Parametric Sound Field Interpolation for Scene-Based Navigable Immersive Audio

Authors:
Publisher website
Google Scholar
Abstract
Modern eXtended Reality (XR) systems are capable of tracking their users' positions within a navigable volume. This opens up the possibility of more realistic interactions with the content not only via head rotations, but also via locomotion. Such capability necessitates the design of audio systems that can generate a perceptually veridical rendering of the sound field at every position which the user can navigate to. While such interactivity can be provided by object-based audio (OBA) renderers such as in 3D computer games and VR applications, the new generation of volumetric visual content requires volumetric capture and rendering of the entire captured scenes, without explicitly identifying/extracting the audio objects therein. This article presents a parametric sound field interpolation method via multi-point sparse plane wave decomposition (SPWD) in the spherical harmonic domain. The proposed approach relies on higher-order Ambisonics (HOA) recordings of sound fields captured at multiple positions. The sound field is separated into direct and diffuse components that are interpolated parametrically and linearly, respectively. The proposed approach is compared with linear interpolation objectively and subjectively to demonstrate its utility.
Funding Information
  • Turkish Scientific and Technological Research Council (119E254)
  • Audio Signal Processing for six-degrees of Freedom (2019-2023)
  • Higher Education Council of Turkey Research Universities (ADEP-704-2024-11499)
  • Development of a deep-learning based MPAI Standard for Scene-Based 6DoF Immersive Audio Coding and Compression (2024-2026)