Moving Target Tracking with Missing Data in 2-D or Higher Dimension

Subspace tracking is the problem to estimate and track a low-dimensional subspace with partially observed data vectors, which has lots of applications on radar, sonar, wireless communication, and surveillance video image processing, etc. Due to the partially observed measurements with missing entries and its online manner, in the subspace tracking problem, it is highly desired to have high performance in recovering missing entries and tracking the low-dimensional space at the same time. In this paper, we consider moving target tracking problem in Direction-of-Arrivals (DoA) with missing entries in 2-D or higher dimension, which can be understood as one of the subspace tracking problems, where the signals are obtained from the multiple of uniform array of sensors organized in 2-D or higher dimension. Especially, we propose to use structural information by expanding measurement space from measurement data in a matrix (or measurement data in a tensor) to a low-rank folded Hankel matrix in higher dimension in order to improve the performance in the recovery of missing entries and obtain improved resolution performance. Through numerical experiments, we demonstrate that expanding the measurement subspace to the folded Hankel matrix form can play a significant role in improving moving target tracking performance with extremely missing measurements.