Virtual Auditory Space: Individual Differences

摘要

Head related transfer functions (HRTFs) capture the direction dependent filter characteristics of the external ears. When a sound is filtered by HRTFs measured from a listener's own ears and played over headphones, the listener hears a virtual source that is well localized in space. When sounds were filtered by other listeners' HRTFs, listeners showed fairly accurate localization in the lateral dimension but showed conspicuous vertical and front/back errors. We examined differences among HRTFs measured from 45 listeners. We quantified differences by subtracting HRTFs between listeners for corresponding locations, then computing the variance of the resulting difference spectra across 393 locations. Interlistener differences could be reduced by shifting HRTFs scaling in frequency. Optimal scalars reduced variances by an average of 20.2% across all pairs of listeners and by more than 50% in 9.5% of listener pairs. The optimal scalar for any pair of listeners correlated highly with the relative sizes of certain physical dimensions. When HRTFs were shifted optimally then used in virtual localization trials, all measures of virtual localization performance tended to improve. In the majority of cases, the performance penalty for use of HRTFs from another listener was reduced by more than half.