Table 2 Classification Performance for BCIC IV 2a subjects 1-9. Comparison of EEGEncoder, ACTNet, TCNetFusion, EEGTCNet, and D-ATCNet Models in Terms of Accuracy and Kappa Coefficient.
From: Advancing BCI with a transformer-based model for motor imagery classification
Subject | EEGEncoder | ATCNet32 | TCNetFusion44 | EEGTCNet45 | D-ATCNet46 | |||||
|---|---|---|---|---|---|---|---|---|---|---|
Acc(%) | Kappa | Acc(%) | Kappa | Acc(%) | Kappa | Acc(%) | Kappa | Acc(%) | Kappa | |
1 | 86.46 | 0.82 | 86.11 | 0.82 | 79.17 | 0.72 | 74.31 | 0.66 | 87.5 | 0.83 |
2 | 74.65 | 0.66 | 72.57 | 0.63 | 64.24 | 0.52 | 52.78 | 0.37 | 70.0 | 0.60 |
3 | 96.53 | 0.95 | 93.06 | 0.91 | 88.54 | 0.85 | 88.89 | 0.85 | 94.9 | 0.93 |
4 | 81.94 | 0.76 | 84.03 | 0.79 | 64.93 | 0.53 | 57.99 | 0.44 | 80.5 | 0.74 |
5 | 84.03 | 0.79 | 77.43 | 0.70 | 71.53 | 0.62 | 72.92 | 0.64 | 79.5 | 0.73 |
6 | 77.78 | 0.70 | 73.61 | 0.65 | 55.56 | 0.41 | 43.75 | 0.25 | 74.4 | 0.66 |
7 | 95.83 | 0.94 | 93.40 | 0.91 | 86.81 | 0.82 | 72.57 | 0.63 | 93.2 | 0.91 |
8 | 89.24 | 0.86 | 86.81 | 0.82 | 80.90 | 0.75 | 77.43 | 0.70 | 87.6 | 0.83 |
9 | 91.67 | 0.89 | 90.97 | 0.88 | 80.21 | 0.74 | 74.31 | 0.66 | 89.6 | 0.86 |
Mean | 86.46 | 0.82 | 84.22 | 0.79 | 74.65 | 0.66 | 68.33 | 0.58 | 84.1 | 0.79 |
ITR(bits/min) | 1400.6 |  | 959.0 |  | 1022.5 |  | 860.0 |  | – |  |
