Integrating codebook and Wiener filtering for speech enhancement

In this paper, we propose a speech enhancement approach integrating codebook and Wiener filtering. Codebook-based method could reserve speech spectrum well, but leaves a large amount of residual noise. Signal to noise ratio (SNR)-estimation-based Wiener filtering method could reduce noise well, but leads to relatively serious speech distortion. To solve the problem, the gain functions of the above two methods are weighted to obtain an integrated gain function by utilizing a sigmoid function related to a posteriori SNR. The performance tests show that the performance of the proposed approach is better than codebook-based method in reducing the background noise and SNR-estimation-based algorithm in retaining speech spectrum.