Abstract: The iterative procedure in closed loop is an important characteristic of the cognitive radar system. In the closed loop including the radar system and operation environment, the cognitive radar can understand the surroundings cycle by cycle to select the proper signal processing method. In this paper, we consider subspace signal detection problem for the radar target embedded in the clutter and thermal noise. The cognitive radar target detection method is proposed based on the iterative and closed loop framework for cognitive radar. The maximum likelihood estimate of the coordinates of the target in the subspace is obtained with the received data at beginning of each loop, and then the transmitted waveform for the next cycle is designed based on the estimated target signature. The computer simulation results indicate that the cognitive radar can obtain the information about the operation environment in more effective manner, due to the iterative closed loop detection method, and the detection performance outperforms that of the traditional radar.