Research Article
3D Skeletal Human Action Recognition Using a CNN Fusion Model
Table 1
Comparisons with other skeletal action recognition methods on the Florence 3D-Action dataset.
| Approach | Accuracy (%) |
| Multipart bag-of-poses [17] (2013) | 82.0 | Motion trajectories [20] (2015) | 87.0 | Lie group [21] (2014) | 90.9 | Bi-LSTM [22] (2018) | 93.0 | MIMTL [23] (2017) | 95.3 | Optical spectra [14] (2018) | 95.6 | JDM [15] (2017) | 96.1 | Proposed (spatial) | 97.3 | Proposed (temporal) | 95.1 | Proposed (fusion) | 98.2 |
|
|