Research On Dynamic And Dynamic Gesture Recognition Based On Kinect Sensor

The development of computer technology continue to change rapidly speed. Along with the computer technology development, the advance of human-computer interaction technology which is closely related with the always towards more and more humanization and simple directions. Gesture is also an important way of human-computer interaction. As a kind of body language contained most quantity of information, gestures are natural, friendly, efficient and other natural advantages. These inherent characteristics of gesture exactly in line with the "harmonious" about the human and computer interaction development trend, which has been widely concerned. Gesture recognition has a variety of ways, and only a gesture recognition method based on computer vision does not require the user to wear peripherals, this advantage enables the users to use more free. But the gesture recognition based on computer vision can be easily influenced by the light conditions and complex background environment. To obtain depth images by Kinect sensor provides some help for solve these shortcoming, Therefore, in order to improve the human-computer interaction, promoting the application of interactive technology in different industries. Research on gesture recognition technique based on depth information has important significance. To sum up, this paper chooses to research on gesture recognition based on depth information, the main contents are as follows:Firstly, the background and significance of study on hand gesture recognition is analyzed, and gives the method of gesture recognition of different hardware devices are classified, then the advantages and disadvantages of these different gestures recognition method is compared. Then the research status of different kinds of gesture recognition methods is introduced.Secondly, Kinect depth sensor which is the main equipment to collect image information in this paper is introduced. The introduction includes Kinect sensor structure, the core theory, calibration of Kinect process and its application prospect. Then introduce the Open NI drivers using in the experimental process.Thirdly, According to the depth of Kinect image sensor outgoing on human skeletal tracking and tracing the hand, and then through the threshold method will be the background and foreground is separated from the image, and then extract the image processing and rival contour.Finally, the gesture is divided into two parts of the static hand gestures and dynamic hand gestures, and through this feature hand displacement will separate the two gestures. For static hand gesture, the contour extraction of Hu matrix as the recognition feature, and then designs the BP neural network as its learning means of identification, and then use genetic algorithm to improve the BP neural network. GA improved BP neural network can be the recognition algorithm to achieve the static hand gesture recognition. For the dynamic gesture, DTW algorithm can recognize the trajectory characteristics and skeletal joint point sequence matching to achieve recognition function, can identify the page, to enlarge and narrow motion. The recognition method is stable, not affected by the background and light conditions, and can satisfy the dynamic hand gesture recognition under the various conditions need.