Principal components of non-individualized head related transfer functions significant for azimuth perception

This paper demonstrates that Principal Component Analysis (PCA) can be used as a very effective tool for identifying the components of Head Related Transfer Function (HRTF) magnitude that are significant for azimuth perception. The research focuses on the azimuth localization accuracy of non-individual HRTFs combined with individualized Interaural Time Differences (ITDs). It shows that at a fixed elevation, magnitudes can be satisfactorily approximated using only two principal components and weights. The high accuracy of such modeling is proved theoretically and with two different localization tests. Not only is a high reduction of data achieved, but also a mathematical formulation of two remainder weights is proposed that enables simple interpolation of data for non-measured positions.