基于小波和先验信噪比维纳滤波的语音增强  

Speech Enhancement Based on Wavelet and a Priori Signal-to-noise Ratio Wiener Filter

作　　者：崔晓[1] 张松炜[2] 

机构地区：[1]郑州师范学院物理与电子科学系,郑州450044 [2]郑州航空工业管理学院电子通信工程系,郑州450015

出　　处：《河南师范大学学报（自然科学版）》2013年第1期43-46,共4页Journal of Henan Normal University(Natural Science Edition)

摘　　要：描述了先验信噪比估计的维纳滤波算法,分析了小波多分辨率分析在信号频谱划分中的作用,提出一种小波和先验信噪比维纳滤波相结合的改进算法.通过小波变换对带噪语音信号进行多尺度分解,然后对不同尺度的小波系数采用维纳滤波,用滤波后的小波系数重构得到增强语音信号.通过计算机仿真实验,将提出的算法与传统维纳滤波算法进行比较.实验结果表明改进算法在低信噪比情况下有效提高了增强效果,对语音成分的影响较小,提高了语音质量.A priori Signal-to-Noise ratio wiener filter is described,and the wavelet Multiresolution analysis in signal spectrum allocation is analysed,then an improved algorithm of combining Wavelet transform and Wiener filter is proposed.By using the wavelet transform,the noisy speech signal is decomposed into multiple scales,on different scales of wavelet coefficients using Wiener filtering,then the filtered wavelet coefficients are reconstructed to get the enhanced speech signal.Through computer simulation,the proposed algorithm was compared with the traditional Weiner algorithm.The experiment showes that the proposed algorithm effectively improves the enhancement result even in low signal-to-noise ratio situations,damages less to the voice component,and improves the quality of speech.

关 键 词：小波 多分辨分析 维纳滤波 低信噪比 语音增强 

分 类 号：TN912.3[电子电信—通信与信息系统]