SSP'05 -- Proceedings

	IEEE/SP 13th workshop on Statistical Signal Processing July, 17-20, 2005 - Bordeaux - France

Information regarding the paper

Title

PHD Filtering for Tracking an Unknown Number of Sources Using an Array of Sensors

Author(s)

Balakumar Balasingam	McMaster University
Abhijit Sinha	McMaster University
Thiagalingam Kirubarajan	McMaster University
James Reilly	McMaster University

Get the paper in PDF format

To obtain Acrobat Reader (version 5 minimum required) necessary to his read.

Abstract

In this paper, direction of arrival (DOA) tracking of an unknown number of sources in a highly non-stationary environment is considered. Conventional DOA estimation techniques, such as MUSIC, fail when the stationarity assumption is violated. Furthermore, the time-varying number of sources makes the problem even more challenging. Recently, a particle filtering approach, which propagates the approximate posterior of the target states and then adopts a reversible jump Markov chain Monte Carlo (RJMCMC) diversity step to resolve the number of targets, was proposed. However, this algorithm is sensitive to incorrect model order initialization.

In this paper, we propose a new algorithm for tracking an unknown number of sources based on the probability hypothesis density (PHD) filter, which propagates only the first moment of the joint posterior distribution of targets in terms of particles, as a computationally efficient alternative to the RJMCMC method. The PHD algorithm provides an automatic way to estimate the number of sources, eliminating the need for a separate model order initialization or update step, which is typically the source of problem in particle-filtering based methods. In addition to the fact that the PHD implementation is simple, simulation results show that, the PHD implementation yields superior performance over the other method.