HearClip - Personalization of Hearing Aids Through Bayesian Preference Elicitation (07605)

Modern hearing aids have evolved into complex digital signal processing devices. Currently, the vast majority of the hundreds of parameters of a hearing aid are fixed by the manufacturer at an average value. This lack of personalization might explain why about 25% of the people that buy hearing aids, do not actually use them. Our aim is to improve the hearing aid fitting procedure such that the hearing aid parameters can be personalized and geared towards the characteristics and preferences of individual users.

Our new fitting procedure will be based upon ideas from Bayesian incremental utility elicitation. With listening tests of the type "which of the two audio samples sounds best?" we elicit responses from the user. The responses will be used to update a probabilistic user satisfaction model. Bayesian techniques for experimental design can be applied to select the most informative listening experiments. The resulting procedures will be tested in audiological experiments.