Adaptive Sampling for Sound Propagation.

Area of Science:

• Computer Graphics
• Computational Acoustics
• Real-time Simulation

Background:

• Precomputed acoustics rely on discrete probes for dynamic sound rendering.
• Uniform probe grids are inefficient, requiring high density for complex geometries.
• Existing methods struggle with accurate sound propagation in narrow or occluded spaces.

Purpose of the Study:

• To develop an adaptive acoustic sampling method for improved sound propagation rendering.
• To enhance the realism of dynamic audio in virtual environments.
• To overcome limitations of uniform sampling and visibility-based solutions.

Main Methods:

• Implemented an offline 3D numerical simulation to precompute sound fields at probe locations.
• Introduced a novel "local diameter" measure based on stochastic path tracing to guide adaptive probe placement.
• Developed a runtime interpolator using geodesic paths and radial weights for smooth acoustic effect rendering.

Main Results:

• Adaptive sampling effectively varies probe density, optimizing resolution in complex areas like corners and corridors.
• The method eliminates undersampling in narrow spaces while coarsening in open areas.
• Achieved smooth acoustic effects respecting scene boundaries for moving sources and listeners, outperforming prior work.

Conclusions:

• Adaptive acoustic sampling significantly improves sound propagation quality and efficiency.
• The novel local diameter measure enables intelligent probe placement for enhanced realism.
• This approach offers a superior alternative to uniform grids and visibility-based methods for dynamic acoustics.