Putting poses on manifold for action recognition | IEEE Conference Publication | IEEE Xplore