Individualizing a monaural beamformer for cochlear implant users

Abstract

Speech intelligibility in noisy environments is still quite limited for cochlear implant (CI) users. Classical beamformers such as the Generalized Sidelobe Canceller (GSC) can provide large improvements in speech intelligibility for CI users. These algorithms have been adopted from hearing aids and multimedia applications into the CI field. However, their optimization taking into consideration the peculiarities of electrical hearing with a CI has not yet been completely investigated. This paper presents a novel method to optimize the performance of a GSC for each individual CI user. We show through a combination of objective and novel subjective measures, how much distortion can be tolerated by a CI user without decreasing speech intelligibility. Experimental results with 5 CI users show that a GSC delivering just noticeable distortion is the one maximizing speech intelligibility for CI users.