Study Of Speech Enhancement In Cochlear Implant Based On Characteristics Of Hearing

Cochlear implant is the only effective equipment which has the ability to renew auditive systerm of deaf people. This Technology has get better effect in the quiet enviroment, but the speech information which coclear implant users gets will be significantly decreased. To ensure the clearness and intelligibility of the speech information, and to extract usefull speech we needed as much as possible from the noisy speech, it is necessary to realize speech enhancement in cochlear implant.In this thesis the CIS speech signal processing is mainly discussed, procesing the improvement speech signal processor arithmetic,which is more adapt to the auditory system. And on the basis of this, proposing the algorithm of speech enhancement in cochlear implant based on characteristics of hearing.The main work of this thesis is below:1,During the discussion on the basis of speech signal processing, proposing wavelet which is very match to the auditory systerm to realize the CIS speech processing in cochlear implant.2,Research speech enhancement in cochlear implant. On the basis of discussing and comparing the traditional algorithms of speech enhancement in cochlear implant, including original spectral subtraction, improved spectral subtraction and self-adaptive filter, proposing the algorithm of speech enhancement in cochlear implant based on characteristics of hearing: Adopting the masking properties of human auditory as the subtract parameter to enhance speech in each channel.3,Carrying on simulating algorithms of speech enhancement in cochlear implant, and presenting the results. Comparing the SNR after using various methods. The result shows that the SNR output can be improved by 20 dB even if the SNR input is low and the enhanced speech is more clearness and intelligibility.4,Research the charater of speech signal in cochlear implant: format and pitch. Extract the format and pitch from the noisy speech,the speech enhanced by original spectral subtraction method, and the speech enhanced by the method of this paper proposing. The result shows that the relative error of the pitch and format extracted from the speech enhanced by the method of this paper proposing is less, and can represent the parameters of the original speech signal,and send more accurate information to cochlear implant users.Finally, the further research and improvement direction of this subject is raised.