Kalman filters for time delay of arrival-based source localization

Abstract

In this work, we propose an algorithm for acoustic sourcelocalization based on time delay of arrival (TDOA) estimation. Inearlier work by other authors, an initial closed-formapproximation was first used to estimate the true position of thespeaker followed by a Kalman filtering stage to smooth the timeseries of estimates. In the proposed algorithm, this closed-formapproximation is eliminated by employing a Kalman filter todirectly update the speaker's position estimate based on theobserved TDOAs. In particular, the TDOAs comprise the observationassociated with an extended Kalman filter whose state correspondsto the speaker's position. We tested our algorithm on a data setconsisting of seminars held by actual speakers. Our experimentsrevealed that the proposed algorithm provides source localizationaccuracy superior to the standard spherical and linearintersection techniques. Moreover, the proposed algorithm,although relying on an iterative optimization scheme, provedefficient enough for real-time operation. Copyright © 2006 Hindawi Publishing Corporation. All rights reserved.