In the scenario of on-orbit servicing, the complex space environment and the strict limitation of spacesuits made the manual control of manipulator extremely inconvenient. Current remote-control methods such as voice control, are commonly affected by the signal ratio, especially in the case of solar wind. Therefore, in this paper, we proposed a gesture-based remote-control system for real-time manipulation of the space manipulator in order to accomplish the onboard tasks. Gesture-based control strategy is more friendly on human-machine interaction than voice control or manual control. The proposed method: utilized gesture recognition technique to perform scheduled tasks according to the preset gestures. For preset missions, specific gestures are recognized and matched through KNN algorithm, and then manipulator is driven to execute the scheduled tasks. To validate the efficiency of the proposed method, extensive experiments are carried out on simulation platforms and compared with the traditional method of gesture recognition using rotation angle and gesture recognition using CNN. Results demonstrated that our method outperforms the other two solutions in terms of accuracy and speed.