Abstract
Efficiency of cochlear stimulation is closely related to the signal processing of the input speech. A variety of techniques were proposed and used for cochlear prosthesis speech analysers. So different seems to be, these techniques are based on two major approaches. The first one consists of stimulating the cochlea according to extracted speech features and the second is based on wide band speech characteristics. Clinically, it is well known that in their initial rehabilitation phase, new cochlear prostheses users face some problems of distinguishing what they hear. This is due to the infinity of stimulation set related to the infinite possibilities of the input speech. In order to overcome this problem and minimise the rehabilitation period, a new stimulation algorithm based upon vector quantization is proposed.