Structure-Preserving View-Invariant Skeleton Representation for Action Detection | IEEE Conference Publication | IEEE Xplore