Modelling Acoustic Propagation with Dominant Paths

Abstract

Hearing is one of our major senses. In addition to being, arguably, our primary form of communication, it affects how we perceive the world around us. Accurate acoustic simulation can complement graphical simulation systems, providing users with more immersive and believable environments. Furthermore, sound can assist users in localizing objects within space, and it has been shown to improve the perceived quality of visuals. Many of the audio properties that our brain subconsciously keys upon are artifacts inscribed into the signal by the environment as the sound propagates from the source to the listener. Diffraction, or the bending of sound around objects, is an important mode of transmission and a key source of reverberation in an audio signal. However, many current acoustic propagation simulation systems do not properly model diffraction.

This work proposes a new method for performing the simulation of diffraction. Our formulation aims to provide accurate results across the frequency spectrum. Accordingly, we avoid using approximative models of diffraction, such as the Unified Theory of Diffraction, which are ubiquitous in the field, but lead to inaccurate results particularly in the lower frequencies. Furthermore, our algorithm can provide a time-domain solution for reverberant sounds and is controllable for stylistic adjustments. In order to evaluate our method, we compare our results against measured and experimental results reported in the literature.