Natural sound statistics and divisive normalization in the auditory system

Abstract

We explore the statistical properties of natural sound stimuli pre- processed with a bank of linear filters. The responses of such filters ex- hibit a striking form of statistical dependency, in which the response vari- ance of each filter grows with the response amplitude of filters tuned for nearby frequencies. These dependencies may be substantially reduced using an operation known as divisive normalization, in which the re- sponse of each filter is divided by aweighted sumof the rectified respons- es of other filters. The weights may be chosen to maximize the indepen- dence of the normalized responses for an ensemble of natural sounds. We demonstrate that the resulting model accounts for non-linearities in the response characteristics of the auditory nerve, by comparing model simulations to electrophysiological recordings. In previous work (NIP- S, 1998) we demonstrated that an analogous model derived from the s- tatistics of natural images accounts for non-linear properties of neurons in primary visual cortex. Thus, divisive normalization appears to be a generic mechanism for eliminating a type of statistical dependency that is