一种基于后验信噪比频域语音增强方法的研究  

Study on a New Spectral Domain Algorithm for Speech Enhancement Based on a Posteriori SNR

作　　者：陈紫强[1] 

机构地区：[1]桂林电子工业学院通信与信息工程系,广西桂林541004

出　　处：《电声技术》2007年第5期56-59,共4页Audio Engineering

基　　金：国家自然科学基金项目(60272038);广西自然基金项目(D20428)

摘　　要：谱减法是常用的单通道语音降噪方法,传统谱减法在抑制背景噪声的同时引入了“音乐噪声”,影响听觉效果。为了抑制音乐噪声,提出了一种基于后验信噪比的频域语音增强新方法,当后验信噪比较高时,采用基于后验信噪比的谱减法增强语音信号;当后验信噪比较低时,采用基于后验信噪比的谱衰减方法对含噪语音信号谱线进行衰减,达到语音增强的目的。仿真结果表明,基于后验信噪比的频域语音增强法具有较好的背景噪声和音乐噪声抑制效果,并保持了较好语音可懂度。Spectral subtraction algorithms are usually used in single-channel speech enhancement. The traditional spectral subtraction algorithms which can suppress the background noises will introduce the ＂music noise＂. The existence of the music noise strongly affects the hearing effect. To restrain the music noise, a new method based on a posteriori SNR is proposed. When the posteriori SNR is high, the spectral subtraction algorithm based on posteriori SNR is used to enhance the speech signal, otherwise the spectral attenuation algorithm is used. Simulating results show that the new method is excellent in the background noise and music noise cancellation while keeping little distortion to speech signals.

关 键 词：后验信噪比 语音增强 谱衰减 

分 类 号：TN912[电子电信—通信与信息系统]