Funded PhD position on accurate numerical solvers for room acoustics modelling

About the Project

Wave-based numerical simulations are an essential tool for room acoustics modelling, supporting research, innovation, and design across areas such as architectural acoustics, immersive audio, and virtual reality. Although wave-based methods are based on physical models or equations, the simulation outputs produced by those techniques inherently contain numerical errors due to the discretization of the physical equations, limiting their accuracy.

The aim of this project is to develop approaches or correction strategies that enhance the accuracy of well-established numerical methods used in room acoustics modelling. The goal is to obtain more reliable and perceptually accurate simulation results.

The project will focus on a specific class of wave-based methods, the finite-difference time domain (FDTD) method, widely used in room acoustics modelling. Building upon previous knowledge on how numerical errors arise and propagate through simulations, the project will explore the use of deep learning as a tool to identify, model, and reduce the numerical errors. The effectiveness of the proposed approaches will be evaluated both in terms of improved numerical accuracy and their ability to reduce audible artefacts in auralisation and virtual acoustics applications, ensuring clear relevance to real-world use.

Please contact Julie Meyer for queries about the project.

Eligibility Criteria

The position is open to candidates with a strong background in Computer Science, Acoustics, Physics or a closely related topic. Further requirements:

• Knowledge of machine learning or willingness to learn it
• Good programming skills (e.g., Python, Matlab)
• Eagerness to develop new skills and knowledge
• Strong English oral and written communication skills
• Knowledge of wave-based simulations and/or room acoustics, as well as an interest in psychoacoustics experiments are a plus