The aim of this paper is to investigate a new control algorithm of gait rehabilitation robots that simultaneously provides more freedom for the patients and corrects their walking trajectory. The controller utilizes a gait-phase dependent reference trajectory and a gait-phase detection algorithm to determine the desired position and velocity of joints based on of their actual positions and velocities. Moreover, the controller uses two separate control blocks for the correction of the path and the cadence of walking of the patient. Since the reference trajectory is time independent, the patient can change the cadence of his/her walking. Furthermore, the separate control structure enables the controller to provide different levels of freedom and assistive force to be delivered to the patients. The control method has been implemented through ARMan , a gait rehabilitation robot and its effectiveness is evaluated on the walking trajectory of three healthy subjects and one stroke patient. The results of the experiments demonstrate that the proposed control method corrects the gait pattern of the subjects as good as impedance control method. In addition, this method provides more freedom for the patients to walk based on their desired cadence.