Covariance Reconstruction for Coherent Signal DOA Estimation With Limited Snapshots

The performance of traditional coherent signal direction of arrival (DOA) estimation algorithms deteriorates significantly in limited snapshot scenarios. This is mainly because when the number of snapshots is insufficient, there exists a weak correlation between different snapshots. To address this challenge, this article proposes a covariance matrix reconstruction algorithm tailored for coherent signal environments. The proposed method leverages the tensor t-product to model signal correlations across temporal and spatial dimensions effectively. By incorporating an oblique projection operator founded on the correlation subspace, a refined covariance matrix is reconstructed. This reconstruction process effectively mitigates noise interference and eradicates intersignal correlation, thereby enhancing the accuracy of covariance estimation even with a limited number of snapshots. Finally, the DOAs are accurately estimated using a subspace-based method. Simulation results demonstrate that the proposed algorithm significantly improves estimation performance under conditions of low SNR and limited snapshot conditions, providing better resolution when the target DOAs are closely spaced.