摘要
本文的目的是阐明一种Mel频率倒谱参数特征的改进算法。该算法是通过线性预测的方法从语音信号中提取出残差相位,同时将残差相位与传统的MFCC相结合,并应用到语音识别系统中。该改进算法比传统的MFCC算法具有更好的识别率。
An improved extraction algorithm for the Mel frequency cepstral coefficient（MFCC） feature is presented.The residual phase is derived from the speech signals by a linear prediction analysis,and the residual phase and the MFCC features are combined. The experiments show that the proposed algorithm can effectively improve the recognition rate.
出处
《计算机工程与科学》
CSCD
北大核心
2009年第12期146-148,共3页
Computer Engineering & Science
基金
面上项目(Z20656)
关键词
语音识别
MEL倒谱系数
残差相位
线性预测
speech recognition
Mel frequency cepstral coefficient（MFCC）
residual phase
linear prediction
