Towards fast, view-invariant human action recognition | IEEE Conference Publication | IEEE Xplore