Evaluation of Fast BEM techniques for binaural synthesis including moving source or listener
Room/cavity acoustic simulations requires solution of very large models and usually are out of reach for standard element based methods. On the other hand, there have been various improvements on the Fast Boundary Element Methods in recent years, which allows tackling of extra-large problems with reasonable computational resources. The PhD research will concentrate on assessment of such advanced numerical methods in the context of room acoustic/cavity simulations first and will have a test validation track afterwards. The numerical part will focus on evaluation and implementation of fast Boundary Element Methods (BEM). This includes comparison of H-Matrix, H2-Matrix and Fast Multipole BEM (FM-BEM) techniques. Most recent algorithmic improvements to these methods will be observed. The next step will be the industrialization of the fast BEM, which will depend on the outcome of the comparison. Emerging software libraries and computing hardware (GPU and Xeon Phi architectures) will be evaluated for their applicability to BEM. The test part will focus on the validation of the developed numerical tools. In particular, binaural synthesis for interior acoustics will be carried out using BEM’s efficient post-processing feature. In addition, the identification of material properties of geometrically-complex structures using the developed models will be investigated.