Wideband grid-free direction of arrival estimation based on covariance matrix fitting

Wideband direction of arrival (DOA) estimation is an important research subject in sonar systems. In this paper, a coherent signal-subspace based modified sparse and parameter approach (C-MSPA) is proposed to achieve high-accuracy and high spatial azimuth resolution wideband DOA estimation. The proposed algorithm uses the focusing matrix to project the sample covariance matrix in each frequency bin onto the focusing frequency bin. After focusing, the covariance matrix fitting criterion is modified by the frequency-selective Vandermonde decomposition to reconstruct the covariance matrix where the DOAs encoded are restricted in the focusing sectors. Finally, the DOAs are estimated by performing the Vandermonde decomposition on the reconstructed covariance matrix. The proposed algorithm does not need to select regularization parameters and avoids the basis mismatch issue. The results of simulation and lake experiment show that the proposed method achieves high resolution and improves the DOA estimation accuracy.