Konstantinos Gkanos will defend his thesis entitled "Interactive, real-time room acoustics"
Supervisors: Cheol-Ho Jeong, Finnur Pind, Hans Henrik Brandenborg Sørensen (DTU Compute)
Censor: Claus Lynge Christensen (ODEON)
The aim of this thesis is to create an Acoustic Virtual Reality system with dynamic source and receiver, meaning that the source/receiver can move freely in the virtual scene. The acoustic simulations and auralization process operate entirely in realtime and the system is able to adjust to any change in the virtual environment or in the acoustical parameters of the space. The two major objectives of this thesis are the combination of the image source method and the ray tracing/secondary source method in a hybrid model as well as the parallel programming of each method and the optimization in terms of performance. Both methods are designed to operate on CPU architecture and the deterministic image source method is also implemented on GPU. The operation time and limitations of each algorithm are measured for various acoustical configurations and the performance of the image source method for both hardware architectures was obtained. The simulation techniques are also validated in terms of computational accuracy and reverberation time, by comparing to Odeon software and to Phased Acoustical Radiosity and Image Source Method (PARISM) model. Lastly, the algorithms are embedded in Unity and Virtual Reality scene is build. Regarding the hybrid model, an energy imbalance between the two methods is investigated and some techniques are suggested. Moreover, it is revealed that the GPU implementation of the IS method is computationally more efficient, and it is suggested to be compared to the stochastic image source method on GPU that has already been developed.