基于小波分析的语音信号增强的研究  

Research on Speech Enhancement Based on Wavelet

作　　者：姚瑶[1] 

机构地区：[1]兰州交通大学自动化与电气工程学院,兰州甘肃730070

出　　处：《信息通信》2010年第3期59-61,共3页Information & Communications

摘　　要：语音增强目的是从带噪语音中尽可能纯净的原始语音,即消除含噪语音信号中的噪声成份,提高输入信号的信噪比。在实际应用环境中,语音都会不同程度受到噪声的干扰,噪声会影响语音质量,严重的情况下将语音完全淹没到噪声中,无法分辨。本文将读入的语音信号加入正态随机噪声,然后对含噪声的语音信号进行小波分解,估计噪声的方差,然后获取去噪的阈值并对小波分解的高频系数进行阈值量化,得到去噪后的语音信号。仿真证明此方法具有很好的增强效果。The objection of speech enhancement is to abstract original speech possiblely, which eliminate noise from speech, and improve signal-to-noise ratio. Speech will be affected by noise in reality. Speech will not be recognized in serious events. This paper added noise to speech and wavelet analysis would be done on speech with noise, variance of noise also would be done. Then the threshold of eliminating noise will be obtained and high-frequency coefficients of wavelet will be quantified, finally speech of eliminate noise will be obtained. Simulation testified that speech enhancement based on wavelet is effective.

关 键 词：小波分析 阈值 语音增强 

分 类 号：TN911[电子电信—通信与信息系统]