Mechanisms of spectro-temporal modulation detection and discrimination in normal-hearing and hearing-impaired listeners

Speech in noise understanding relies on the ability of our auditory system to extract relevant spectro-temporal modulations from noise: sensitivity of hearing-impaired listeners for detecting elementary spectro-temporal modulations is found to predict their speech-in-noise performances. However, we do not have a full computational understanding of this connection. Here, we used a two-fold approach combining psychophysics and modeling to probe the mechanisms underlying spectro-temporal modulation processing in both normal-hearing and hearing-impaired listeners. We ran several psychophysical experiments using a newly developed methodological framework based on reverse correlation deployed in the spectro-temporal modulation space and used system identification tools as well as current auditory models to determine the potential architecture of this processing and its underlying components. Both normal-hearing and hearing-impaired listeners were asked to detect or discriminate (upward vs downward) elementary spectro-temporal modulations called ‘ripples’ that were embedded in ‘ripple noise’, i.e. noise made of ripples of other spectro-temporal modulations with random energy. First, our results indicate that listeners rely on finely tuned but not fully directional band-pass filters to extract the target modulations. Second, they show that hearing-impaired listeners with similar audiometric loss exhibit a large variety of computational strategies. In order to further understand this variability, we used modeling tools to determine the impact of the different stages as well as their combined contribution in the processing. Overall, this interdisciplinary approach paves the way toward a computational characterization of human spectro-temporal modulation processing and should therefore contribute to a better understanding of supra-threshold auditory mechanisms and their deficits in general.