DOA Estimation Based on RBFNN for Minimum Redundancy Linear Array (MRLA)

The mutual coupling matrix (MCM) of uniform linear array (ULA) can be modeled as a banded symmetric Toeplitz matrix. However, the MCM of MRLA is a symmetric but not a Toeplitz matrix. Many conventional calibration algorithms based on the banded symmetric Toeplitz matrix for ULA can´t be applied to MRLA. Based on RBF neural network, a DOA estimation algorithm in the presence of mutual coupling for MRLA is proposed in this paper. Since the array correlation matrix is symmetry and there is no DOA information on its diagonal, the upper triangular half of the matrix is extracted as the input vectors. This method not only reduces the dimension of the input vectors but also present a modified preprocessing scheme to handle the problem at the endfire angles of the array. Simulation results demonstrate the proposed algorithm is efficient and valid.