Research Article
A Hierarchical View Pooling Network for Multichannel Surface Electromyography-Based Gesture Recognition
Table 3
Gesture recognition accuracy achieved by HVPN, VS-L1VP, VS-L2VP, and VS-ONLY on five databases.
| | Database | Evaluation methodology | HVPN | VS-L1VP | VS-L2VP | VS-ONLY |
| | NinaProDB1 | Intrasubject | 86.8% | 86.5% | 86.2% | 85.8% | | NinaProDB2 | Intrasubject | 84.4% | 84.1% | 83.9% | 83.4% | | NinaProDB3 | Intrasubject | 68.2% | 67.7% | 67.5% | 67.2% | | NinaProDB4 | Intrasubject | 70.8% | 69.9% | 69.7% | 68.5% | | NinaProDB5 | Intrasubject | 88.6% | 87.9% | 88.3% | 87.2% |
| | NinaProDB1 | LOSOCV | 83.1% | 82.6% | 82.5% | 81.9% | | NinaProDB2 | LOSOCV | 79.0% | 78.7% | 78.7% | 78.1% | | NinaProDB3 | LOSOCV | 65.6% | 65.5% | 65.0% | 64.7% | | NinaProDB4 | LOSOCV | 67.0% | 66.3% | 65.7% | 65.2% | | NinaProDB5 | LOSOCV | 87.1% | 86.2% | 86.5% | 84.7% |
|
|
Results in bold entries indicate best performance.
|
