Research Article
Research on Discriminative Skeleton-Based Action Recognition in Spatiotemporal Fusion and Human-Robot Interaction
Table 1
Ablation study on the benchmark of NTU-RGB + D.
| Methods | CS (%) | CV (%) |
| ST-GCN [8] | 81.5 | 88.3 | 2s-AGCN [18] | 88.6 | 95.2 | ST-GCN + LPM-TAM (ours) | 84.1 | 88.9 | 2s-AGCN + LPM-TAM (ours) | 90.3 | 95.7 | ST-GCN + LPM-CAM (ours) | 84.5 | 89.0 | 2s-AGCN + LPM-CAM (ours) | 90.7 | 95.9 | ST-GCN + PM-STF (ours) | 84.7 | 89.2 | 2s-AGCN + PM-STF (ours) | 91.1 | 95.9 | ST-GCN + PM-STFGCN (ours) | 85.7 | 89.9 | 2s-AGCN + PM-STFGCN (ours) | 91.9 | 96.5 |
|
|