Existing Kinect-based rehabilitation training systems,which focus on active rehabilitation training executed by the patients themselves and are seldom applied to robot-aided rehabilitation training,have certain limitations.To solve these problems,a Kinect-based robot-aided upper-limb passive rehabilitation training method was proposed.Firstly,Kinect-based upper-limb skeleton tracking and inverse kinematics computation were conducted.Secondly,the virtual rehabilitation environment based on Linux/QT and Kinect was developed.Using the reference trajectories of the training task demonstrated by the therapist and the impaired limb's real-time motion position feedback,a fuzzy-based passive rehabilitation training controller was developed.Finally,using Kinect and the 4-DOF WAM^TM-compliant manipulator of Barrett Inc.,the experimental system platform was constructed to evaluate the effectiveness of the proposed method.The results show that the proposed method can fully learn the therapist's rehabilitation experiences,make the robot manipulator smoothly stretch the upper-limb of the patient along predefined training task trajectories,and achieve human-robot collaborations between therapist-robot-patient.