DOA estimation using SNR-aware joint sparse representation

Direction of arrival estimation is one of the most studied problems in array signal processing. While subspace-based methods achieve high accuracy at higher SNR levels, their accuracy deteriorates at lower SNR. Furthermore, they lack accuracy when there is correlation between sources, and when the number of snapshots is limited. Compressed sensing or sparse approximation-based methods were proposed to improve accuracy in such conditions. However, they suffer from higher computational complexity. Furthermore, their estimation accuracy at high SNR is not as high as subspace-based methods. In this work, we propose a joint sparse representation-based algorithm which combines the power of subspace methods and sparse representation methods. We propose a measure of SNR based on the eigenvalues of the covariance matrix of the sensor measurements. The algorithm utilizes the SNR measure and noise subspace information to improve estimation accuracy. Simulations show that our algorithm outperforms other related algorithms, even at lower SNR and even if the sources are correlated and the number of snapshots is limited, faster than related sparse algorithms.