A robust gesture recognition based on depth data

In this paper, we propose a novel method for gesture recognition using depth data captured by Microsoft Kinect sensor. Conventionally, the features which have been used for gesture recognition are divided into two parts, hand shape and arm movement. In conventional methods, only two-dimensional hand features are used because human´s hand consists of the multiple joint structure. Furthermore, conventional arm movement feature are influenced by environmental changing, such as individual differences in body size, camera position and so on. Therefore, to assist in the recognition, a method of feature extraction is proposed, which involves the hand shape with 3D feature and the arm movement with angle between joints of body. In order to show effectiveness of the proposed method, performance for gesture recognition is compared with conventional methods using Japanese language.