Speech Enhancement Based on Masking Properties and Short-Time Spectral Amplitude Estimation
一种基于短时谱估计和人耳掩蔽效应的语音增强算法

In this paper a single channel speech enhancement method for noisy speech signals is proposed, which is based on masking properties of the human auditory system and power spectral density estimation. During the estimation of power spectrum of the speech, the estimator can be modified by the MMSE method and the estimator of short -time spectral amplitude filter is determined by masking properties. In this way, the best trade off among the reduction of noise, the speech distortion and the level of musical residual noise can be found. Experimental results demonstrate the improved performance of the proposed algorithms.