Human action recognition based on Kinect and PSO-SVM by representing 3D skeletons as points in lie group | IEEE Conference Publication | IEEE Xplore