Calculation of listener-specific head-related transfer functions: Effect of mesh quality

Abstract

The geometry of head and ears defines the listener-specific directional filtering of the incoming sound. The filtering is represented by the head-related transfer functions (HRTFs), which provide spectral features relevant for the localization of sound-sources. HRTFs can be acoustically measured or numerically calculated based on a geometric representation of the listener. While the acoustically measured HRTFs usually provide localization performance similar to that obtained in free-field listening, the performance obtained with numerically simulated HRTFs, however, heavily depends on the quality of the geometric and acoustic model of the listener used for the simulation. In this study, we show how to calculate listener-specific HRTFs with spectral features similar to that from acoustically measured HRTFs for the entire audible frequency range. We review the boundary-element method coupled with the fast-multipole method and we present details on the prerequisites like the geometry-capture technique, acoustical parameters, and the numerical algorithms. Further, the effect of the mesh quality on the HRTFs was investigated by systematically varying the average edge length from 1 to 5 mm. The HRTF amplitude spectra were analyzed and evaluated by visual comparison and in a localization model. The optimal average edge length for a fast calculation of HRTFs yielding potentially good localization performance is discussed.