Can You Hear the Shape of a Room?

Title

Can You Hear the Shape of a Room?

Speaker

Tom Sprunck - University of Strasbourg (France)

Abstract

Since its introduction by Allen and Berkley in 1979, the image source method has remained one of the most popular techniques for modeling room impulse responses (RIR) in acoustics. This method models each reflection of a sound impulse off the walls of a rectangular (or polyhedral) room as a Dirac impulse source outside the room boundary, derived from simple geometric criteria. Recent studies explore estimating the three-dimensional shape of a room by analyzing the time of arrival of echoes in recorded room impulse responses. However, this method has limitations, including the precise temporal localization of echoes and their labeling.

The method outlined in this presentation allows for the reconstruction of the 3D positions of image sources without requiring prior labeling of reflections.The inverse problem is formulated as an infinite-dimensional sparse reconstruction problem on the image-source distribution. The observation operator to be inverted involves a free-field Green function, i.e the solution of the wave equation with a Dirac source term. The dimensions of a cuboid can then be accurately estimated using the reconstructed image source point cloud. The presentation will conclude with the introduction of an alternative approach currently under development, based on shape optimization and the method of fundamental solutions. This approach aims to extend the analysis beyond rectangular rooms.

Bio

Tom Sprunck is a PhD student in applied mathematics at IRMA in Strasbourg, under the supervision of Yannick Privat, Antoine Deleforge and Cédric Foy. His thesis focuses on the inverse problem of reconstructing the shape of a room using filtered acoustic measurements of Room Impulse Responses, using sparse methods and shape optimization.

When

Wednesday, May 15th, 16:00

Where

DIBRIS/DIMA, Via Dodecaneso 35, Room 706