Abstract:

A signal-processing model is proposed for the head-related impulse response (HRIR). As in earlier work by Genuit [K. Genuit, ``A model for the description of outer-ear transmission characteristics,'' Ph.D. thesis, Rheinisch-Westfalischen Technichen Hochschule Aachen (1984)], different components of the model account for the directionally dependent filtering by the head, shoulders, and pinnae. Interaural-polar coordinates produce a natural separation of the azimuth- and elevation-dependent effects. The parameters of the head model depend only on head diameter. The parameters of the pinna model depend on pinna shape, and were derived from an image representation of the HRIR. The relative timing of the pinna ``echoes'' was found to be much more critical than their amplitudes or spectra. In a formal evaluation, subjects compared the apparent locations of virtual sources synthesized using the model to virtual sources synthesized using the subjects' individualized HRIRs. The effectiveness of the model was evaluated by having three subjects match the apparent location of a sound source. With a 500-ms Gaussian white noise burst heard through Etymotic Research ER-2 in-ear phones, the average error in matching elevations was less than 23 deg. [Work supported by NSF.]