Abstract
Cochlear prostheses are now widely used as a mean for restoring hearing to profoundly deaf people. Their performances depend on anatomic consideration as well as on the stimulation strategies used. The rehabilitation process of patients includes a reeducation period targeting to fit the device to the individual and to train him to distinguish different sounds. The stimulation algorithms used nowadays suffer from a weak immunity against noise and necessitate a relatively long reeducation period. In order to overcome this problem, we propose a new stimulation algorithm based on vector quantization technique to process the sound signal. This aims to minimize the sound segments to be identified by using a well defined finite sea of basic elements, and also to decrease the surrounding noise effect making easier the rehabilitation process and shortening the reeducation period.