The goal of this research is to develop efficient algorithms for physics-based room acoustics modeling and real-time auralization. Given the room geometry and wall materials, in addition to listener and sound source positions and other properties, the auralization system aims at reproducing the sound as would be heard by the listener in a corresponding physical setup. A secondary goal is to predict the room acoustics parameters reliably. The thesis presents a new algorithm for room acoustics modeling. The acoustic radiance transfer method is an element-based algorithm which models the energy transfer in the room like the acoustic radiosity technique, but is capable of modeling arbitrary local reflections defined as bidirectional reflectance distribution functions. Implementing real-time auralization requires efficient room acoustics modeling. This thesis presents three approaches for improving the speed of the modeling process. First, the room geometry can be reduced. For this purpose an algorithm, based on volumetric decomposition and reconstructions of the surface, is described. The algorithm is capable of simplifying the topology of the model and it is shown that the acoustical properties of the room are sufficiently well preserved with even 80 % reduction rates in typical room models. Second, some of the data required for room acoustics modeling can be precomputed. It is shown that in the beam tracing algorithm a visibility structure called "beam tree" can be precomputed efficiently, allowing even moving sound sources in simple cases. In the acoustic radiance transfer method, effects of the room geometry can be precomputed. Third, the run-time computation can be optimized. The thesis describes two optimization techniques for the beam tracing algorithm which are shown to speed up the process by two orders of magnitude. On the other hand, performing the precomputation for the acoustic radiance transfer method in the frequency domain allows a very efficient implementation of the final phase of the modeling on the graphics processing unit. An interactive auralization system, based on this technique is presented.
|Translated title of the contribution||Efficient physics-based room-acoustics modeling and auralization|
|Publication status||Published - 2010|
|MoE publication type||G5 Doctoral dissertation (article)|
- room acoustics modeling
- virtual reality