Efficient Interactive Sound Propagation in Dynamic Environments

The physical phenomenon of sound is ubiquitous in our everyday life and is an important component of immersion in interactive virtual reality applications. Sound propagation involves modeling how sound is emitted from a source, interacts with the environment, and is received by a listener. Previous techniques for computing interactive sound propagation in dynamic scenes are based on geometric algorithms such as ray tracing. However, the performance and quality of these algorithms is strongly dependent on the number of rays traced. In addition, it is difficult to acquire acoustic material properties. It is also challenging to efficiently compute spatial sound effects from the output of ray tracing-based sound propagation. These problems lead to increased latency and less plausible sound in dynamic interactive environments. In this dissertation, we propose three approaches with the goal of addressing these challenges. First, we present an approach that utilizes temporal coherence in the sound field to reuse computation from previous simulation time steps. Secondly, we present a framework for the automatic acquisition of acoustic material properties using visual and audio measurements of real-world environments. Finally, we propose efficient techniques for computing directional spatial sound for sound propagation with low latency using head-related transfer functions (HRTF). We have evaluated both the performance and subjective impact of these techniques on a variety of complex dynamic indoor and outdoor environments and observe an order-of-magnitude speedup over previous approaches. The accuracy of our approaches has been validated against real-world measurements and previous methods. The proposed techniques enable interactive simulation of sound propagation in complex multi-source dynamic environments.Doctor of Philosoph