基于奇异谱分析的盲源分离单通道语音增强算法研究

针对盲源分离算法无法直接对单路含噪语音信号进行分离的问题,提出一种基于奇异谱分析的盲源分离单通道语音增强算法.通过对单路含噪语音信号进行奇异谱分析,将其低频分量作为第二路观测信号,利用改进的最大信噪比盲源分离算法进行处理,从而实现语音信号和噪声的分离.实验结果表明,该算法能够有效抑制噪声,提高信噪比,起到良好的语音增强效果.

Research of Single-Channel Speech Enhancement Algorithm Based on Blind Source Separation and Singular Spectrum Analysis

This paper proposes an algorithm based on the combination of singular spectrum analy-sis and blind source separation for single-channel speech enhancement to deal with the problem that the blind source separation algorithm cannot separate single-channel signals mixture directly. First,through singular spectrum analysis of the single-channel noisy signal,the second road ob-served signal is obtained by the reconstructed sequence of its low-frequency component.Then, with these two observed signals as the inputs of the improved blind source separation algorithm based on maximum SNR,the speech signal and noise are successfully separated.The experimental results show that this algorithm can effectively reduce the noise,improve the signal-noise ratio of the speech,and achieve a good enhancing effect.