Public defence in Acoustics and Audio Signal Processing, M.Sc. Georg Götz

The title of the thesis: Data-driven room-acoustic modelling

Doctoral student: Georg Götz
Opponent: Prof. Cheol-ho Jeong, DTU, Denmark
Custos: Prof. Ville Pulkki, Aalto University School of Electrical Engineering, Department of Information and Communications Engineering 

Room-acoustics research has traditionally been of interest in architectural planning and design as it studies sound propagation and the resulting sound field in enclosed spaces. With the spread of virtual- and augmented-reality technology, room-acoustic modelling has also become increasingly relevant for audio engines because it can make virtual sound sources sound like they originate from the same (virtual) environment as the listener. The dynamic and fast-paced nature of such applications requires audio rendering systems to operate in real-time. This means that whenever the listener or the sound sources move, the audio content must be adapted as soon as possible. However, accurate state-of-the-art room-acoustic-simulation technology is often computationally expensive, limiting its use for audio engines. Data-driven methods can be used to extract relevant room-acoustic information from large datasets, and offer the potential to bypass expensive simulations, while ensuring convincing perceptual experiences. This dissertation works towards data-driven audio engines by exploring the interaction between room-acoustic modelling and data-driven methods. 

As sound propagates through a room, it interacts with various surfaces like walls and furniture, leading to a gradual energy decay over time. The properties of this energy decay significantly influence the acoustic impression evoked by a room, making it a widely studied topic in room-acoustic research. This thesis introduces a neural network for sound-energy-decay analysis, which can be applied in complex geometries that consist of multiple rooms or feature a non-uniform distribution of absorptive materials. Moreover, spatial and directional variations of sound-energy decay are investigated, and a compact representation to model them is proposed. Data-driven approaches often require large datasets, but room-acoustic measurements are tedious and time-consuming. However, the measurements can be automated. To this end, two autonomous robot systems were developed. While the first one demonstrates the general idea and the design constraints of a practical system, the second one extends the measurement strategy to multi-room environments. 

Finally, a new approach for data-driven room-acoustic rendering is proposed, which builds upon the new room-acoustic theories developed in the beginning of the thesis. The presented approach is computationally efficient and can model spatially and directionally varying reverberation.

Keywords: room acoustics, sound-energy decay, inhomogeneous and anisotropic late reverberation, autonomous measurement robots, automatic data acquisition, machine learning, late-reverberation rendering

Thesis available for public display 10 days prior to the defence at: https://aaltodoc.aalto.fi/doc_public/eonly/riiputus/

Contact:

Doctoral theses in the School of Electrical Engineering: https://aaltodoc.aalto.fi/handle/123456789/53