Research on Single-Channel Speech Enhancement Algorithm for Hearing Aids

At present, multi-channel loudness compensation algorithm is widely used for the hearing aids. However, the conventional multi-channel algorithm ignores the protection of speech feature, which will lead to the loss of recognition rate and deformation of speech structure. To solve this problem, a new single-channel speech enhancement algorithm based on multi-resolution wavelet is proposed in this paper. Multi-resolution wavelet is involved to decompose and reconstruct the signal’s spectrum, which can help to extract the envelope of signal spectrum. Then, calculate the required gain of each feature point based on the input-output curve. After that, the gain of whole signal spectrum is obtained by using the interpolation algorithm. The proposed algorithm shows fairly good performance in both stimulation and subjective test. The experiment and simulation indicate that the proposed algorithm has high practical value. It can not only compensate the loudness of voice, but also effectively protect the speech feature, which improves the speech recognition rate.