Effects of masker type on spatial release from masking in acoustic simulations of bilateral cochlear implants

Bilateral cochlear implant (CI) users show reduced benefit from spatial separation of target speech and background noise compared to normal hearing listeners. This is consistent with the fact that the availability of important spatial cues is severely limited by characteristics of typical CI speech processing strategies, such as the elimination of temporal fine structure (TFS). Additionally, the ability to use those spatial cues that survive CI speech processing may be limited because neural processing of binaural information has been degraded due to long-term deprivation of auditory input. The use of acoustic simulations of CI processing may help to isolate limitations due to CI speech processing from those associated with auditory deprivation. However, previous simulation studies have provided conflicting evidence. Garadat et al. (2009) observed differences in speech reception thresholds (SRTs) of up to around 8 dB due to spatial separation of target and masker, suggesting that even after elimination of binaural TFS cues, sufficient interaural envelope cues remained to allow substantial spatial release from masking (SRM). In contrast, Schoof et al. (2013) found only very limited evidence of SRM. The present work assessed the role of differences in target material and type of masker in this discrepancy. SRTs were measured for male target speech (spondees or BKB sentences) masked with either 20-talker babble or a single competing male talker. Spatial separation of speech and masker was simulated by applying a spherical head model prior to 6-channel noise-excited vocoding. Target speech was always presented with 0° azimuth. Masker azimuth was either 0° or 90°. SRM was significantly affected by masker type but not by target material. Mean differences in SRT due to masker azimuth were substantial for single-talker maskers (3.4 dB for BKBs, 6.7 dB for spondees) but were only around 1 dB in babble for both types of material. Differences in masker type therefore appear to underlie the contrast in previous findings, perhaps reflecting greater potential for SRM in the presence of informational masking.