Using channel-specific models to detect and remove reverberation in cochlear implants

Abstract

Reverberation results in the smearing of both harmonic and temporal elements of speech through self-masking (masking within an individual phoneme) and overlap-masking (masking of one phoneme by a preceding phoneme). Self-masking is responsible for flattening formant transitions, while overlap-masking results in the masking of low-energy consonants by higher-energy vowels. Reverberation effects, especially the flattening of formant transitions, are especially detrimental to cochlear implant listeners because they already have access to only limited spectral and temporal information (Kokkinakis and Loizou, 2011). Efforts to model and correct for reverberation in acoustic listening scenarios can be quite complex, requiring estimation of the room transfer function and localization of the source and receiver. However, due to the limited resolution associated with cochlear implant stimulation, simpler processing for reverberation detection and mitigation may be possible. This study models speech stimuli in a cochlear implant on a per-channel basis both in quiet and in reverberation, where reverberation is characterized by different reverberation times, room dimensions, and source locations. The efficacy of these models for detecting the presence of reverberation and subsequently removing its effects from speech stimuli is assessed. [This work was funded by the National Institutes of Health (NIDCD), R01-DC-007994-04.]