Physics-informed machine learning for sound field estimation and control

Speaker: Prof. Shoichi Koyama
National Institute of Informatics (NII), Tokyo, Japan

DEIB - BIO1 Room (Building 21, 2nd floor)
September 3rd, 2024 | 2 pm

Contact:  Prof. Mirco Pezzoli

Research Line: Signal processing for multimedia and telecommunications

Abstract

On September 3rd, 2024 at 2 pm the seminar "Physics-informed machine learning for sound field estimation and control" will take place at DEIB BIO1 Room (Building 21).

Sound field estimation and control are fundamental problems in acoustic signal processing, which are aimed at capturing and reproducing a spatial acoustic field with a discrete set of microphones or loudspeakers. These essential technologies have a wide variety of applications, such as VR/AR audio and active noise cancellation in a spatial region. For such problems, it is not sufficient to simply apply general signal processing and machine learning techniques; it is essential to establish techniques that appropriately incorporate prior knowledge of physical properties. An overview of our work on physics-informed signal processing/machine learning techniques for sound field estimation and control and their applications is presented.

Short Bio

Shoichi Koyama received his B.E., M.S, and Ph.D. degrees from the University of Tokyo, Tokyo, Japan, in 2007, 2009, and 2014, respectively. He is currently an Associate Professor at the National Institute of Informatics (NII), Tokyo, Japan. Prior to joining NII, he was a Researcher at Nippon Telegraph and Telephone Corporation (2009-2014), and Research Associate (2014-2018) and Lecturer (2018-2023) at the University of Tokyo, Tokyo, Japan. He was also a Visiting Researcher at Paris Diderot University (Paris 7), Institut Langevin, Paris, France (2016-2018), and a Visiting Associate Professor at Research Institute of Electrical Communication, Tohoku University, Miyagi, Japan (2020-2023). His research interests include audio signal processing/machine learning, acoustic inverse problems, and spatial audio.