基于滤波器组的MCAF噪声对消语音增强

MCAF Noise Cancellation Speech Enhancement Based on Filter Banks

在自适应噪声对消语音增强系统中,为了更好地加快自适应收敛速度,又不增加系统的计算复杂度,同时达到较好的增强效果,提出基于滤波器组的多通道自适应滤波(MCAF)语音增强;给出分析滤波器组与综合滤波器组的原型滤波器设计的具体方法。自适应滤波部分采用经典的LMS算法,同时结合多通道自适应滤波(MCAF),实现对含噪语音的处理,以达到增强效果。实验结果表明,相对于传统的子带LMS算法,基于滤波器组的多通道自适应滤波具有更好的性能,且加快了计算速度。

In adaptive noise cancellation speech enhancement system, in order to better speed up the convergence rate adaptive,not increase the complexity of the system,and to better enhance the effect,methods based on filter banks and the Multichannel Adaptive Filtering （MCAF） speech enhancement are proposed, the specific design methods of analysis of filters and filter group prototype filter are given. The adaptive filtering part uses the classic LMS algorithm combination of Multi -channel Adaptive Filter （MCAF）, realization of the noise to deal with to enhance the effect. The results show that, compared to the traditional sub - band LMS algorithm, based on the multi - channel filter banks adaptive filter has better performance, and accelerate pace of the calculation.