The Research Of Gesture Recognition Based On RGB-D Depth Information

In recent years, with the rapid development of science and technology, human-computer interaction also continued to adjust. Traditional methods for interaction, the use of a keyboard,mouse and other hardware, the gesture is the most direct way to interact within the most effective way, and most conform to people’s habits. To this end, a gesture to identify with a broad scope of application of skills, and this is the most talked about a range of topics within the human-computer interaction. Gesture recognition method based on monocular camera is highly vulnerable to illumination changes and kind of color of skin, making the recognition effect is often short of expectations. In this paper, the design of the latest Microsoft Kinect camera lens, can significantly reduce the environmental factors interference gesture identification. This study was based on RGB data source depth information, proposed and implemented the static and dynamic gestures gesture recognition under complex background track recognition, the work is as follows:1. First Custom nine kinds of static gestures. Then the hand region is extracted from complex background via depth image and Skin color model. Using Moore neighborhood algorithm to obtain gesture contour from gestures binary figure, using fingertips threshold test positioning fingertips, and determining the hand index fingers while calculating the angle.Then appearance features are integrated to build the decision tree model which based on hands and largest index angle of the fingertips for hand gesture recognition.Nine common gestures with complex background were tested in the system. Experiments show that the method can implement efficiently and has a strong robustness.2. Custom dynamic gesture template. Using the distance variation method get the palm coordinates from the gesture area is extracted for the data. By the continuous static method to obtain the start and end points, and then get the gesture trajectory models of dynamic gesture dynamic. Extraction of trace X, Y coordinate, calculation of the angle between the adjacent points value as a gesture features. Using 8 direction chain code get the feature sequence.Dynamic time warping(DTW) algorithm is used for training for the gesture recognition, the recognition of the custom dynamic hand gestures, the average recognition rate of 96.42%.