Particle-Velocity-Field Difference Smoothing for Coherent Source Localization in Spatially Nonuniform Noise

This communication considers the problem of estimating 2-D directions of arrival (DOAs) of multiple coherent signals under spatially nonuniform noise (spatially inhomogeneous temporary white noise) using an array of vector hydrophones. A novel preprocessing method called particle-velocity-field difference smoothing (PVFDS) is proposed. The key idea underlying the PVFDS is to remove the spatially nonuniform noise by using the matrix difference of pairs of particle-velocity data correlation matrices, and to decorrelate the coherent signals by summing these difference correlation matrices. Unlike most of other existing preprocessing techniques, such as spatial smoothing and forward-backward averaging, the PVFDS processing does not decrease the array aperture. For arbitrary array geometries, the PVFDS can resolve up to four coherent signals, and for centro-symmetric arrays, forward-backward averaging can double this number to eight. Monte Carlo simulations illustrate that the PVFDS-based eigenstructure algorithms can offer better performance than the particle-velocity-field smoothing (PVFS)-based counterparts.