Real-Time Speech Enhancement by Adaptive Spectral Subtraction Method

Abstract

Non-stationary noise and strong background noise is difficult to extract the actual audio signal problem, an adaptive spectral reduction algorithm is proposed. A dynamic threshold algorithm is devised, iterative update mechanism and the specific implementation are contrived in the clean speech spectrum and noise spectrum estimating. Simulation experiments show that the algorithm can effectively de-noising filter, significantly improve the intelligibility of speech recognition system performance and read, and the method is robust in different noise environments and SNR. The algorithm complexity low, the computational cost is small, real-time, easy to implement, so that the effectiveness and real-time dual meet.