基于含噪语音统计特征的小波去噪算法分析与实现  

Wavelet de-noising algorithm analysis and implementation based on statistical feature of noised speech

作　　者：陈卒 周群彪[1] 陈正茂[1] 吕学斌[1] 

机构地区：[1]四川大学计算机学院,成都610064

出　　处：《四川大学学报（自然科学版）》2008年第2期291-295,共5页Journal of Sichuan University(Natural Science Edition)

基　　金：国家863高技术研究发展计划项目(2006AA12A104)

摘　　要：小波分析在信号处理方面具有优越性.本文将其应用于语音信号处理,所做的主要工作:证明了Daubechies小波语音能量守恒;基于含噪语音的统计特征,提出了样本方差阈值小波分解算法及重构算法;将新的阈值算法和Waveshrink算法用于语音除噪,并运用于多组实验进行对比.仿真数据表明,本算法效果较优.Wavelet analysis has advantages in signal processing, which is applied to process speech signals in this article. The main work includes： Energy Conservation Law of Daubechies wavelet transform of speech signals is proved. Based on the statistical feature of noised speech, Decomposition and Reconstruction algorithm of sample variance threshold wavelet is proposed. The new algorithm and Waveshrink algorithm are applied to speech denoising. Finally the contract between both results of the tests is provide . The simulation result show the new algorithm is superior.

关 键 词：小波分析 语音去噪 统计特征 阈值函数 

分 类 号：TN911[电子电信—通信与信息系统]