Abstract: Space station manipulators are one of the important technology for space stations. For such practical problems as high complexity and difficulty of control, low interaction efficiency, many limitations and large time delay between heave and earth, etc. in space environment a human-machine interaction system for space manipulator teleoperation is designed. First, the gesture recognition algorithm is proposed and improved based on Kinect camera. The images are processed and recognized through skeleton tracking and deep processing and classification of information to provide feedback for the operators. Then the remote control schemes of manipulators are designed combined with simulation software. The micrograuity virtual reality environment is configured. The teleoperation time day control scheme based on the state forecast is proposed combined with model prediction and control. The simulation experiments show that the method utilizes the visual feedback and virtual state prediction of gesture interaction and can guide the operators to control the manipulators to grasp and move the objects under the microgravity. It can realize the teleoperation of space manipulators for human in the loop and can reduce the time delay effectwely.