High-performance 2D DOA estimation and 3D localization for mixed near/far-field sources using fourth-order spatiotemporal algorithm

Abstract

In this paper, a novel method called fourth-order spatiotemporal algorithm (FOSA) is presented for classification, 2D DOA estimation and 3D localization of mixed near-field and far-field sources using a uniform circular array with a center sensor. By constructing two spatiotemporal cumulant matrices, without range information, and regarding the fourth-order stationary property, the 2D DOA estimation matrix is extracted. By applying the eigenvalue decomposition, the virtual steering matrix is estimated. The sensor in the center of the array is used to extract the valid virtual steering vectors. The 2D DOAs are estimated by a least-squares method and without the use of spectral search. Ranges of near-field sources are estimated by constructing a spatial cumulant matrix and using the 1D MUSIC technique. FOSA provides a reasonable classification of signals types. It can detect near and far field sources possessing the same azimuth or elevation DOAs. Furthermore, it has a much low computational complexity. FOSA prevents array aperture loss without any pairing parameters. The simulation results validate its high performance in terms of estimation accuracy, resolution and the probability of correct classification of the signals types.