Direction of Arrival Estimation in Low SNR Environment using Two Stages Sparse Reconstruction

Sparse-based reconstruction for direction of arrival estimation (DoA) offers an advantage of small data size compared to the conventional DoA estimation algorithm such as MVDR, MUSIC, or ESPRIT. Sparse-based reconstruction algorithm can even estimated the DoA using one snapshot. Given this advantage, the sparse-based reconstruction algorithms such as $L$1-norm minimization using CVX-programming or greedy algorithm usually suffers in high noise environment (low SNR) which manifest by a lot of false spikes in DoA estimation spectrum. In this paper we proposed two-stages sparse reconstruction method to estimate the DoA to mitigate this problem. In this scheme, DoA is estimated twice using a greedy based algorithm to get a high resolution DoA estimate, and then the false spikes are removed using the L1 - L2 algorithm. Compared to the conventional method, the proposed method has advantage of much smaller data and robust in low noise condition.