Recognition of human gestures represented by depth camera motion sequences

 The method of gesture recognition useful in construction of hand controlled user interfaces is proposed in the paper. To validate the method, a database containing seven selected gestures, performed by six subjects is collected. In acquisition Microsoft Kinect device was used. In the first stage of introduced method, linear dimensionality reduction in respect to depth images of human silhouettes is carried out. Furthermore, the proper recognition is based on Dynamic Time Warping technique. To assess general features of human gestures across subjects, different strategies of dividing captured database into training and testing sets are taken into consideration. The obtained classification results are satisfactory. The proposed method allows to recognized gestures with almost 100% precision in case of training set containing data of classified subjects and with 75% accuracy otherwise.