Fast acoustic obstruction with proximity cost differencing (PCD)

Acoustic obstruction is a difficult challenge faced in virtual reality and games. Obstruction is the loss in audio loudness due to objects and doorways that intervene the line-of-sight between a sound source and listener. Visual obstruction using a single ray-cast is fast to compute, but too abrupt for audio, sounding obviously unrealistic. On the other hand, using a full wave simulation is accurate but computationally expensive, requiring pre-computation. In this paper, we try to balance the two extremes with a new heuristic approach (PCD) that is orders-of-magnitude faster than wave simulation and shows promise for evolving into a real-time technique. We show that PCD captures the qualitative behaviors expected from wave diffraction. The proposed PCD algorithm for acoustic obstruction (middle row) is able to capture the qualitative variations from a small aperture compared to reference FDTD (bottom row) at more than 100 times speedup. In comparison, simple heuristics used in games today (top row) produce unrealistic results where a small pinhole can still transmit a large amount of energy, which results in easily audible artifacts.