Physically Based Sound Synthesis for Large-Scale Virtual Environments

By offering a natural, intuitive interface with the virtual world, auditory display can enhance a user’s experience in a multimodal virtual environment and further improve the user’s sense of presence. However, compared to graphical display, sound synthesis has not been well investigated because of the extremely high computational cost for simulating realistic sounds. The state of the art for sound production in virtual environments is to use recorded sound clips that events in the virtual environment trigger, similar to how recorded animation sequences in the earlier days generated all the character motion in the virtual world. Although this technique has the clear advantage of being fast and simple, it has two main drawbacks. First, the sound generated is repetitive. Real sounds depend on how objects collide and where impact occurs, and prerecorded sound clips fail to capture such factors. Second, recording original sound clips for all the sound events in a virtual environment is a labor-intensive and tedious process. Physically based sound synthesis, on the other hand, can automatically capture the subtle shift of tone and timbre due to factors such as

Physically based sound synthesis, on the other hand, can automatically capture the subtle shift of tone and timbre due to factors such as change in impact location, material property, and object geometry. However, physically based sound synthesis has two computational requirements:

• An underlying physics engine. A physics engine informs the sound system of the exact collision geometry and the impact forces involved for every contact in the scene. Many recent commercial physics engines, such as the Havok Engine (see http://www.havok.com (opens in new tab)), can fill this requirement.
• Greater computing resources. Physically based sounds take significantly more computing resources than recorded sounds. Therefore, a brute-force sound simulation will not be able to achieve real-time performance.

In this article, we describe several techniques for accelerating sound simulation, thereby enabling realistic, physically based sound synthesis for large-scale virtual environments.