Master Theses - Details

Spatial audio makes it possible to place sound sources in space and create an immersive listening experience. An example of this application can be found in virtual/augmented reality (VR/AR). Here, acoustic environments are created that enable users to immerse themselves in the virtual surrounding and experience the scene realistically.

Complete information about a sound field is available through knowledge of the sound pressure at the surface of a single sphere and the spherical harmonics, which appear as solutions of the wave equation in spherical coordinates. The higher the order of these functions, the more complex patterns can be described.

One challenge is the so-called physical sweet spot, which affects the spatial reproduction outside an ideal listening area, resulting in a loss of immersion and sound quality. In practice, the size of this area is limited by the order of the spherical harmonics. An interesting area of research is to investigate how and to what extent the sweet spot can be extended. To accomplish this, machine learning models can be used to scale up the order of the spherical harmonics and thus improve the sound quality.

The following points can be dealt with in this context as part of a final thesis:

• Identification and analysis of suitable assessment measures and implementation of training processes based on these measures
• Implementation and evaluation of new model-based and/or machine learning-based models (such as Generative Adversarial Networks, Diffusion Models, etc.) for HOA upscaling
• Evaluation of existing/new methods through measurements or hearing tests

If you are interested, further details can be discussed in a personal interview. The exact task of the thesis will then be determined in close consultation and according to the interests of the candidate.