基于改进噪声方差估计的小波阈值语音增强  

Speech Enhancement Based on Improved Noise Variance Estimation in Wavelet Domain

作　　者：李玉[1] 于凤芹[1] 范树凯[1] 

机构地区：[1]江南大学通信与控制工程学院,江苏无锡214122

出　　处：《电声技术》2008年第3期56-58,共3页Audio Engineering

摘　　要：提出了一种估计语音噪声方差的改进方法,利用语音信号小波系数尺度间的相关性,去除了语音信号细节小波系数对噪声方差估计的干扰。仿真结果表明,方法能准确估计语音信号的噪声方差,尤其当语音噪声较小时,方法比传统的Donoho法更准确。An improved method is proposed to estimate the noise variance of noise, which eliminates the detail coefficients of the speech using correlation of the speech wavelet coefficients between scales, The simulation shows that the proposed method can provide better results than the traditional Donoho method, especially when the noise variance is small.

关 键 词：小波变换 噪声估计 语音增强 

分 类 号：TN912.35[电子电信—通信与信息系统]