The effects of cochlear-implant processing and current spread on the weighting of spectrotemporal speech information in noise

Abstract

CI listeners show a different weighting of acoustic information in quiet compared to NH listeners. For example, they rely less on cues such as formant transitions. However, less is known about how CI listeners weight spectrotemporal speech information in noise. It is hypothesized that the spectrotemporal regions of noise that are most critical to intelligibility for CI listeners are distinct from those for NH listeners. Specifically, CI processing and current spread are hypothesized to shift importance towards coarse temporal regions. NH listeners should weight regions associated with formant transitions and rapid transients more heavily. To test these hypotheses, an experiment was designed to reveal spectrotemporal importance functions for speech-in-noise intelligibility in NH listeners presented unprocessed and CI-simulated stimuli. The stimuli will include unprocessed vowel-consonant-vowel words presented in “bubble” noise, which randomly attenuates different spectrotemporal regions. Listeners (N = 8 expected) will also be presented with vocoded stimuli with attenuation rates simulating broad or shallow current spread. Spectrotemporal importance functions will be assessed through point-biserial correlation by comparing how different bubble noise regions impact intelligibility. These data could help inform improvements in CI processing and noise-reduction strategies, which often assume relatively uniform weighting of acoustic information.