Guest Speaker: Prof. Cheol-Ho Jeong
Department of Electrical and Photonics Engineering
Technical University of Denmark
Prof. Cheol-Ho Jeong is an Associate Professor in the Acoustic Technology group at the Department of Electrical and Photonics Engineering, Technical University of Denmark (DTU). He received his BSc, MSc, and PhD degrees in Mechanical Engineering from the Korea Advanced Institute of Science and Technology (KAIST), completing his PhD in 2007. He joined DTU as an Assistant Professor the same year and became an Associate Professor in 2011.
His research focuses on room and architectural acoustics, computational acoustics, acoustic material characterization, and data-driven methods for acoustic prediction and measurement. His broader interests include auditory virtual reality, speech and music perception in rooms, and multimodal perception, particularly how auditory and other sensory information jointly shape the experience of built environments. His recent work explores scientific machine learning for efficient sound field prediction, reconstruction from sparse measurements, and estimation of acoustic boundary properties.
Alongside his academic research, he develops practical acoustic tools, including ATLAS, a compact system for in-situ acoustic characterization. He supports the translation of acoustic research into industrial applications and served as a scientific advisor to Treble Technologies, an Icelandic start-up specializing in room acoustics simulation, during its early development. He is an Associate Editor of Applied Acoustics and received DTU’s Lecturer of the Year award in 2019.
Abstract
Accurate prediction and characterization of room acoustics require reliable computational models and boundary conditions, and sufficient measurement data. In practice, numerical simulations can be computationally demanding, acoustic material properties are often uncertain, and measurements provide only limited spatial information. Scientific machine learning (SciML) offers new opportunities to address these challenges by combining physical models with data-driven methods. This seminar presents recent developments in SciML for three interconnected tasks: sound field prediction, sound field reconstruction, and boundary condition estimation. The first concerns efficient prediction of room acoustic responses, including scenarios involving moving sources and changing boundary conditions. The second explores how physical constraints enable the reconstruction of spatially distributed sound fields from sparse pressure measurements. The third addresses the inverse estimation of acoustic boundary properties, including surface impedance and absorption, connecting sparse pressure measurements in sound fields to the characteristics of boundary surfaces. Examples will illustrate physics-informed and physics-constrained learning, with emphasis on computational efficiency, physical consistency, and experimental validation. Finally, current limitations and opportunities for integrating prediction, measurement, and inverse modelling to support acoustic design, material evaluation, and noise control will be discussed.
