Virtual localization improved by scaling nonindividualized external-ear transfer functions in frequency.

This study examined virtual sound localization in three conditions that differed according to the directional transfer functions (DTFs) that were used to synthesize the virtual targets. The own-ear and other-ear conditions used DTFs measured from listeners' own ears and those measured from other subjects, respectively. The scaled-ear condition employed other-ear DTFs that were scaled in frequency to minimize the mismatch between spectral features in the listener's and the other subject's DTFs. All measures of localization error typically were lowest in the own-ear condition. In other-ear conditions, all error measures tended to increase in proportion to the inter-subject differences in DTFs. When spectral features in an other-ear set of DTFs fell systematically lower in frequency than in a listener's own DTFs, low frontal targets typically were reported as low in the rear, and high rear targets were reported as high in front. When spectral features in a set of DTFs fell systematically higher in frequency than in a listener's own DTFs, elevation judgements showed an upward bias. In the scaled-ear condition, all measures of performance tended to improve relative to the other-ear condition. In the majority of cases, frequency scaling more than halved the penalty for use of another subject's DTFs.