Relationship between pinna anthropometry and the frequency scaling of pinna-related transfer function principal components

Previous studies have shown that head-related transfer functions can be expressed as a weighted sum of basis vectors using Principal Components Analysis (PCA) and that, taking body height as a surrogate for head and pinna overall size, the PCs scale (shift in log frequency) toward higher or lower frequencies for shorter or taller individuals, respectively. Here, we directly investigated the relationship between various pinna dimensions and the frequency scaling of the PCs of Pinna Related Transfer Functions (PRTFs). Our data included three-dimensional (3-D) laser-scans of 70 pinnae(46 adults). These were transformed to 3-D voxel models, and the finite-difference time-domain method was used to calculate PRTFs at 1250 locations a distance 1 m from the ear-canal entrance. Thirty-one anatomical landmarks were manually defined on each pinna surface, and the Euclidean distance between every pair of landmarks was calculated. For each pair of landmarks, the PRTF dataset was divided into two groups based on the median distance between the landmarks, and the frequency scaling between the two groups was quantified for the top 5 PCs. As a result, the 10 largest frequency scaling amounts (5.7–6.7%) that were in accord with previous studies, corresponded to various measures of triangular fossa height. Notably, frequency scaling in the opposite direction (longer distances scaling the PCs up in frequency) were also found, the 10 largest of which (2.6–4.7%) corresponded mainly to various widths and depths of the cavum concha and cymba.