Kinect Based Hand Gesture Recognition And Its Applications

Hand gesture is widely-used as an important approach of human-computer interaction. It improves naturality and flexibility of user experience, and enhances productivity, safety of industry and quality of our life to some extent. The applications of hand gesture promote innovation and social progress. Therefore, hand gesture recognition becomes one of the research hot topics during recent years.Vision-based hand gesture recognition faces different kinds of challenges, such as illumination, scene noises and occlusion, etc. Recently, lots of hardware devices have been developed represented by Kinect which can capture synchronized depth data. With these hi-tech devices, we get a broader solution space. Based on studying basic knowledge and key techniques about hand gesture recognition in recent years, this paper proposes a framework of hand gesture recognition with depth information, and then proves its effectiveness through experiments. Subsequently, a hand control system for presentation based on hand gesture recognition is designed and realized.This paper splits hand gesture recognition into two parts:static hand posture recognition and dynamic hand gesture recognition, where the former is the prerequisite to the latter. In static hand posture recognition, the first step is hand segmentation by making full use of depth information. Next, a novel hand tracking method based on rules is provided to realize hand tracking in real-time. Then, do hand feature description and classify static hand postures by using supervised learning method. In dynamic hand gesture recognition, layer-based definition strategy is used which defines dynamic hand gesture as combinations of basic unit of hand gesture, trajectory and other hand gestures. Therefore the semantics of dynamic hand gestures can be inferred by using finite-state machine and depth data. Experiments show that the framework of hand gesture recognition proposed in this paper is accurate, robust and real-time enough. At last, based on the framework, a hand control system for presentation based on hand gesture recognition is designed and realized, which can be an instance of its applications as a reference.