Abstract: Based on the idea of jointly sparse matrix recovering, a new DOA estimation method for narrowband signals is proposed in this paper. By applying vectorization on the computed array covariance matrix in each frame, a pseudo data matrix is formed. Then, the jointly sparse signal model is generated by constructing the over determinate dictionary of array direction matrix. Afterwards, the optimal solution to the jointly sparse matrix is solved by using joint l_2,0 approximation approach, and the DOA estimation is obtained accordingly. Since the vecorization operation extends the array aperture, the proposed method can resolve more signals than the number of arrays, be suitable to the short stationary or stationary narrow band signals, and need not estimate the number of signals in advance. Simulation results validate the effectiveness of the proposed method.