Single person hand gesture recognition using support vector machine

Abstract

A simple and easy-to-use system is designed for recognition of single person gestures using their movement of hands while expressing feelings. Here Kinect sensor is employed to generate the 20 body joint coordinates for a person. The gestures are comprised of six single hand gestures as well as four double hand gestures. In this paper, the authors have processed only one (right- or left-hand joint) coordinate while sculpting single hand gestures, whereas two (both right- and left-hand joints) coordinates are taken into account when double hand gestures are considered for each frame using the Kinect sensor. Once the coordinates are obtained, then normalization is carried out based on the coordinate of the hip centre for the first frame. The recognition procedure is based on support vector machine and produces high accuracy rate of 94.3%.