Durand R. Begault
Mail Stop 262-2, NASA---Ames Res. Ctr., Moffett Field, CA 94035
The relationship between measured and synthetic head-related transfer function (HRTF) headphone localization performance was measured by gathering direct localization estimates of speech material. Thirteen people participated in a ``within subjects'' design, and evaluated target positions at 30-deg azimuth intervals, all at 0-deg elevation. The measured HRTFs were of a nonparticipating subject. The synthetic HRTFs were derived by two procedures. (1) The magnitude transfer function of the original 512 coefficient impulse response was analyzed via a DFT, and then supplied with a specific weighing scheme to a Parks--McClellan FIR linear phase filter design algorithm, which output a 65-coefficient filter. (2) The interaural delay was implemented by averaging interaural group delay values between 0.2--6 kHz for each target position, resulting in a constant interaural time delay across frequency. Results showed that for each subject, localization judgments were nearly identical between conditions. These results parallel those found for HRTFs synthesized with a principal components analysis model using minimum phase filters [D. J. Kistler and F. L. Wightman, J. Acoust. Soc. Am. 91, 1637--1647 (1989)].