An auditory-nerve model with an instantaneous frequency glide and its application to modeling psychophysical results

Abstract

A computational model was developed to simulate the responses of auditory nerve (AN) fibers. The model consists of a time-varying band-pass filter with a nonlinear feed-forward control path, which changes the bandwidth and gain of the signal path. This model produces realistic response features to various stimuli, including pure tones, two-tone combinations, wide-band noise and clicks. The parameters of the band-pass filter were estimated by fitting the model revcor functions to revcor functions from experimental data of cat. The Marquardt method was used to minimize the difference between the model revcor function and the cat revcor function at various characteristic frequencies. Instantaneous frequency (IF) glides in the reverse correlation function of the models response to broad-band noise were achieved by simple configuration of the locations of the poles and zeros in the band-pass filter. The locations of the poles are continuously varied by the control signal to change the gain and bandwidth of the signal path, without affecting the IF profile, which is level independent. Other important properties, such as nonlinear compression, two-tone suppression and reasonable Q10 values are also included. Applications of this model for studying AN responses to speechlike signals will be discussed.