Identification of the spectrotemporal modulations that support speech intelligibility in hearing-impaired and normal-hearing listeners

Abstract

Accurate encoding of the spectrotemporal envelope of speech is essential for intelligible perception. We have recently developed a psychophysical procedure that identifies the relative contributions of particular spectrotemporal modulations (STMs) to intelligibility. Here, we use the procedure to test the hypothesis that intelligibility is supported by different patterns of STMs in hearing-impaired versus normal-hearing listeners. A group of 20 hearing-impaired listeners and an age-matched group of 13 normal-hearing (≤ 25 dB HL from 0.25-4 kHz) listeners performed a speech recognition task in which acoustically-degraded sentences presented over headphones were repeated back verbally and scored for keywords correctly identified. Different patterns of STMs were removed on each trial by applying a randomly-shaped filter to the 2-D modulation power spectrum. Reverse correlation was used to identify STMs that predicted performance (i.e., intelligibility) across trials. Three main findings describe the results: (1) the group-average patterns of STMs supporting intelligibility did not differ between hearing-impaired and normal-hearing listeners; (2) greater individual variability in STM patterns was observed within the hearing-impaired group; and (3) hearing-impaired listeners required more overall STM information to perform the task. The results suggest hearing-impaired listeners rely on the same STM information as normal-hearing listeners but encode this information less efficiently.