The stochastic Cramér-Rao bound for source localization and medium tomography using vector sensors

A direct version for the stochastic Cramér-Rao bound (CRB) for parameters of Gaussian signals with additive Gaussian noise is introduced. The formulation applies to passive and active radars/sonars/seismics/structures with vector observations from multiple sources. These sensors include pressure, vector velocity, and/or acceleration sensors for ocean and structural acoustics, seismometers, polarized receivers for electromagnetics, and vector current meters for oceanography. The observations may contain partially coherent signals such as multipath. The parameters represent (i) signal localization or (ii) tomographic ones. As such, their embedding is very general using a Green's function vector and is not limited to direction of arrival problems. This formulation leads to simplified expressions for the stochastic CRB using just three quadratic forms involving just the Green's function and its derivatives with the inverse of the noise matrix for the norm. The number of the parameters sets the dimensions of these quadratic forms, so performance studies over the parameter space can be done with much smaller matrices as the noise covariance is inverted just once. The formulas are applied to vector sensors in jamming with both analytical and numerical results. The results are also compared to often cited papers on the CRB.