with Unknown. Arbitrary Covariance Matrices
J.P. Reilly, Kon Max Wong, Park M. Reilly · 1989
A new method for estimating directions of arrival of plane waves onto arrays of sensors is proposed. The method is particularly well-suited to the case where the background noise field is non-isotropic, with arbitrary covariance matrix. The joint posterior probability density function of the signal parameters and the noise covariance matrix C is formed, and then the dependence on C is integrated out, after a suitable noninformative prior p(C) is defined. The resulting estimator structure is then modified to substantially reduce the computational requirements. Significantly improved performance over the MUSIC algorithm, particularly with regard to threshold. is observed. The field of array processing has been an active area of signal prwessing throughout the 1970's and 80's. The object of this paper has to do with estimating the directions of arrival (DOA's) of plane waves onto arrays of sensors, in the presence of non-isotropic Gaussian noise, having unknown and arbitrary covariance matrices. There have been several algorithms proposed for dealing with this issue, the most well-known of which is no doubt the ESPRIT algorithm [11. The ESPRIT algorithm attempts to deal with the effects of correlated noise by assuming the noise field in which the array is immersed is invariant under a spatial displacement of the array, whereas the desired signal is not. However, as indicated by Speisar [21. there are shortcomings associated with ESPRIT, in that there can be grating lobes introduced if the two subarrays are too widely separated, and also the method is highly dependent on the spatial invariance of the noise field.