SNR improvement, filtering and spectral equalisation in cochlear implants using wavelet techniques

Abstract

This paper contains a brief outline of the engineering requirements for a cochlear implant speech processor based on wavelet techniques. Signal processing in cochlear implants typically involves filtering and noise removal, spectral estimation (to the frequency response of the hearing-impaired individual) and equalisation across desired frequencies. These various signal processing functions are typically implemented using separate hardware and microprocessor stages. In contrast, wavelet analysis inherently supports feature extraction, ambient noise removal and spectral estimation through the use of selected frequency-localised functions. Appropriate basis selection can incorporate noise, spectrum and equalisation (scaling) considerations. Thus it is possible to integrate the various signal processing requirements into a single DSP-based (digital signal processor) solution.