In this paper, a joint direction of departures (DODs) and direction of arrivals (DOAs) estimation algorithm for bistatic Multiple-Input Multiple-Output (MIMO) radar based on real signal subspace is presented. Firstly, the received signal data can be reconstructed by forward-backward spatial smoothing, and the covariance matrix of the constructed signal data can be transformed into real-valued matrix. Then, the real signal subspace can be achieved by singular value decomposition (SVD) of real covariance matrix, and the DODs and DOAs can be obtained by Unitary ESPRIT. Compare with ESPRIT method, the proposed method achieved a better estimation performance. Furthermore, all of singular-value decomposition (SVD) or eigen-value decomposition (EVD) are in real-valued space, so the computational complexity of the proposed method is lower than ESPRIT. Simulation results show that the proposed algorithm has better estimation performance than other methods. (5 pages)