Sparse feature auto-combination deep network for video action recognition | IEEE Conference Publication | IEEE Xplore