Low Order Modeling Of Head Related Transfer Functions Based On Spectral Smoothing And Principal Component Analysis

Abstract

The presented paper discusses methods of reducing the order of HRTF filters. Two alternative methods of HRTF preprocessing were proposed: wavelet based approximation of HRTF magnitude spectrum and dimensionality reduction by means of Principal Component Analysis. Frequency components of HRTFs were weighted according to the non-linear frequency resolution of human hearing via transformation to the bark scale. The preprocessed HRTFs were modeled with either recursive or non-recursive filters in linear and non-linear frequency domains.