Abstract : We present a real-time, diffraction-aware sound–effect framework for VR/game environments that balances physical plausibility and computational efficiency. The core idea is to approximate obstacle-induced wave behavior by (i) interior curl vectors derived from obstacle geometry and branch points, and (ii) exterior vectors obtained as normalized gradients of a convolution-based smoothing map around the obstacle. The two fields are combined with distance-based weights to synthesize a stable propagation flow, and a magnitude field is formed by blending source–distance attenuation with boundary-aware attenuation to yield a final sound magnitude map. The method integrates cleanly in Unity3D and runs in real time. Across diverse scenes and source placements, the approach reproduces boundary-guided diffusion and corner diffraction with continuous vector fields and monotonic decay. Quantitatively, our metrics show high monotonic attenuation and shadow suppression (MAR up to 0.92, SSI up to 0.71), visible corner gains (CDG up to 0.22), strong field continuity (FCS 0.84∼0.88), low content–preservation error when modulating real music (CPE 0.08), and frame costs below 1 ms at 5122 resolution, supporting interactive deployment. The results indicate that curl–vector approximation coupled with convolutional boundary cues is an effective alternative between exact wave simulation and overly simplified models for immersive audio rendering.

[paper]