Arrival angle estimation of correlated sources with unknown, spatially uncorrelated and nonstationary noise

Most methods for direction of arrival angle (DOA) estimation with array processing assume spatially white Gaussian noise of a known covariance matrix. However, the additive sensor noise in a practical situation may not be known. The paper presents a DOA estimation scheme for correlated sources including uncorrelated ones under an unknown spatially nonstationary white noise environment, i.e., the noise has a diagonal covariance matrix whose diagonal elements have unequal noise power. . The method is applied to DOA estimation for coherent, noncoherent, and mixtures of these narrow band signals. The proposed method employs the difference between the forward backward averaging covariance matrix and its transform to eliminate the noise effects which can improve the performance of DOA estimation. Numerical results show that our proposed algorithm gives accurate estimation even if the sources are correlated, whereas the method of A. Moghaddamjoo (see IEEE Trans., vol.SP-39, p.219-21, 1991) fails to resolve the DOAs when the sources are coherent.