基于自适应滤波的噪声抵消法  

Noise Cancellation Based on Adaptive Filter

作　　者：郑海啸[1] 刘珩[1] 

机构地区：[1]中国农业大学信息与电气工程学院,北京100083

出　　处：《信息与电子工程》2007年第6期444-448,共5页information and electronic engineering

摘　　要：语音降噪就是从带噪语音信号中提取尽可能纯净的原始语音。文中介绍了一种基于自适应滤波的噪声抵消法,采用归一化最小均方误差算法,采集实际噪声环境下各种不同信噪比的带噪语音样本进行降噪处理,实验结果表明,处理后信号的信噪比得到了较大程度的提高,大大改善了听音效果,具有很高的可懂度,且语音自然度好,没有失真;并与谱减法进行了比较,自适应噪声抵消法的降噪幅度比谱减法有一定提高,在听音效果上,用自适应噪声抵消法处理后的语音在清晰度、自然度方面优于谱减法。The primary purpose of noise reduction is to extract more pure speech from noisy speech. This article introduces a method of noise cancellation based on adaptive filter, adopting normalized least mean square（NLMS） algorithm. The experiment is to collect different SNR noisy speech in factual noisy environment and then reduce the noise in computer. The emulation result shows that the SNR is increased largely and the effect of hearing is improved obviously with high intelligibility and no distortion. Compared with spectrum subtraction, the method of noise cancellation based on adaptive filter is more effective in terms of noise reduction extent and has higher definition and better naturalness of the speech in hearing.

关 键 词：自适应滤波 噪声抵消 降噪 归一化最小均方误差 

分 类 号：TN911.4[电子电信—通信与信息系统]