基于人耳感知掩蔽效应的子空间语音增强算法研究

Ear Perceptual Masking Effect Based on Subspace Speech Enhancement Algorithm

针对传统子空间语音增强算法中,因语音增强方法中去除噪声而出现的音乐噪声和失真问题,提出了一种人耳感知掩蔽效应的子空间语音增强算法,并结合频域到特征值域的变换,在Bark域内实现人耳的感知掩蔽效应的语音增强。实验结果表明,该算法在白噪声和有色噪声的背景下,与传统子空间语音增强算法相比,不仅提高了语音信号的信噪比,而且减少了语音失真和音乐噪声,提高了增强后语音的听觉质量。

To address the speech distortion and music noise in the traditional subspace speech enhance- ment algorithm.This paper presented a perceptual masking effect of human ear subspace speech enhance- ment algorithm and incorporated into frequency domain to the transform of eigenvalue domain.At the same time,to realize perceptual masking effect of human ear for speech enhancement in Bark domain.Experimental results show the proposed algorithm not only improves the signal-to-noise ratio of speech signals,but also reduces speech distortion and music noise than the traditional subspace speech enhancement algorithm in white noise and colored noise background,to improve the quality of the enhanced speech hearing.