Headphone Virtualization: Improved Localization and Externalization of Non-Individualized HRTFs by Cluster Analysis
Research and experimentation is described that aims to prove the hypothesis that by allowing a listener to choose a single non-individualized profile of HRTFs from a subset of maximally different best representative profiles extracted from a database improved localization, and externalization can be achieved for the listener. k-means cluster analysis of entire impulse responses is used to identify the subset of profiles. Experimentation in a controlled environment shows that test subjects who were offered a choice of a preferred HRTF profile were able to consistently discriminate between a front center or rear center virtualized sound source 78.6% of the time, compared with 64.3% in a second group given an arbitrary HRTF profile. Similar results were obtained from virtualizations in uncontrolled environments.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is temporarily free for AES members.