Identification of spectral features as sound localization cues in the external ear acoustics

The direction-dependent acoustic filtering of the external ear plays an important role in sound localization. It is believed to be essential for the detection of sound elevation, whereas sound azimuth follows from binaural difference cues in intensity and phase. Sound elevation is believed to be coded by prominent and characteristic features in the sound power spectrum at the sensory level, like peaks and valleys. However, it is still not known what the relative importance is of each of these features and how sound elevation can be calculated from them. This paper presents a localization model that enables extraction of the features that can be expected to be relevant in elevation detection. A feature was characterized as a spectral derivative of a certain order and scale, where a simple two-layer network was used to determine its importance in computing sound elevation.