Neural networks simulating the frequency discrimination of hearing for non-stationary short tone stimuli

Abstract

This paper addresses the question of frequency discrimination of hearing for non-stationary (short) tone stimuli (duration < or = 125 ms). Shortening of the stimulus duration leads to widening of the frequency spectrum of the tone. It can be shown that for hearing no acoustical uncertainty relation holds and thus some nonlinear elements must be present in hearing physiology. We present neurophysiological and psychoacoustical findings supporting the hypothesis that frequency discrimination of non-stationary short tone stimuli is performed in neural networks of the auditory system. Neural network architectures that could process the temporal and place excitation patterns originating in the cochlea are suggested. We show how these networks (temporal coincidence network processing the temporal code and lateral inhibition network processing the place code) can be combined to show performance consistent with auditory physiology. They might explain the frequency discrimination of hearing for non-stationary short tone stimuli. We show the fitting of psychophysical relations based on these networks with the experimentally determined data.