Room Geometry Estimation from Acoustic Echoes using Graph-Based Echo Labeling

Conference Paper (2016)
Author(s)

Ingmar Jager (Student TU Delft)

Richard Heusdens (TU Delft - Signal Processing Systems)

Nikolay D. Gaubitch (TU Delft - Signal Processing Systems)

Research Group
Signal Processing Systems
DOI related publication
https://doi.org/10.1109/icassp.2016.7471625
More Info
expand_more
Publication Year
2016
Language
English
Research Group
Signal Processing Systems
Pages (from-to)
1-5
ISBN (electronic)
978-1-4799-9988-0
Reuse Rights

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons.

Abstract

A computer being able to estimate the geometry of a room could benefit applications such as auralization, robot navigation, virtual reality and teleconferencing. When estimating the geometry of a room using multiple microphones, the main challenge is to identify which reflections, or echoes, originate from the same wall and can, therefore, be modeled by a virtual source outside the room using the mirror image source model. In this paper we present a new and efficient method to disambiguate the echoes using a graph theoretical approach where echo combinations are modeled as nodes in a graph and the problem is stated as a maximum independent set problem. Once the echoes are correctly labelled, we know the locations of the virtual sources from which we can infer the room geometry. Experiments for shoe-box shaped rooms show that we can reliably estimate the room geometry within seconds on contemporary hardware and achieve centimeter precision on finding the vertices of the room.

Files

Heusdens16icassp2.pdf
(pdf | 0.325 Mb)
License info not available