Efficient Combination of Numerical Models for Full Audio Range Simulations in Virtual Reality

Due to the large range of audible frequencies, from 20 Hz to 20 kHz, the physical behavior of sound waves cannot be captured with a single numerical method. Therefore, different acoustic modelling strategies exist, based on the assessment if the sound waves can be categorized as low-, mid- or high frequency. For low-frequency problems the wave propagation is usually modeled with the Finite Element Method (FEM) or the Boundary Element Method, for mid frequencies wave-based methods are more efficient and for high frequencies methods like geometrical acoustics are preferred. However, for VR/AR applications it is generally desired to model the full audible frequency range and none of the above numerical methods will be suitable on its own. Therefore, this thesis will investigate how to combine the above mentioned techniques to obtain a single model that can render sound in the full audible range. Parallel filter banks for different frequency bands will be investigated first. They have already been tested with a reduced FEM model. With geometrical acoustics simulations available for the higher frequency bands, computationally efficient rendering of acoustical environments could be possible for the entire audible frequency range. Attention will be given to the effective use of massive parallel computing and the minimization of the error introduced by the combination of the individual models.