Research On3D Hand Tracking Based On The Relationship Of Hand Variables

Research on3D hand tracking is the fundamental and key project. As a way of HumanComputer Interaction the hand has the advantage of natural, convenience and harmonious,which is better than the mouse, keyboard, data gloves and other equipments. In particular,precise hand tracking can promote the development of many applications such as humancomputer interaction, sign language tracking and recognition, control the robot, and so on. Infact, it will be a great contribution to more harmonious human computer interactionHand is a complex multi-segment body with a high degree of freedom and small size. Asan elastic object, hand has diversity and ambiguity, and it’s position was in three-dimensionalspace. So it is difficult to track the moving hand. During the movement, the self-occlusion ofhand will make the hand reconstruction and recognition becoming more difficult. Therefore,hand tracking based on vision is a challenging research.this paper is supported by National Natural Science Foundation of China(No.60773109):Fused into the research on natural hand particle filter tracking for real-time and robustness, andNational Natural Science Foundation of China(No.60973093): Natural hand sampling methodand its application in the tracking based on cognitive psychology. This paper researches onsome key problems in3D hand tracking and focuses on the interactive analysis of handmovements to establish the corresponding interactive models to guide the tracking, in order toimprove the accuracy and real-time. The following are innovative contributions of this paper:First, the method of removal noise in a complex background image. Firstly, use grid toseparate the whole image. Then, count the pixel number of foreground pixel and backgroundpixel in every grid. Finally, according to the preset threshold to treat each grid, and eliminatethe image of irrelevant information.Second, get hand feature of the palm. Palm size is an important parameter of hand. In thispaper, we use the features of hand to get the maximum inscribed circle, hand position and palmsize. Then, the hand is divided into two parts: palm and fingers, and accord to the proportion ofthe palm and fingers in the whole hand to guide the hand tracking and recognition.Third, the method of hand tracking based on the analysis of the relationship among handvariables. In this paper, we make a large number of experiments on the data glove and position track platform. And use cognitive psychology to Analysis of human cognition and behavior,then divide the state. The key part is to analysis the rule of the joint variables in the process ofgrasping and releasing objects. Then use of Matlab curve fitting toolbox to fit the original dataand establish the mode of joint variables. The experiments show that the proposed algorithmcan track the movement of hand robust and accurately.In this paper we puts forward a new method for3D hand tracking, by analysis the rule ofthe hand joint variables in the process of grasping and releasing objects, and our experimentalresults demonstrate that, compared with other methods, our method is feasible in improvingaccuracy reducing time-cost in our virtual assembly platform.