The effects of pitch, reverberation, and spatial separation on the intelligibility of speech masked by speech in normal-hearing and hearing-impaired listeners

Understanding speech in noisy and reverberant environments is a task that normal listeners can perform quite well and with which the hearing impaired have great difficulty. Comparing the performance of these two groups of listeners while varying perceptual cues gives insight into the segregation abilities of both normal and impaired auditory systems.;The ability of normal and hearing-impaired listeners to use the segregation cues of pitch and spatial separation was measured in a speech intelligibility task, and the detrimental effects of reverberation on those cues was also measured. The target was always simulated as straight ahead (at 0 degrees azimuth). Three spatial configurations of the two masking talkers were simulated: (1) both maskers colocated with the target at 0 degrees; (2) maskers symmetrically placed at +60 and -60 degrees; and (3) both maskers located at +60 degrees azimuth. Pitch contours of all sentences were monotonized with the target always at 100 Hz. Three masking sentence pitch conditions were tested: (1) both at 100 Hz; (2) both at +1 semitone (106 Hz); and (3) both at +4 semitones (126 Hz). The conditions were all tested under simulated anechoic, mild, and moderate reverberation levels.;Spatial release was correlated with amount of hearing loss, and decreased with increasing reverberation. Listeners in both groups were able to use the non-spatial segregation cue, pitch, to achieve lower thresholds in all spatial configurations and room conditions. Both segregation cues gave advantages, and the cues combined when both were available, due to the difficult segregation problem the experiment presented. Predictions of benefit gained from head-shadow and detriments due to audiogram thresholds were partially explained by SII model results, although all models failed to account for the variability of performance between subjects. Pitch factors were not included in the model, and thus was unable to predict the ability of subjects to use pitch. Some hearing-impaired listeners were able to get the same or more benefit from pitch as from spatial separation, suggesting non-spatial segregation cues could be exploited to improve speech intelligibility in hearing-impaired listeners in complex environments.