Covariance-tuned EKF resampling based particle filter

Abstract

Signal processing in complex environments is a challenging issue in many applications. Recent studies show that variants of particle filters achieve significant performance in positioning and tracking applications under complex environments. In this article, a covariance-tuned EKF resampling based particle filter (CTEKF-PF) is proposed. In CTEKF-PF, a double-resampling of prior particles is introduced during the resampling stage, in which the latest observed measurements are effectively integrated into each already resampled particle, rather than into the sampled particle, using a covariance-tuned Extended Kalman Filter (EKF). Thereby facilitating the motion of resampled particles towards the high likelihood regions. Experimental results from the GPS receiver position estimation application show improved estimation accuracy, reduced computational load, and reduced computation time for the proposed CTEKF-PF compared to Least-squares particle filter (LSPF) and standard particle filter.

Keywords

• ekf
• particle filter
• resampling
• signal processing