Multi-Modal Fusion With Observation Points For Skeleton Action Recognition | IEEE Conference Publication | IEEE Xplore