Accurate and Fast Auditory Virtual Reality Implementation

Nikolas Borrel-Jensen

Background
In many applications, it is desirable to model and simulate acoustics, such as architectural design, computer game audio, and virtual reality systems (VR). Often the simulations are limited by real-time constraints and especially for architectural design and VR, the correctness is important. Wave-based methods are solving the underlying wave equation numerically and are therefore precise from a physical point of view, but the drawback of these methods is that they are computationally expensive.

Project
The objective of this PhD project is to develop new methods for simulating sound in spaces (“room acoustics”) with unprecedented accuracy and efficiency. We will investigate a new hybrid method combining an inflexible - but efficient - method running in the bulk of the domain where only air is present, with a flexible method taking complex geometries and boundary materials into account. We hope that running a highly effective solver restricted to simple geometries in the bulk of the domain can speed up the overall simulation time many-fold while still ensuring accurate results. Another task important for practical applications is to understand the accuracy needed for various frequency bands for the errors to be negligible. We will focus on determining the upper cross-over frequency limit where wave phenomena can be approximated by non-wave-based methods such as geometrical acoustics.

Perspective
We hope that the outcome of this work can be implemented in software applications reaching near-real-time room acoustic simulations.

The project will be completed in 2023.

Name of supervisors
Cheol-Ho Jeong, Allan Peter Engsig-Karup, Maarten Hornikx

 


Contact

Nikolas Borrel-Jensen
PhD student
DTU Electro