Subjective Selection of Non-individual Head-related Transfer Functions

The individualization of head-related transfer functions (HRTFs) for virtual acoustics is a key technique for the optimization of the directional reproduction. This paper presents a subjective selection method for a fast, individual selection of one HRTF from a catalogue of non-individual ones. The selection method gives random access to sounds filtered with the HRTFs. In a first selection step a group of HRTFs is chosen out of which a final HRTF is singled out in a second step according to multiple criteria. The results of the two selection-steps were evaluated through a localization experiment. It is found that the selection minimizes the variance of the localization responses and the number of inside-the-head localizations. Localization error as well as the number of front-back confusions is small. As the selection method is fast, easy to implement, and operable for subjects without training, it is suitable for various applications, such as telephone conference systems or computer games.