A Modified a Priori SNR Estimation for Spectral Subtraction Speech Enhancement

In this paper, a modified a priori SNR estimation method used in spectral subtraction is presented to address the speech enhancement problem. To better track the speech change speed, an adaptive smoothing factor changing with respect to the frame energy variation is introduced into the modified decision directed (MDD) method. Simulation results show the musical noise can be suppressed at a low level, while weak speech components are preserved by using the proposed method. Improved results are obtained in terms of speech quality measures for various types of noise at different SNR levels.