Abstract: SWEDE (Subspace method Without Eigen DEcomposition) algorithm is one kind of DOA (direction of arrival) algorithms. This algorithm doesn’t need eigenvalue decomposition to obtain the signal subspace and noise subspace, which greatly reduce the computation complexity. However the weakness of this algorithm is to reduce the maximum number of the signals detected by the ULA (Uniform linear array). This paper proposes an improved SWEDE algorithm: NC-SWEDE algorithm by the combination of the nature of signals with maximum noncircularity rate and SWEDE algorithm. NC-SWEDE algorithm uses the augmented data model of the non-circular signals with maximum rate, which is equivalent to double the number of array elements to increase the maximum number of sources measured by SWEDE algorithm. Originally this algorithm should resort to 2D search for spatial spectrum due to the phase of the non-circular signals. By getting the extremes of the phase, we can reduce dimension and the computation complexity. Several computer simulations are conducted to compare performance of NC-SWEDE with SWEDE algorithm respectively. Simulation results verify that NC-SWEDE algorithm can improve the resolution performance and the estimation accuracy compared with SWEDE algorithm.