MODEL-BASED ASSESSMENT OF SPECTRAL WEIGHTING SCHEMES FOR MEDIAN-PLANE LOCALISATION

Abstract

Monaural spectral cues help humans to localise sound-sources in sagittal planes. These spectral cues mainly originate from the direction-specific filtering of the pinnae to the sound. While there is evidence that certain spectral regions contain crucial information used by the auditory system to infer the direction of the sound source, previous studies suggest that certain frequency regions are more important than others. To test this hypothesis, a sagittal-plane auditory localisation model has been modified to adopt a spectral weighting scheme. This weighting scheme assigns relative importance to the spectral cues within a frequency region to analyse its effect on predicting listener-specific patterns of localisation responses. Various model variants, which differed on their assigned spectral weights, were compared by means of Bayesian model selection. Our results show that the preferred spectral weighting was listener-specific with no clear preference at group level. Thus, our findings suggest that listeners apply different decoding strategies of pinna cues for sound localisation.