Ithaca, NY - When computer animators discovered how to render hair, suddenly movies were full of cuddly creatures instead of plastic toys and reptiles.
But watery animations have remained thin on the ground, as up to now fluid sounds, such as splashing and pouring, have been hard to create. Movie makers lacked the physically based algorithms to synthesize them in computer animation or interactive virtual environments.
Now, computer scientists at Cornell University have proposed a practical method for synthesising bubble-based sounds from 3D fluid animations.
To avoid audio-rate time-stepping of compressible fluids, the team acoustically augmented existing incompressible fluid solvers with particle-based models for bubble creation, vibration, advection, and radiation. Sound radiation from harmonic fluid vibrations was modeled using a time-varying linear superposition of bubble oscillators.
Each oscillator was weighted by its bubble-to-ear acoustic transfer function, which was modeled as a discrete Green's function of the Helmholtz equation. To solve potentially millions of 3D Helmholtz problems, the team proposes a fast dual-domain multipole boundary-integral solver, with cost linear in the complexity of the fluid domain's boundary.
Enhancements are proposed for robust evaluation, noise elimination, acceleration, and parallelization.
The team has produced examples of harmonic fluid sounds for water drops, pouring, babbling, and splashing, often with thousands of acoustic bubbles and hundreds of thousands of transfer function solves.