Database matching of sparsely measured head-related transfer functions

This paper discusses a method of database matching applied on sparsely measured Head- Related Transfer Function (HRTF) datasets. Such an approach minimizes data collection durations through a selective measurement procedure without compromising spatialization accuracy by providing users with the best-fitting densely measured HRTFs from an existing repository. The coordinates of the sparse acoustic measurements are selected using Linear Discriminant Analysis (LDA) and represent the individuals' spatialization characteristics in the most discriminative manner. The sparse HRTF subsets consist of 68 measured positions non-uniformly sampled across five elevations between -30° and 30°. The designed database matching system was assessed by way of a binaural localization study. Results confirmed that the system is successful in pairing users with pre-measured HRTFs, as long as the objective similarity between measured and matched data lies above a target threshold.