Considerations for the Acceleration Structure of Sound Propagation on Mobile Devices: Kd-Trees Versus Multi-Bounding Volume Hierarchies

Sound propagation algorithms can provide immersive auditory experiences for users in various domains, such as games, virtual/augmented reality, and other applications. Recently, there has been a growing trend to apply spatial sound effects not only on desktops but also on mobile devices with limited computational resources. In geometric-acoustic (GA) sound propagation utilizing ray tracing methods for spatial audio effects, the choice of acceleration structure is an important factor that can either degrade or enhance performance. From this perspective, we seek to provide insights into the essential considerations for selecting acceleration structures for sound rendering on mobile devices. In this paper, we propose guidelines for mobile devices that address both how to select acceleration structures for sound rendering depending on the scene characteristics and how to optimize the selected acceleration structures. We used kd-trees and multi-bounding volume hierarchies (MBVHs), both of which are widely used acceleration structures for ray tracing. According to our experiments, our optimization approach, when compared against the baseline kd-trees, not only achieved performance improvements of up to 1.33× and 1.44× for optimized kd-trees and MBVHs, respectively, with minimal increases in power consumption on a Google Pixel 8a, but also enabled analysis of the advantages and disadvantages of each acceleration structure in various test scenes.We expect that our research will serve as a valuable reference for future studies on sound propagation and the broader multimedia community.