The Research On Speech Enhancement Of Digital Hearing Aids

Digital hearing aids can largely compensate for hearing loss,and choosing a reliable hearing aid can improve the hearing ability of hearing-impaired people.Based on the theory of speech signal processing,this paper focuses on the improvement and optimization of speech enhancement algorithms.The main work of the dissertation can be listed as follows:Firstly,the basic features of the speech signal,the perceptual characteristics of the human ear,the noise characteristics and the evaluation criteria of the speech quality are studied.The working principle of digital hearing aids and the research status at home and abroad are analyzed.The research status and principles of the key algorithms of digital hearing aids,such as sound source localization,loudness compensation,echo cancellation,are briefly summarized.Secondly,the wavelet threshold denoising algorithm is studied.Based on the hard and soft threshold functions,a new improved threshold function is proposed,which can solve the problems existing in hard and soft threshold function when reconstructing signals.The threshold function and the selection of the wavelet function directly determine the enhancement effect is good or bad.Thus,an improved continuous threshold function with tuning parameters is studied,which can deal with wavelet coefficients of noisy speech signals in wavelet domain.The feasibility of the new threshold function is analyzed,and artificial fish swarm algorithm is selected for the parameter selection in the new threshold function to achieve the minimum mean square error.Simulation results show that the new algorithm has better performance in speech enhancement.Thirdly,the generalized sidelobe canceling structure based on microphone array is studied,and an improved algorithm is put forward according to the existing problems.The wavelet threshold de-noising method is effectively combined with the beamforming algorithm,the wavelet threshold is firstly denoised for each signal of the microphone array,and the improved LMS algorithm is adopted for the adaptive GSC beamformer,The iterative step of the improved LMS algorithm changes with the signal to minimize the noise output power and speed up the renewal of the weight coefficient at the same time.The voice of the GSC processed voice activity detection(VAD)to distinguish the voice frame from the noise frame,Sexually speaking,speech frames and noise frames are respectively enhanced and attenuated,so as to compensate for the speech leakage.