Cochlear implant stimulation based on vector quantization

Cochlear stimulation algorithms usually extract some features of the voice like fundamental frequency, formants, energy in a passband filter, before exciting the cochlea through electrodes. They generally expose the patient to an infinite set of stimuli. In fact, it is well known that patients, on the first stimulation, don't know what they are hearing. To limit the number of stimulation sets, we propose to use vector quantization. This data compression algorithm is based on the use of a codebook which contains N referenced signals. The received signal is compared with all signals in the codebook and the one that is the nearest to the input signal is chosen. It is then possible to perform the stimulation of the cochlea on the basis of the chosen signal instead of the input signal. In that way, it is possible to minimize the number of stimulation sets of the cochlear implant.