Contribution
Investigating Adaptive Order Schemes for Efficient Head-Related Transfer Function Computation
* Presenting author
Abstract:
Recent advancements in numerical methods, such as the Boundary Element Method (BEM) and Finite Element Method (FEM), have enabled efficient computation of Head-Related Transfer Functions (HRTFs) on consumer-grade devices. However, these simulations remain computationally demanding due to the wide frequency range and the intricate ear geometry, necessitating fine discretisation.This study explores the use of Adaptive Order (AO) schemes for BEM and FEM formulations to enhance accuracy and efficiency in HRTF simulations. These methods utilise hierarchical shape functions and dynamically adjust element orders based on geometry and frequency, ensuring that numerical errors remain within a pre-defined threshold. This approach is particularly advantageous for non-uniform meshes, commonly used in HRTF computations.Benchmarking against a state-of-the-art Fast-Multipole BEM (FM-BEM) technique, often used for numerical HRTF acquisition, demonstrates the potential of BEM AO and FEM AO for achieving high accuracy, while providing tuneable precision for specific simulation needs. Validation against experimental data shows minimal deviation between numerical and measured HRTFs, in line with FM-BEM results. However, AO formulations offer greater flexibility, allowing users to balance computational cost and accuracy, making them ideal for scenarios demanding high precision or limited computational resources.