Research On Moving Hand Tracking Based On Behavior Analysis

In HCI (Human Computer Interaction, HCI) research field, the interaction between users and virtual objects in virtual environment has been the current hot topics at home and abroad. In human computer interaction, gesture interaction is a new human-computer interaction. In the gesture interaction, the moving hand tracking as the key issues in the study, but because of the specificity of nature hand, it becomes difficult to track them. To solve the problems, there has been in-depth research in recent years, but but there is still not fully resolved. In this paper, we combine with cognitive psychology view of vision-based hand tracking, which has been the great significance for the advancement of the subject.Hand is a special articulated object, analysis the physical characteristics of it, it is composed of the palm and fingers. 5 fingers are both independently and closely linked, they are interact and constraint each other. Hand gesture has characteristics of more meanings and difference in time and space, and it is also a joint chain object in high degree which makes it hard in tracking. In hand tracking it is not only to ensure a higher accuracy but also in good real-time, which is an extremely challenging in research. This paper is supported by National Natural Science Foundation of China (No.60773109), National Natural Science Foundation of China (No.60973093), Natural Science Foundation of Shandong Province (Y2007G39), Natural Science Foundation for Distinguished Youth Scholar of Shandong Province (No.JQ200820), Key Project of Natural Science Foundation of Shandong Province (2006G03), Science and Technology Plan of Shandong Province Education Department (J07YJ18).In this paper, we introduce a new element in gesture interaction--cognitive psychology. We also use nature hand in virtual assemble for gesture interaction, and present a method based on behavior analysis in moving hand tracking. Under a single camera, this paper is discussed from gesture segmentation and recognition, gesture block, three-dimensional automatic initialization hand, hand tracking which based on behavioral analysis and so on. The main study are as follows: (1) We built two virtual assemble platforms, one is based on camera and the other is based on data glove. The research in this paper is discussed under the two platforms. We could acquire real-time gesture frame from the camera plat, and the data glove plat is used to get the changement data from the gesture interaction which are used to built cognitive model.(2) Gesture segmentation and gesture recognition is the key point in hand tracking. In this paper, combined with the characteristics of the hand, we establish two models to segment the hand from the complex background. Firstly, we use the RGB color space of the hand to establish the first preliminary segmentation model. Secondly, combined with the gray,lightless and the characteristics of the hand to establish the second segmentation model. Lastly, combine with density distribution of gestures and geometric moments, we propose GDF to discribe spatial distribution of gestures. Use this method we could identify the gestures from the video.(3) Address the problem of hand block itself and three-dimensional problem of automatic initialization. This paper judges the block among the fingers and the block between fingers and the palm, then eliminate the interference of projection. Meanwhile, we propose a method of three-dimensional manual initialization to resolve the problems of worse accuracy, human capacity and long-time consumption in the traditional initial method.(4) This paper we propose the moving hand tracking method which based on behavior analysis. We introduce new element-cognitive psychology, to the degree of cognitive psychology, we analysis the change of the hand parameters in the virtual platform from the perspective of time model, distance model, hand successive deformation, hand mutation, move speed, the size of the object, then built the corresponding cognitive model, and use the model into the forecast and sampling in the method of particle filter for hand tracking, and to sampling regularly from the perspective of the channel probability, so improve the robustness, real-time, accuracy of hand-gesture tracking.