A real-time hand gesture recognition algorithm for an embedded system

With the development of technology, intelligent robots will play an important role in our daily life. More and more intelligent robots based on embedded systems are used for security, detection, service, etc. Interaction with human is an important part of intelligent robots. Hand gesture is a convenient and fast method for human-robot interactions. In this paper, we propose a method which is suitable to detect hand area and recognize hand gestures for video stream on an embedded system. In order to make the recognition process run in real-time, the proposed method is based on Gaussian Mixture Model (GMM) which not only helps to build the skin model but also helps to classify the gestures. An embedded system with our algorithm is almost the same with a PC(Intel Core i3 500, 4G DDR3) real-time performance. At the same time, the average recognition rate is more than 75%. Both the embedded system and the real time hand gesture recognition algorithm is used to control an intelligent robot.