A shrinkage-based particle filter for tracking with correlated measurements

This paper studies the problem of tracking with wireless sensor networks (WSNs) using received signal strength (RSS) measurements. The log-normal shadowing associated with RSS measurements from a mobile terminal is correlated both in space and time. We propose a particle filter that exploits the temporal and spatial correlation and estimates the covariance matrix of the measurement noise using the shrinkage technique. Simulation results show that using the estimated covariance matrix in the tracking filter improves considerably the filter performance. It is also demonstrated via simulations that the shrinkage-based particle filter exhibits superior performance to the particle filter without shrinkage when limited measurements are available. Results with high accuracy of tracking using the proposed method are presented.