Evaluation of photogrammetry-based near-field head related transfer function estimation for immersive sound reproduction

Abstract

Binaural reproduction algorithms are commonly applied in virtual and augmented reality applications to create immersive audio environments. Nonetheless, flexible near-field binaural rendering remains challenging. In this work, we explore the acoustic accuracy of photogrammetry-based, near-field head-related transfer function estimation. Photogrammetry measurements of the Head and Torso Simulator (HATS) are taken using a photogrammetry rig consisting of 44 synchronized Canon 1300D DSLR cameras mounted on ten aluminium poles arranged in a circle of 0.8m radius. An Apple iPhone 14 Pro mobile phone is used to acquire the ear data. Commercial photogrammetry software is then used to create a 3D mesh from the still images. Finally, numerical acoustic simulations using the fast multipole boundary element method (FM-BEM) technique are applied to the mesh to generate the near-field HRTFs. We present comparisons of acoustically-measured and photogrammetry-based near-field HRTFs.