Table 3 The accuracy comparison between our trained K-NN model with existing models.
Study | Algorithm | Number of subjects | Number of poses exercises per subject | Accuracy |
|---|---|---|---|---|
Ince et al.42 | K-NN | 10 people | 7 (human poses) | \(86\%\) |
Ince et al.42 | LSTM | 10 people | 7 (human pose) | \(82\%\) |
Ince et al.42 | Radom Forest | 10 people | 7 (human pose) | \(80.6\%\) |
Jalal et al.39 | Multi-fused features; Accumulated HMM | 15 people | 15 (daily routine activities) | \(93.3\%\) |
Li et al.43 | CNN | 40 patients with chronic stroke | 1 (Single leg-stance) | \(84\%\) |
Our model | K-NN | 30 physical therapists | 3 (Lower limb recovery) | \(93.3\%\) |
