Article

R.Keerthana, PSNA College of Engineering and Technology; Dr.S.Mythili ,

Hand gesture recognition system’s performance and accuracy are increased by one of the pre-processing techniques. i.e., gesture segmentation. In some real time applications, due to low accuracy in gesture matching process used to lead unexpected response. Earlier days, contour model based hand gesture recognition, accelerometer based hand gesture recognition and scene classifications based on the contextual semantic information of an image are the recognition techniques used in hand gesture recognition system. Here, these techniques arise many problems. They are lagging in focus on detecting hands, some of them directly use marker based motion capture devices, it cannot provide a rotation and scale invariance, it does not provide the hand parts segmentation, and it capture the same gestures and manually labelled them for each subject. These draw backs can be overcome by implementing the new image segmentation algorithm. The proposed algorithm uses binaryzation and K Nearest Neighbor (KNN) algorithms. The binaryzation used for back ground subtraction. And KNN classifier for classifying the hand features. Here, a sample image is processed with the graph cut algorithm. And then the parameter is observed. The parameter has been compared with parameter observed in CSS (Curvature Scale Space) algorithm. Finally, the result shows that graph cut algorithm gave the better accuracy than CSS algorithm. In real-time application, the obtained contour descriptor will be matched with contour descriptor in the data base. Once it matched, then it will trigger an event to drive an application.