Abstract: Traditional DOA(direction of arrival) estimation algorithms often fail to deal with coherent signals， so a new DOA estimation method with high accuracy based on reconstructed noise subspace is proposed. This method constructs an augmented matrix as a new covariance matrix by using the auto-covariance and cross-covariance information of uniform liner array， and then， the corresponding noise subspace and the eigen value matrix can be obtained through singular value decomposition on the augmented matrix. To obtain more accurate signal vectors， a new noise subspace can be reconstructed by the eigenvectors associated with the new eigen value matrix. Finally， DOA estimation is completed through spectral peak searching. The proposed algorithm doesnot affect the estimation effect of independent signals. Compared with the IMMUSIC(improved multiple signal classification) algorithm， the proposed algorithm has higher estimation accuracy， especially under the conditions of low signal-to-noise ratio and small signal incidence interval. The simulation results show that even for the conditions of low signal-to-noise ratio and low sampling snapshot number， the improved algorithm can effectively estimate the DOA.

Key words: DOA(direction of arrival) estimation; coherent signals; augmented matrix; reconstruction; noise subspace; MUSIC(multiple signal classification) algorithm