Table 1.
Muscle activation simulation results evaluated against measured MVC-normalized EMG from healthy subjects.
| Pearson's r | RMSE | |||||
|---|---|---|---|---|---|---|
| Muscle name | SO | Proposed | Improvement | SO | Proposed | Improvement |
| TA | 0.669 | 1.00 | 49.5% | 0.0691 | 0.0617 | 10.7% |
| GAS | 0.469 | 1.00 | 113% | 0.0253 | 0.0167 | 33.9% |
| SOL | 0.604 | 1.00 | 65.6% | 0.0335 | 0.0313 | 6.46% |
| RF | 0.924 | 1.00 | 8.20% | 0.0657 | 0.0274 | 58.3% |
| VAS | 0.849 | 1.00 | 17.8% | 0.0581 | 0.0266 | 54.2% |
| BFL | 0.826 | 1.00 | 21.0% | 0.0500 | 0.0266 | 46.7% |
| BFS | 0.964 | 1.00 | 3.80% | 0.0617 | 0.0226 | 63.4% |
| GMAX | 0.982 | 1.00 | 1.90% | 0.0698 | 0.0551 | 21.0% |
| RA | 0.782 | 1.00 | 28.0% | 0.1005 | 0.0865 | 14.0% |
| ES | 0.778 | 1.00 | 28.5% | 0.0279 | 0.0241 | 13.8% |
With reference to MVC-normalized EMG, an inter-method simulation accuracy comparison was done between the proposed joint torque-based algorithm and SO based on muscle activation. Muscle activation profile conformity and peak amplitude agreement are evaluated and compared by Pearson's correlation (−1 ≤ r ≤ 1) and root-mean-square error (RMSE), respectively, between the proposed and SO algorithms. Within-muscle simulation accuracy improvements by the proposed algorithm are reported as percentages.