Two-Microphone Noise Reduction for Cochlear Implants

Sensimetrics Corporation, Gloucester MA

Investigators

Abstract

DESCRIPTION (provided by applicant): Although there has been dramatic progress in the performance of cochlear implants (CIs), users of these devices continue to be severely hampered by background noise. Recent work with rapidly adapting two microphone noise reduction algorithms has shown promising improvements in speech-to-noise ratio with multiple noise sources. However, these results, and the few measurements of intelligibility with normal hearing listeners, were obtained in unrealistically favorable (low-reverberation) conditions. The proposed work will develop and assess new two-microphone noise reduction algorithms for cochlear implants. Two such algorithms will be implemented, one of them derived from prior work on linear array processing and the other from models of binaural hearing. Each algorithm will be implemented in real time and function as a preprocessor to the listener's implant. CI users will serve as subjects in tests of speech reception in acoustic conditions that include multiple noise sources and realistic degrees of reverberation. Performance with the two-microphone preprocessors will be compared to that without preprocessing. If the results of Phase I show clear intelligibility benefits of either preprocessing algorithm in realistic acoustic environments, work in Phase II will involve field trials with CI users, algorithm optimization and resource minimization for productization, and exploration of the algorithm(s) for binaural delivery to hearing aid users. Success in this project will result in enhanced speech reception by cochlear implant users in everyday noisy backgrounds.