2D DOA estimation with propagator method for correlated sources under unknown symmetric Toeplitz noise

The paper employs the propagator method (PM) to find two-dimensional (2D) direction of arrival angles (DOAs). The proposed algorithm can be applied to the situation when the incident sources are mixed, either with uncorrelated or fully correlated (coherent) sources. Like S. Pasad et al. (see IEEE Trans., vol.ASSP-36, no.5, p.631-41, 1988), we assume a symmetric Toeplitz covariance matrix for the unknown noise model. But the proposed algorithm, when compared with Pasad´s, has two main advantages: it does not require any eigen decomposition to find the DOAs, so that the complexity can be significantly smaller than in Pasad´s method; it requires the number of sensors, N, to be larger than the number of sources, K, i.e., N>K, but Pasad´s method requires N>2K. The proposed method is based on the covariance matrix difference between the average of the forward-backward covariance matrices of the received data and its Hermitian of the backward. This difference is introduced to eliminate the noise components from the array structure. The numerical results verify that the proposed method performs successfully under the environment where Pasad´s method fails, and, furthermore, gives better performance and less computation than the conventional MUSIC.