Information-theoretic Performance Analysis of Azimuthal Localization for Spatial-Auditory Display of Beamformed Sonar Data

Abstract

An information-theoretic model of azimuthal localization is presented. The number of distinct source locations that can be encoded by a set of head-related impulse response functions (HRIR) is predicted in terms of information transfer as a function of the properties of the source signal and a quantization interval that is related to the level of internal perceptual noise. The model also predicts how source locations should be distributed in azimuth in order to maximize the information transferred through the set of HRIR for a given set of input conditions. The predictions are related to design considerations for a spatial-auditory display of beamformed sonar data in which time series associated with fixed beams of a one-dimensional array are mapped to virtual sources located at fixed radius from the listener in the horizontal plane.