Contribution
A filter-based physically-informed spherical head model for arbitrary source distances
* Presenting author
Abstract:
The rigid sphere represents a fundamental geometric shape suited to approximate the scattering behavior of more complex three-dimensional objects, such as the human head. Although the analytical solution for the scattering of a rigid sphere is well-established, its computational demands make it impractical for real-time audio applications. For plane-wave incidence, computationally-efficient, parametric filter approximations exist, mimicking main effects of head-related impulse responses (HRIRs), thus being suited for dynamic binaural rendering of far-field sources. This paper introduces a physically-informed, low-order digital filter approximation to model diffraction effects for the rigid sphere, accommodating arbitrary source and receiver positions. In the low-frequency (LF) range, spherical harmonics terms of the analytical wave solution are expressed by low-order digital filters. In the high-frequency range, the basic properties of the diffraction pattern are approximated by modelling two paths, which either reflects or bends around the sphere. Time- and frequency-domain results reveal the general validity of the proposed model for arbitrary distances. It is demonstrated that the current LF approximation explains the existing heuristic shelving filter approximation known as spherical head model and extends it to arbitrary source distances. Moreover, our LF approximation offers a physics-based filter solution to introduce near-field effects in existing HRIR models or measurements.