SUBSPACE FITTING WITH DIVERSELY POLARIZED ANTENNA-ARRAYS

Diversely polarized antenna arrays are widely used in RF applications. The diversity of response provided by such arrays can greatly improve direction finding performance over arrays sensitive to only one polar ization component. For d emitters, directly implementing a multidimens ional estimation algorithm (e.g., maximum likelihood) would require a search for 3d parameters: d directions of arrival (DOA's), and 2d pola rization parameters. In this paper, we present a more efficient soluti on based on the so-called noise subspace fitting (NSF) algorithm. In p articular, we show how to decouple the NSF search into a two-step proc edure, where the DOA's are estimated separately. The polarization para meters are then obtained by solving a linear system of equations. The advantage of this approach is that the search dimension is reduced by a factor of three, and no initial polarization estimate is required. I n addition, the algorithm can be shown to yield asymptotically minimum variance estimates provided no perfectly coherent signals are present . Simulation examples are included to compare the NSF approach with a similar generalization of the MUSIC algorithm.