The improvement and realization of speech enhancement algorithm based on Wiener filtering

In the speech enhancement algorithm, adding Mel-frequency domain processing will make the processed more in line with the characteristics of human ears. The implementation of this approach is based on the Wiener filter, which will be improved obviously of the sound quality, but will keep too much background noise. For this reason the output SNR(signal-to-noise ratio)reduced. At the same time, we all know the key to solve the Wiener filter is to make noise spectrum estimation accuracy. So, in this paper, these two aspects are designed to make an improvement. After the Mel-frequency domain processing, adding the gain factor based on priori SNR on each frame. What´s more, also designed a fast and effective noise spectrum estimation method, making the Wiener filter computation more efficiency. Experimental results proved that the processed signal SNR is improved obviously, the intelligibility of speech is good with a high quality.