COMBINING HIGHER ORDER REFLECTIONS WITH DIFFRACTIONS WITHOUT EXPLOSION OF COMPUTATION TIME: THE SOUND PARTICLE RADIOSITY METHOD

Abstract

The simulation of sound propagation in large rooms and urban environments is mainly performed by geometric simulation methods like ray tracing or the Sound Particle Simulation Method (SPSM). Hence, a severe deficiency is that wave effects are not included, especially if screening or diffraction effects are important. A method to introduce diffraction is the Uncertainty relation Based Diffraction (UBD) model, which has been successfully evaluated recently. To find close edges as sources of diffraction, a subdivision of the room into convex subspaces is performed by virtual walls. However, this causes a recursive split-up of Sound Particles (SPs) at each diffraction event. This effect should be compensated by a reunification of SPs. Therefore, the Sound Particle Radiosity (SPR) has been found that combines the SPSM with an advantage of the radiosity method: the re-unification of sound energy that uses a discretization of the surface into small patches. Now, SPR has been extended to 3D for the first time. To increase the available memory and to decrease the computation time, a parallelization has been implemented for the first time. First results indicate that the discretization of the virtual walls into patches yields additional but tolerable errors in the simulation of diffraction. However, even in 2D, SPR requires a huge memory. To solve this problem in 3D remains a great challenge, even more for more complex rooms. Also a method for a convex subdivision to 3D still has to be found.