New method based on single-channel separation algorithm using Gammatone filterbank for cochlear implants

This paper presents a new approach of coding music signals for cochlear implants; the proposed method is based on a Gammatone filter bank and an unsupervised single channel separation of non stationary signals algorithm. This method can increase recognition of musical instrument by cochlear implant users because the stimulus is separated for each instrument. Experimental tests show that the proposed method for cochlear stimulation is efficient with a mixture of signals.