Wideband Signal Processing Using Sequential MCMC Methods

Submitted to IEEE Transactions on Signal Processing, August 2003.
William Ng, J.P. Reilly, and T. Kirubarajan

Abstract   This paper proposes a novel online tracking and detection algorithm for wideband array signal processing. Based on the wideband data model in [1], this method extends the sequential Monte Carlo methods (otherwise known as particle filtering) techniques for recursive state estimation. This proposed method finds application in various areas of telecommunications, including radar, sonar, and other wireless communication problems, where locating moving targets and extracting the waveforms for classification are needed.

The proposed method recusively estimates the direction-of-arrival (DOA) information corresponding to the sources. The proposed method first detects the number of unknown wideband sources using a hypothesis testing procedure, followed by the sequential estimation of the DOAs of these identified sources and the waveform extraction of these sources using a particle filter. Computer simulations demonstrate the ability of the proposed method to detect the time-varying nunber of sources in different scenarios as time progresses. The results also show that the DOA trajectories of the identified sources are well tracked and their signal waveforms are also well restored. Finally, the posterior Cram´er-Rao bound (PCRB) [2] is also presented to demonstrate that the estimates by the proposed method lie consistently within the bound.