traits of the speakers from their voices is the key of voiceprint recongnition.This paper chiefly introduces Mel-Frequence Cepstral Coefficients and it’s improved algorithm(3Q+1),analyzes and gives detailed count process,and also gives the differences from the normal algorithm by experiment.

Key words: voice recognition, characteristics extraction, Mel-Frequence Ceptral Coefficients(MFCC)