Research On Vision Recognition Of Hand Gesture Based On ARM Embedded System

With the development of human-computer interaction techniques, vision-based recognition of hand gesture has become one hot point of researches on natural human-computer interaction, which can be used for human-computer interaction in virtual environment, intelligent control, as well as sign language recognition and snatch by robot arm. However, at present, the embedded technology get good advanced and widely used in nearly every intelligent device, like intelligent home appliances, apparatus and instrument, industry control, intelligent construction, palmtop computer and so on. It undoubtedly brings a more natural, convenient and intelligent human-computer interaction mode with the vision-based recognition of hand gesture is used in embedded devices. Therefore, the research on vision-based recognition of hand gesture based on embedded system is not only of great theoretical significance, but also of wide application prospect and high practical value.Based on the analysis and study of basic theories and algorithms on topic in the video capture, preprocessing to gesture image, gesture dynamic segmentation with complex background and the methods of feature extraction under ARM Linux system,the control system with embedded vision-based recognition of hand gesture is designed and realized based on technology of ARM11and Zigbee. It has the functions for controlling the attributes of video source and the format of image captured, setting the rates of frame, as well as real-time hand gesture recognition and non-contact hand gesture control. Finally, the each function of system is tested and the tested result is brought out in this thesis.The main work consists of three parts:(1) The process of video capture under embedded Linux system is detailedly analysed after basing on Linux video capture mechanism and combining with features of the embedded. Then the model of image capture under embedded Linux system is implemented.(2) On the method of hand gesture segmentation, the method of combining with skin color information and motion information is presented. A good segmentation of hand gesture under complicated background is realized through combining the adaptive threshold segmentation of Cr component of skin color and ViBe foreground segmentation of hand gesture. In addition, more calculations are greatly decreased on follow-up feature extraction of hand gesture.(3) On the implementation of embedded hand gesture recognition algorithm, a set of hand gesture control system based on Zigbee technology is developed with OpenCV2.3.1in order to testify the feasibility and the real-time performance. Aiming at the embedded transplant problem of Highgui section which depends on GTK2.0+libraries, the Highgui section of OpenCV is replaced by the Qt/QtEmbedded C++Ui libraries which have fine compatibility and cross platform capability.At last, some tests turn out that the vision recognition of hand gesture based on ARM embedded system has good performance and wide application prospect.