Table 2.

The performance of the latest state-of-the-art 3D skeleton-based methods on NTU-RGB+D 120 dataset

NTU-RGB+D 120 dataset
Rank	Paper	Year	Accuracy (C-Subject)	Accuracy (C-Setup)	Method
1	Wang et al. [55]	2023	92.0	93.8	Two-stream Transformer
2	Xu et al. [146]	2023	90.7	91.8	Language knowledge-assisted
3	Zhou et al. [56]	2022	89.9	91.3	Transformer
4	Duan et al. [134]	2022	89.6	91.3	Dynamic group GCN
5	Chen et al. [136]	2021	88.9	90.6	Topology refinement GCN
6	Chen et al. [147]	2021	88.2	89.3	Spatial-temporal GCN
7	Liu et al. [103]	2020	86.9	88.4	Disentangling and unifying GCN
8	Cheng et al. [148]	2020	85.9	87.6	Shift GCN
9	Caetano et al. [90]	2019	67.9	62.8	Tree structure + CNN
10	Caetano et al. [89]	2019	67.7	66.9	SkeleMotion
11	Liu et al. [149]	2018	64.6	66.9	Body pose evolution map
12	Ke et al. [150]	2018	62.2	61.8	Multitask CNN with RotClips
13	Liu et al. [151]	2017	61.2	63.3	Two-stream attention LSTM
14	Liu et al. [12]	2017	60.3	63.2	Skeleton visualization (single stream)
15	Jun et al. [152]	2019	59.9	62.4	Online+Dilated CNN
16	Ke et al. [153]	2017	58.4	57.9	Multitask learning CNN
17	Jun et al. [82]	2017	58.3	59.2	Global context-aware attention LSTM
18	Jun et al. [76]	2016	55.7	57.9	Spatiotemporal LSTM