非平稳噪声环境下结合听觉掩蔽的语音增强

摘要

Aiming at the problem of residual noise,a modified speech enhancement algorithm combined with the masking proper-ties of human auditory system was presented.A two-stage speech enhancement algorithm framework combined with MMSE-LSA spectral estimation method and an optimum perceptual enhancement filter was constructed,which effectively removed the resi-dual noise using the masking properties.Detailed operation steps and the theoretical derivation of the perceptual enhancement fil-ter were given out.Results of the simulation indicate the speech distortion and residual noise are significantly reduced in condi-tions of different kinds of non-stationary background noise,and the subj ective and obj ective qualities are improved.%针对现有语音增强算法面临残留噪声这一问题，提出一种结合人耳听觉掩蔽的改进算法。将MMSE-LSA谱估计法和一种最优感知增强滤波器融入一个两极语音增强算法框架，利用人耳听觉掩蔽去除残留噪声；给出算法实施的具体步骤和最优感知滤波器的理论推导。实验结果表明，在非平稳噪声环境下，该算法可以有效降低语音失真和残余噪声，提升增强语音信号的主观和客观质量。