Individualised and self-adapting sound processing for cochlear implants.

Cochlear implants are successful auditory prostheses that enable people with deafness to hear through direct electrical stimulation of the auditory nerve. In a cochlear implant sound processor, a sound signal is converted into a sequence of electrical pulses. This conversion entails many parameters that should ideally be fine-tuned (fitted) for every individual patient, to account for various anatomical and physiological differences, such as nerve fibre survival and neural adaptation. Current clinical fitting methods are very time consuming, so only the bare minimum number of parameters is fitted individually. However, it has been shown that for many other parameters, for which currently the same default values are used for all patients, better speech intelligibility can be achieved with individual fitting. We aim to provide better fitting to individual patients by recording brain activity evoked by speech from the electroencephalogram (EEG), and automatically fitting a wide array of parameters accordingly using a genetic algorithm. This will lead to improved speech intelligibility in noise and thus better communication and quality of life. For the clinic this means improved efficiency and the ability to better fit devices.

Keywords:Speech perception, Fitting, Eeg, Evoked responses, Cochlear implant

Disciplines:Otorhinolaryngology, Speech, language and hearing sciences, Modelling, Multimedia processing