基于平方根维纳滤波算法的噪声抑制研究

Research on Noise Suppression Based on Square Root Wiener Filtering Algorithm

摘要: 针对噪声抑制的语音增强问题，本文提出一种基于平方根维纳滤波的语音增强算法。通常增益函数引入的两种类型的语音失真的影响包括：当增强语音信号振幅的放大失真和衰减失真。事实上，当被处理的语音只包含衰减失真和小于6分贝的放大失真时，对语音质量的影响度较小。目前的数据表明，现有算法不能高质量提高语音清晰度的一个原因是，它们允许放大失真超过6分贝。本文主要提出一种针对幅度失真大于6分贝的噪声抑制算法，根据平方根维纳滤波增益函数特性，推导出基于先验信噪比约束条件的噪声抑制函数。实验结果表明，本文提出的算法能有效的提升语音质量，提高语音的可懂度。

Abstract: A speech enhancement algorithm based on square root Wiener filter is proposed for speech en-hancement with noise suppression. Generally speaking, the influence of the two types of speech distortion introduced by the gain function is to enhance the amplification distortion and attenua-tion distortion of the speech signal amplitude. In fact, when the processed speech contains only attenuation distortion and amplification distortion less than 6 dB, the influence on speech quality is relatively small. Current data show that one of the reasons that existing algorithms cannot improve speech intelligibility is that they allow amplification and distortion more than 6 dB. In this paper, a noise suppression algorithm for amplitude distortion more than 6 dB is proposed. Based on the gain function characteristic of square root Wiener filtering, a noise suppression function based on the constraints of a priori SNR is derived. Experimental results show that the algorithm proposed in this paper can effectively improve speech quality and improve speech intelligibility.