A rigid sphere model for the head-related transfer function and its application to audio signal processing.

摘要

In this thesis a physical model for sound localization is examined. The head related transfer function (HRTF) is determined by calculating the pressure on the surface of a rigid sphere ensonified by a distant point acoustic source. The rigid sphere is used to model the human head. To determine the scattering of the spherical wave from the rigid sphere, the incident pressure Pi and scattered pressure P s are calculated using spherical harmonics. The magnitude of HRTF is the even valued function of the azimuthal angle. The group delay, dg, at low frequency and small angle is lower than the lower its asymptotic value. The phase increases linearly with frequency. The resulting magnitude and group delay can be used to localize the direction of the source. The results of the spherical model transfer function are compared with that of Brown and Dudas head related transfer function.; To render spatial localization effects of the model an audio signal is convolved with the spherical model HRTF in real time. To deal with the convolution of a short duration impulse response with the long duration input signal the fast convolution and the Overlap-add method is used. OpenGL and OpenAL libraries are used to develop the software required to render the audio signal using a computer.