A Cochlea Prosthesis Implementation Using Spectral Coding

In this work, we are interested in the implementation of a new speech coding strategy for cochlear implants. In order to restore hearing for deaf people, we were asked to find a numerical model which reflects the functioning of the auditory system. This model is based on analysis of the speech signal by a Gammachirp filter bank and the generation of electrical stimulation of the auditory nerve fibers. Stimuli with selected number and rank are applied to the electrodes by using a coding strategy which preserves the frequency and temporal information. To improve intelligibility, we used an adaptive coding strategy (ASAS-GF) which is based on a reduced number of channels and variable stimulation rates. The validation of our method was performed by subjective and objective tests such as the evaluation of the SNR, MQE, Recognition rate and listening tests in various environments (noisy, multi speakers). A second validation step has been developed by comparing the results of our strategy with those obtained with CIS and n of m.