Research Article
Sports Competition Assistant System Based on Fuzzy Big Data and Health Exercise Recognition Algorithm
Table 6
Experimental results on MSR Daily Activity 3D.
| Approach | Method | Accuracy (%) |
| Depth | Only LOP features | 66.25 | Only LOP features | 42.5 | DSTIP (DCSF) | 83.6 | Unsupervised training (LSTM) | 86.9 |
| Skeleton | Only joint position features | 65.63 | MSD-HMM + skeleton | 83.3 |
| Hybrid | Actionlet ensemble | 85.75 | DSTIP (skeleton + DCSF) | 88.2 | MMMP (skeleton + LOP) | 88.13 | Our method | 89.38 |
|
|