Evaluation of three different standard head-related transfer functions using an elevation discrimination task for virtual sources displayed in the median plane

Abstract

The performance in controlling elevation effect of virtual sources displayed in the median plane was tested for three different sets of standard head-related transfer functions (HRTFs) using an elevation discrimination task. The three candidate sets compared included a representative subject's customized head-related impulse response (HRIR), mean-HRIR derived from the CIPIC database, and mean-HRIR of individually customized HRIRs proposed by Jo et al. [7]. The goal of the current study was to find the best standard HRTF dataset for many listeners, with particular emphasis on its ability to work well without requiring a given listener to engage in any customization procedure. A two interval two alternative forced choice (2-AFC) discrimination for virtual source elevation was used to find how easy it was for subjects to distinguish between elevation angles within 15 or 30 degrees of each other. The simple question asked of each subject was “Which sound source is perceived as higher between the two presented.” To compare the performance of elevation discrimination, individually measured HRTF, individually customized HRTF, and HRTF of KEMAR dummy were also used. The results from six subjects showed that even though the listeners' own measured HRTFs were used, they could not discriminate elevation well for sources differing only in terms of polar angle in the median plane. Similar poor performance was observed when using HRIRs from KEMAR. With regard to the search for standard filters for general use across listeners, a representative subject's customized HRTF, which was based on the method provided by Hwang, et al. [8] was preferred as it supported the best discrimination performance between the tested candidate HRTF datasets.