Elevating Skeleton-Based Action Recognition with Efficient Multi-Modality Self-Supervision | IEEE Conference Publication | IEEE Xplore