Based on the analysis of classical adaptive filtering algorithm,an effective speech enhancement algorithm is proposed,which could reduce noise and distortion effectively.According to the auditory masking properties,this algorithm uses masking thresholds of the key frequency segments as dynamic coefficients,to further improve the effect of speech enhancement and estimate error function dynamically.Signals with different SNR are input to simulate this proposed algrithm on Matlab.The high SNR and low residual noise of the output signals show that the algrithm is better than traditional filtering algrithms.