loading page

Robust  spatial release from masking for spectrally degraded vocoded speech
  • Antje Ihlefeld
Antje Ihlefeld
New Jersey Institute of Technology
Author Profile


For the majority of cochlear-implant (CI) users, CI devices offer substantial benefits in quiet but not when background sounds interfere. To suppress unwanted sound, CI users can deploy spatial auditory attention. Specifically, when target and background speech originate from different directions, a subset of bilateral CI users can understand target speech better than when these sources are co-located, an improvement quantified as spatial release from masking (SRM). A known factor limiting CI sound quality is spectral resolution. Testing normal-hearing (NH) listeners with simulated bilateral CIs, two experiments explore how spectral resolution affected SRM. To assess SRM, closed-set bisyllabic target words were masked by sentences that were either co-located with the target or spatially separated. Target identification was measured as a function of both spectral resolution and the target-to-masker energy ratio (TMR). Experiment 1 contained stimuli with either 4, 8, or 16 spectral bands. In experiment 2, the spread of excitation along the cochlea was simulated. Specifically, in each speech band, the spread was attenuated with either 1, 2 or 4 dB per mm. To control for effects of stimulus uncertainty on SRM, experiment 3 assessed SRM as a function of response set size to include 2, 4 or 25 words. Generalized linear mixed-effects modeling revealed that 1) spectral resolution in isolation was not a major factor limiting SRM in simulated CI listening, and that 2) TMR shaped SRM, an effect that was most pronounced at low spectral resolution. Spectral resolution and TMR interacted, jointly affecting SRM.