基于维格纳分布的语音特征参数提取方法  

Extraction of Speech Feature Based on Wigner Distribution

作　　者：修国浩[1] 简伟[2] 

机构地区：[1]燕山大学信息科学与工程学院,河北秦皇岛066004 [2]解放军理工大学通信工程学院研究生4队,江苏南京210007

出　　处：《军事通信技术》2004年第3期8-11,共4页Journal of Military Communications Technology

摘　　要：文中将维格纳分布运用到语音信号处理中,对语音信号进行了时频分析。通过改进维格纳-威利(WV)谱的计算方法,提出了修正的WV谱的定义,提取了一种抗噪声语音特征参数WV-MFCC。仿真实验证明WV-MFCC具有优于传统MFCC的抗噪声性能。In this paper, Wigner distribution is applied in the processing of speech, by use of which speech is described in the time-frequency way. Through the improvement on the Wigner-Ville spectrum, the definition of modified WV spectrum is proposed, based on which one robust speech feature named WV-MFCC is extracted. The recognition experiment results show the WV-MFCC's robustness is prior to that of MFCC.

关 键 词：Mel频率倒谱参数 维格纳威利谱 修正的维格纳威利谱 维格纳威利-Mel频率倒谱参数 

分 类 号：TN912.34[电子电信—通信与信息系统]