Gesture Recognition And Application Based On Kinect Skeleton Information

With the rapid development of computer and information exchange technology, a lot of changes take place in communication and interaction between human and computer. From one dimensional information in the form of text message, to two dimensional information in the form of graphics message, to three dimensional information incorporating graphics with depth, the transformation promotes development of human computer interaction(HCI) technology. Natural User Interface(NUI) has become a trend of HCI, which is to communicate with computer with the most primitive form of communication, such as language, human movement, expression, etc. Gesture recognition is a necessary technology in HCI because it is flexible, intuitive and information-rich. Kinect can capture RGB color image, depth image and skeleton image, which provides a good condition for gesture recognition.This dissertation proposes a high-performance gesture recognition method based on Kinect skeleton information, and develop a motion-sensing browser system which can manipulate computer by gesture when surf the Internet. The main work is as follows:1) Briefly introduced the research status at home and abroad in robot control, medical rehabilitation and human computer interaction.2) Improved traditional DTW method by constraining warping path, which achieves high performance, In order to improve gesture recognition accuracy, we assign different weights to different gesture features according to the contribution of different gesture feature. Combining with unique characteristic of Kinect, we set different weights to different gesture features secondly, which makes the gesture recognition result is more reliable.3) A motion-sensing browser system was designed and implemented. The whole system is running smoothly, has obtained the good user experience. In addition, it also can be extend to realize more abundant function.