A weighted spatial smoothing scheme for high-resolution direction-of-arrival estimation in unknown correlated noise

We first show that any weighted spatial smoothing of subarray covariance matrices with nonzero weights will lead to a decorrelation of the signals, and discuss the consequences of using negative weights. We then propose a weighted spatial smoothing which will eliminate a class of correlated noise frequently discussed in the literature. We next develop a procedure based on weighted spatial smoothing for determining the number of coherent signals in correlated noise. Finally, we present numerical results to demonstrate the potential of our methods