基于小波变换的双通道相干语音增强  

Two-Channel Coherence Speech Enhancement Based on Wavelet Transform

作　　者：任济生[1] 任鹏[2] 吉爱国[2] 曹茂永[3] 

机构地区：[1]山东科技大学机械电子工程学院,山东青岛266510 [2]青岛理工大学通信与电子工程学院,山东青岛266033 [3]山东科技大学信息与电气工程学院,山东青岛266510

出　　处：《电子器件》2008年第4期1352-1356,共5页Chinese Journal of Electron Devices

基　　金：国家自然科学基金资助项目(60472047);山东省优秀中青年科学家科研奖励基金资助项目(02BS011)

摘　　要：提出了一种基于小波变换的双通道语音增强方法。该方法将双通道接收的带噪语音分解到各小波子空间,在对噪声非平稳特性的精确估计的基础上有效地进行相干增强处理。针对语音和噪声在各小波子空间的不同时-频特性,采用不同参数值的MMSE-STSA估计器进行后续滤波,可以将遗留噪声进一步滤除。文章最后的实验结果证明,与传统的双通道相干语音增强方法相比,该方法具有更好的处理效果。One method of two-channel speech enhancement based on wavelet transform is proposed. The noisy speech observed in either channel is split into wavelet subspaces, in which coherence function is efficiently applied based on the accurate estimation of the characteristics of non-stationary noise. MMSE-STSA estimator acts as a post-processing filter in every wavelet subspace. It takes relevant parameter values according to the time-frequency features of the components in each subspace so that the residual noise can be removed to a further degree. The experimental results are presented in the end. The proposed two- channel method shows better capability of speech enhancement, compared with conventional ones.

关 键 词：小波变换 相干函数 MMSE-STSA估计器 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]