Marked poisson point process PHD filter for DOA tracking

Abstract

In this paper we propose a Track Before Detect (TBD) filter for Direction Of Arrival (DOA) tracking of multiple targets from phased-array observations. The phased-array model poses a new problem since each target emits a signal, called source signal. Existing methods consider the source signal as part of the system state. This is inefficient, especially for particle approximations of posteriors, where samples are drawn from the higher-dimensional posterior of the extended state. To address this problem, we propose a novel Marked Poisson Point Process (MPPP) model and derive the Probability Hypothesis Density (PHD) filter that adaptively estimates target DOAs. The PPP models variations of both the number and the location of points representing targets. The mark of a point represents the source signal, without the need of an extended state. Recursive formulas for the MPPP PHD filter are derived with simulations showcasing improved performance over state-of-the art methods.

Variational Bayesian cardinalized probability hypothesis density filter for robust underwater multi-target direction-of-arrival tracking with uncertain measurement noise

“…For the scenario of moving targets, the traditional DOA estimation methods cut the measurement of the output of sonar array signal into small periods in time to process, which ignores the kinematic characteristics of the targets [6][7][8]. The DOA tracking methods not only use the measurement information but also rely on the kinematic characteristics of the underwater targets [9][10][11][12][13][14][15][16][17][18]. Therefore, by not only depending on the current measurements but also utilizing the prior kinematic information of an unknown underwater target, the DOA tracking methods can provide more robust and accurate results than the traditional DOA estimation methods.…”

“…Similar to the PHD filter, the CPHD filter also has no closed-form solution in general and has to be implemented on the basis of the SMC or GM model [30]. Considering that the computational complexity of the RFS-based method is lower than that of the data association method, researchers have proposed many multi-target DOA tracking methods based on the PHD and CPHD filters [13][14][15][16]. These methods use the output of the array signal as measurements with the non-linear measurement model, thus implementing them based on SMC.…”