Better-ear glimpsing for fully and partially degraded speech signals

Abstract

Normal-hearing listeners seem to take advantage of “better-ear glimpsing” when a target signal is masked by interfering sounds with different fluctuations in the two ears, presumably by analyzing small spectro-temporal bins and synthesizing the bins from the ear with the best signal-to-noise ratio (SNR). This study investigated the glimpsing mechanism using degraded inputs associated with bilateral cochlear implants (BICIs). BICI and NH listeners identified three-digit sequences spoken by a target talker at 0° with interfering talkers that were co-located at 0° or placed at ±60°. Glimpsing was evaluated by comparing the co-located control condition with either a standard bilateral condition (where the better-ear glimpses alternated across the ears) or a better-ear processed condition (where all the better-ear glimpses were artificially moved to one ear). Both NH and BICI listeners obtained substantial spatial benefit in the better-ear processed condition. However, the spatial benefit in the standard bilateral condition was greatly reduced for the BICI listeners and for NH listeners listening to vocoded signals. Varying the number of vocoder channels from 4 to 128 had little effect on the spatial release. The results provide insights into the cues listeners use to extract speech from fluctuating maskers. [Work supported by NIH R01-DC014948 (Goupell).]