Web Page Controlling Software Design Based On Gesture Recognition Of Kinect

With the rapid development of computer science and technology,human-computer interaction technology has entered a multi-channel,multi-media intelligent human-computer interaction stage. Replacing the mouseand keyboard, it is hoped that computers can understand the feeling of humanand a variety of human action, which will enhance the natural and efficienthuman-computer interaction, and it has become a hot topic in the field ofhuman-computer interaction. Human gesture as an intuitive, natural inputoperation, which has become an important branch of human-computerinteraction area. It has board application prospects and market value.In the research of human-computer interaction using human gesture,Microsoft’s Kinect sensor shows its outstanding advantage with good researchand application value. Thus, this paper is based on Kinect sensor. First, weanalyzed the technical principles of Kinect sensor, then using the Kinect sensorto get the depth image in the angle,through which,we can identify the humanbody in the image.Finally we get the3D coordinate information of the humanskeleton point from the depth image. We studied the static gesture recognition based on depth image and dynamic gesture recognition based on the humanskeletal information. Finally, we combined the dynamic gesture recognition andthe browser web page control technology, which achieved a system that usinghuman gesture control the browser web page. The main researches andconclusions are as follows:1. From the gesture image capture and pro-process, gesture imagesegmentation, gesture feature extraction and the choice of classification tointroduce the static gesture recognition based on Kinect sensor in detail, whichinclude OTSU threshold segment algorithm, Hu distance, the number of fingersextraction algorithm and the SVM which using in this paper.2. In the dynamic gesture recognition, Kinect sensor for obtaining depthinformation is analyzed to obtain the3D coordinates of skeleton points of themain body, six points are selected as the feature reference for hand movement;In order to eliminate the deviation caused by the difference depth position ofeach subject and the different size of each human body, the experimental datahas been centralized and normalized. For the feature acquired for the experiment,we made some adaptions for the DTW algorithm. In order to improve the rate ofrecognition and the identification speed of the system, we using a DTWalgorithm look-up table in the template training and recognizing dynamic handgesture. The experimental result show that: the method has a high identify speedand recognition rate for the dynamic gesture. At the same time, it also has strongrobustness for the complex background and light intensity change. 3. Finally, the dynamic gesture recognition is applied to control the browserweb page. Six common gestures in our daily life have been defined to achievedynamic gestures control the browser web page. The test result shows that thesystem has a high recognition and identify speed for the predefined dynamicgestures. At the same time, it also has strong robustness for the light intensitychange and complex background, which can meet the actual demand.