Signal Processing Strategies For Enhancement Auditory Perception In Cochlear Implant

Cochlear implant (CI) is the only medical intervention device that can restore and improve hearing to profoundly deafened persons via directly electric stimulation of the residual auditory nerve. With the development of bionic ear technology, CI has achieved high speech recognition rate and been applied in clinical prosthetics. However, clinic reports show that there still exist some problems to be improved. For most prelingually deafened CI users, both of tone recognition rates and tone production performance are bad and the music perception is worse. The further studies on tone and music auditory perception are need and the relative coding strategies need for improving. Moreover, in noisy environments, the CI user auditory perception performances decrease rapidly. So the effective noise reduction strategy also is an important issue in CI improvement.This dissertation first examined the conventional CI processing strategies and the up-to-date research state. Then, focusing on the tone perception, the CI synthetic sound fidelity and noise reduction, some novel signal processing algorithms and strategies are presented. (1) For transmitting finer signal to electrode array, the all-phase DFT (APDFT) filter bank is introduced to CI waveform coding strategy. In virtue of the excellent characteristics of APDFT filter, the better simulation synthetic signal and auditory perception than that of the conventional non-linear phase filter are achieved and these testify the filter characteristics, especial for phase characteristic, can influence the CI auditory perception. (2) Face to the tone reception mechanism for normal-hearing, the auditory chimaera perception studies are presented, and the envelope and fine time structure are decomposed by APDFT filter banks and Hilbert transform. The tone reception is audibly tested though several manners auditory chimaera. The results show that the fine structure is more important to tone perception than the envelope, and mainly locates on the low frequency channels. Controlling the pulse firing time on corresponding electrodes is vital for improving CI tone reception. Then, an improved processing strategy is introduced, which can selectively track peaks and sequences so as to emphasize the effective signal perception. (3) For enhancing the CI auditory reception in background noise, two noise reduction methods are studied and improved. Aim at the problem of over threshold of conventional wavelet speech enhancement method, a time adaptive threshold (TAT) wavelet packet algorithm is presented. In each frame, the TAT can be adjusted by estimating the speech-present probability so as to make the recovered signal in CI simulation clearer, especially for word boundary perception. In order to speed up the noise reduction and to eliminate non-stationary noise in CI, a novel spectrum subtraction is presented based on the speech endpoint detection and the dynamic noise-estimation method. The power spectrum of noise speech is divided into Bark frequency sub-bands and the noise is tracked by the improved minima controlled recursive averaging method. The recovered sound simulated by ACE coding strategy is more intelligible and clearer than traditional spectrum subtraction, and the uncomfortable"music noise"is removed completely. (4) As our research tool, the CI Simulation Experiment Platform is developed and implemented; its functions and constituent frame are briefly introduced. Finally, a future landscape in amplification of CI is presented.