This event is free and open to the public, with no registration required.

If you are unable to attend in person, you can join via Teams.

Abstract

Reconstructing a sound field from sparse measurements is a central problem in spatial audio, with applications in immersive rendering, room acoustics, telepresence, and sound control. In realistic acoustic environments, the challenge is to interpolate pressure values and recover a field that is physically consistent with wave propagation.

This talk presents physics-informed machine learning as a framework with the application for sound field reconstruction. I will discuss how acoustic prior knowledge can be embedded into learning models either as regularization terms or as structural constraints. Particular attention will be given to physics-informed neural fields, which provide continuous representations of acoustic quantities over space.

The talk will highlight the distinction between physics-informed and physics-constrained approaches, their connection to classical acoustic modeling.

Mirco Pezzoli

Mirco Pezzoli is an Assistant Professor at Politecnico di Milano. His research focuses on spatial audio, musical acoustics, and audio signal processing, specializing in sound field reconstruction, extended audio reality, and the integration of machine learning with physics-based models. He earned his Ph.D. in Information Technology from Politecnico di Milano in 2021. His work spans immersive 6DoF audio systems and networked music performance. He has contributed to major projects such as the PNRR-funded FUNMedia and the Horizon Europe REPERTORIUM. In 2023 and 2024, he was a visiting researcher at the National Institute of Informatics in Tokyo. Currently he is visiting scientist at MIT through the Rocca Fellowship. Since 2025, Dr. Pezzoli has been a member of the EURASIP-ASMSP and EAA-ASP Technical Committees.