Passive Localization of Mixed Near-Field and Far-Field Sources Without Eigendecomposition via Uniform Circular Array

Abstract

In this paper, we employ the geometry of uniform circular array to achieve classification and localization of mixed near-field and far-field sources. Considering that the eigendecomposition of the covariance matrix requires high computational cost, we develop the propagator method to obtain the noise subspace and reduce complexity. Firstly, since the direction parameters of far-field sources at centrosymmetry sensors hold a conjugate structure while the covariance matrix of near-field sources holds a Hermitian structure, we exploit the covariance differencing approach to extract the pure near-field sources from mixed sources. Then, we improve the ESPRIT-like method and one-dimensional MUSIC method to determine the 2-D direction-of-arrival (DOA) and range of near-field sources, respectively. Finally, by calculating the noise power of mixed sources, we utilize the oblique projection approach to extract the pure far-field sources and exploit the 2-D MUSIC method to determine the 2-D DOA of far-field sources. Simulations demonstrate that the proposed algorithm can avoid the pseudo-peaks in the 2-D DOA spatial spectrum of far-field sources and provide the satisfactory performance of mixed source localization.