A hemispheric two-channel code accounts for binaural unmasking in humans
This provides a unified understanding of binaural unmasking for auditory neuroscience and hearing technology, though it is incremental as it builds on existing physiological models.
The paper tackled the problem of binaural unmasking, where sound detection improves when target and noise sources are spatially separated, by proposing a two-channel model based on the complex-valued correlation coefficient to quantify temporal fluctuations in interaural differences, and it accounted for 98% of variance across eight psychoacoustic studies.
Sound in noise is better detected or understood if target and masking sources originate from different locations. Mammalian physiology suggests that the neurocomputational process that underlies this binaural unmasking is based on two hemispheric channels that encode interaural differences in their relative neuronal activity. Here, we introduce a mathematical formulation of the two-channel model - the complex-valued correlation coefficient. We show that this formulation quantifies the amount of temporal fluctuations in interaural differences, which we suggest underlie binaural unmasking. We applied this model to an extensive library of psychoacoustic experiments, accounting for 98% of the variance across eight studies. Combining physiological plausibility with its success in explaining behavioral data, the proposed mechanism is a significant step towards a unified understanding of binaural unmasking and the encoding of interaural differences in general.