Spatially transparent binaural beamforming for noise reduction in cochlear implant processors

Abstract While cochlear implants (CIs) have provided remarkable benefits to patients with severe-to-profound hearing loss, CI users still often struggle to perceive speech in noisy environments. A number of improvements have been developed to help CI users in such environments, including a) the use of directional microphones and related beamforming technologies to improve the signal-to-noise ratio, and b) the addition of a second cochlear implant on the contralateral ear (i.e., bilateral CIs), which not only allows the user to sense the directions of sounds (localization), but also improves speech perception in the presence of interfering noise from other directions. Most beamforming technologies and directional microphones currently improve the signal-to-noise ratio for the target sound, but at the cost of degrading some of the spatial cues that, if preserved, could lead to further intelligibility improvements. Other technologies require complex mathematical transformations that increase computational complexity or introduce significant delays. This project proposes an innovative and practical binaural beamforming algorithm suitable for implementation on today’s CI processors, and potentially on hearing aid platforms in the future. The goal of the algorithm is to provide highly directional spatial filtering in real-world noisy environments, while still retaining the advantages of binaural cues. Our approach will achieve this with efficient, real- time processing, free of the delays inherent with many existing approaches. The project will demonstrate feasibility of the proposed algorithm. First, a proof-of-concept beamformer based on our recently patented Virtual Reality (VR) audio processing algorithms will be improved and enhanced, with iterative evaluation and feedback from bilateral CI users. Second, the feasibility of implementing the resulting software in current CI processors will be ascertained. Third, an assessment of improvements in speech perception with background noise and sound source localization will confirm the ability of the algorithm to provide the foundation of a practical commercial product.