Desgin And Implement On The Industrial Robot Operation Control Method Based On Hand Gesture Recognition

With the development of automation technology, more and more robots appear in our life and production activity. The interaction between human and robot become more and more frequently, and therefore how to deal with the robot in a better way, is the key of the study of human-computer interaction technology. In recent years, the direction of the development of human-computer interaction technology is humanization and simplification. Hand gesture is a kind of human languages which contains maximum amount of information of all. It expresses information in a more natural, friendly, and effective way. Accordingly, hand gesture, as an important way of human-computer interaction, will occupies a higher position in the future.Nowadays, the human-computer interaction mode is rather mechanized. In the past, the operator often need to use the mouse or teaching box to control the robot. Even use digital data glove, it is not natural and convenient since the expensive glove is needed. Faced with this situation, now we tent to use the bare hand gesture to control industrial robot creatively. While the previous bare hand gesture recognition based on computer vision, which is vulnerable to the effects of light conditions and the complex background environment, we have abandoned the use of conventional color camera as a hand gesture image capture devices. We use Kinect sensor to collect depth information in order to exclude the effects of light conditions and the complex background environment. It is believed that the study of hand gesture recognition has great significance for achieving a natural interaction between human and industrial robots.This paper proposes a hand gesture control system of using Kinect sensor as a hand gesture acquisition equipment to control robots. In the system, using the method of combining the depth threshold method and the hand bones point to extract the hand gesture images accurately from the data obtained by Kinect infrared camera. The operator does not need to wear any equipment, while it has no operator location requirement and no background environmental requirement.Then combining the use of deep autoencoder network and Softmax classification as the method of hand gesture recognition. And the recognition experiment results satisfactorily. Finally, we use the client program and the server program designed by this article to have simulation and remote experiments of robot-gesture control. Both bi-handed and single-handed gesture experiments have achieved good results, proving the feasibility and availability of hand gesture control among all the industrial robot controling system.