EE Seminar: An Efficient Simulation Tool For the Auditory System By Parallel Processing
Speaker: Yonatn Koral
M.Sc. student under the supervision of Prof. Miriam Furst-Yust and Prof. Shlomo Weiss
Wednesday, December 5th, 2018 at 15:30
Room 011, Kitot Bldg., Faculty of Engineering
An Efficient Simulation Tool For the Auditory System By Parallel Processing
Abstract
This work presents an auditory-emulation program that simulates the auditory neural response and determines hearing level based on a parallel time-domain nonlinear solution of the cochlea. Previous such simulations used a parallel calculation of the basilar membrane velocity and then calculated the serial neural response in the temporal and longitudinal dimensions. However, this approach suffers from limited efficiency because it requires copying large arrays and performing serial computation. In contrast, the proposed approach uses the graphics processing unit to do massively parallel computing, which accelerates threshold derivation based on the neural response calculation by a factor of 80 to 400. This is valuable both to calculate the hearing level for single-pitch signals and spoken words (with various types of noise or in silence) and to determine the benefit of different hearing aids to the hearing impaired.