Multiple Source Localization Via Redundancy Exploitation With Geometrically Redundant Planar Sensor Arrays

Abstract

A new scheme for estimating the respective radial directions of multiple radiating sources is introduced. The main contribution is the incorporation of a technique referred to as "Pseudo Forward Backward averaging" which judiciously exploits the inherent redundancy in a planar sensor array exhibiting the ESPRIT structure [2]. ESPRIT is a recently proposed bearing estimation algorithm that works with an array of M matched sensor pairs (2M sensors total) referred to as doublets, grouping one element of each doublet into an X array, and the other into a Y array. The displacement vector, denoted d→, between the two members of a given doublet is assumed to be the same for each doublet and the two members of each doublet are required to possess identical radiation patterns. ESPRIT allows us to estimate the angle between d→ and the unit normal to the i-th arriving wavefront, denoted by r^ i , i.e., d→⋅r^ i . In contrast to spatial spectrum estimation algorithms such as MUSIC and Maximum Entropy which require the plotting and searching of a multi-modal, two dimensional surface, the estimates obtained via ESPRIT are simply found by solving for the generalized eigenvalues of a judiciously formed data pencil. More specifically, we shall employ a refinement of ESPRIT, referred to as PRO-ESPRIT [1], the development of which was predicated on exploiting the inherent redundancy built into the ESPRIT array structure described above.