Abstract:

As for enhancing the speech recognition rate of the skinhearing system，we propose a  multiband spectral subtraction based speech enhancement algorithm. Colourednoise can depress the speech quality sharply in the experimental test of the skinhearing system, we embed the multiband spectral subtraction based speech enhancement algorithm into the skinhearing system. The multiband spectral subtraction based speech enhancement algorithm divides the whole frequency bands to multi bands that each band can be processed by its  features, and we emulate the algorithm by using Matlab and realize it by DSP hardware, which outputs the processed speech to the skinhearing system. The experimental results prove that it  can enhance the reliability and practicality of the skinhearing system.

Key words: skinhearing；spectral subtraction；speech enhancement；multichannel；DSP