By exploiting the spatial and temporal properties of most communication signals, we propose a forward-backward linear prediction (FBLP) approach for estimating the directions-of-arrival (DOA) of coherent signals impinging on a uniform linear array. In the proposed method, the evaluation of the cyclic array covariance matrix is avoided and the difficulty of choosing the optimal lag parameter is alleviated. As a result, the proposed method has two advantages: the computational load is relatively reduced and the robustness of estimation is significantly improved. It is shown through numerical examples that this approach is superior in resolving closely spaced coherent signals with small length of array data and at relatively low signal-to-noise ratio (SNR).
展开▼