Improving the Robustness of Hand Gesture Recognition with One-shot-learning Features of Static HOG and SVM by Using Kinect V2 Method

The innovative technologies like dynamic action and hand gesture recognition has improved considerably in the past few years. However, in this paper we are implementing the concept of one short learning with the major features like histogram orientation of gradients and support vector machine for hand gesture recognitions. Here the major utilities of RGB and depth images are pre-recorded by using sensors and 3-D camera of Kinect V2. The combined appearance of the depth images and RGB are very important for motion descriptors. These are used for gesture recognition and temporal segmentation. To measure the differences in Histograms we capture the relationships of cross-bins by using the Quadratic-Chi distance family. Here the proposed methods of HOG descriptors has the ability to remove all unwanted frames from the captured images by using Kinect V2 and it can also trim the video frames by using new algorithms using a combination method in HOG, descriptors together with Support Vector Machines technique to deliver good results in the hand gesture recognition technology.