Precomputed Wave Simulation for Real-Time Sound Propagation of Dynamic Sources in Complex Scenes

摘要

We present a method for real-time sound propagation that capturesrnall wave effects, including diffraction and reverberation, for multiplernmoving sources and a moving listener in a complex, static 3Drnscene. It performs an offline numerical simulation over the scenernand then applies a novel technique to extract and compactly encodernthe perceptually salient information in the resulting acousticrnresponses. Each response is automatically broken into two phases:rnearly reflections (ER) and late reverberation (LR), via a thresholdrnon the temporal density of arriving wavefronts. The LR isrnsimulated and stored in the frequency domain, once per room inrnthe scene. The ER accounts for more detailed spatial variation,rnby recording a set of peak delays/amplitudes in the time domainrnand a residual frequency response sampled in octave frequencyrnbands, at each source/receiver point pair in a 5D grid. An efficientrnrun-time uses this precomputed representation to perform binauralrnsound rendering based on frequency-domain convolution. Ourrnsystem demonstrates realistic, wave-based acoustic effects in realrntime, including diffraction low-passing behind obstructions, soundrnfocusing, hollow reverberation in empty rooms, sound diffusion inrnfully-furnished rooms, and realistic late reverberation.