Efficient Numerical Acoustic Simulation on Graphics Processors Using Adaptive Rectangular Decomposition 

Accurate acoustic simulation can enable realistic auralization that leads to enhanced immersion for visual applications, as well as facilitates accurate predictions for practical room acoustic scenarios. Numerical simulation provides realistic impulse responses that properly account for interference and diffraction effects by modeling the physics of wave propagation. However, it has posed a tough computational challenge owing to its large computation and memory requirements. We present a technique which relies on an adaptive rectangular decomposition of 3D scenes that yields two key advantages: Firstly, its key computational routine is DCT which can be efficiently parallelized on Graphics Processors. Secondly, the analytical solution of the Wave Equation in rectangular domains is known, which can be exploited to gain in accuracy and perform simulations on coarse simulation meshes, reducing both the computation and memory requirements further. Our technique is able to achieve a gain of at least a hundred-fold in computation and ten-fold in memory compared to a standard Finite Difference Time Domain (FDTD) implementation with comparable accuracy.