Abstract
Estimating the elevation of sound sources is a challenging task for the auditory system due to the lack of interaural cues. In the 1930s, Wallach developed the first comprehensive theory on auditory elevation perception by means of auditory motion parallax during head movements. Years later, Blauert found that stationary listeners' perceived localization of narrowband sound sources correlated with the center frequencies of the narrowband sound sources rather than with their actual locations, leading to the idea of directional bands. Returning to Wallach's original idea, which has been modeled also by Zhong, Sun, and Yost in a robot hearing context, a new binaural algorithm will be presented to estimate elevation by auditory motion parallax. The model uses a front end with gammatone filterbanks and cross-correlators to compute binaural activity maps as a function of azimuth, elevation, and frequency. By continuously compensating for head rotations in the horizontal plane, all but the actual sound source position wash out, removing ambiguous positions from the cone of confusion. We will simulate binaural data in two primate species with vast differences in head size—humans and marmosets—to evaluate the generality of using the strategy of auditory motion parallax in elevation localization. [Work supported by NSF BCS-1539276, BCS-1539376, NIH F32 5F32DC017676 & ERC #618075.]
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call