Abstract: After a noisy signal is denoised by auditory masking,the noise estimation errors lead to voice distortion.In the use of auditory masking value of the coefficient spectrum,an improved method of calculation is proposed and amended by adding parameters to curb excessive attenuation voice and reduce voice distortion,then based on the MMSE guidelines,to enhance the voice spectrum further smoothing treatment and further inhibit music noise.Experiments show that the algorithm improves signal to noise ratio and eliminates music noise without effects on voice distortion.The quality of subjective listening to voice improves markedly.