ASA 129th Meeting - Washington, DC - 1995 May 30 .. Jun 06

2pPP46. Infinite-Impulse-Response filter models of the head-related transfer function.

Abhijit Kulkarni

H. Steven Colburn

Dept. of Biomed. Eng., Boston Univ., 44 Cummington St., Boston, MA 02215

Directional-transfer-functions (DTFs) derived from head-related transfer functions (HRTFs) measured from human subjects in an anechoic environment were modeled using auto-regressive (AR) and auto-regressive moving-average (ARMA) modeling techniques. The HRTFs, which were provided by Dr. Fred Wightman (Wightman and Kistler, 1989), were decomposed into a directional-transfer-function (DTF) and a mean-function (computed as the mean of all measured HRTFs for that subject). Model reconstructions of the HRTFs was done under the assumption that HRTFs are minimum-phase functions. The AR estimator which is derived from the theory of linear prediction, corresponds to an all-pole model of the DTF. The ARMA model extends the all-pole model to a pole-zero model and is derived by posing a weighted-least-squares formulation of the modified-least-squares problem proposed by Kalman (1958). The validity of each model was assessed psychophysically using a 4I, 2AFC paradigm. Using a 80-ms white noise stimulus and testing a different position on each trial, for a 25-pole all-pole model and a 6-pole 6-zero pole-zero model, subjects performed close to chance for most of the 13 representative positions tested. [Work supported by NIDCD (Grant DC00100).]