On the Impact of Training HRTF-Based Auralisation

Abstract

Auralisation with generic, non-individualised, Head Related Transfer Functions (HRTF) is common practice, as obtaining individualised HRTFs poses very serious practical difficulties. It is therefore extremely important to assess to what extent this hinders our 3D sound localisation capabilities. Here, we address this issue from a learning perspective. We carried out a set of experiments to better understand how the 3D virtual source localisation performance of listeners using generic HRTF is influenced by training. In experiment 1, we observed that listeners perform fairly well in terms of azimuth discrimination, but mere exposure to the tests does not cause performance improvement. In experiments 2 and 3 we implemented a short training period on both azimuth (exp2.) and elevation (exp. 3) discrimination. Training involved active learning and feedback and led to significantly better results. We therefore propose that in order to fulfil its perceptual potential, auralisation based on generic HRTF sets should always be preceded by a period of training.