Table 5 CWM: Cross-validation and testing performance after feature selection (18 features total for e-Glass and 33 for the full-cap configuration) and hyperparameter optimization to reduce the model size (200 trees for e-Glass and 400 for the full-cap). We use 19 EEG channels on the full-cap evaluation.
From: EEG glasses for real-time brain electrical activity monitoring
System Config. | CV (%) | Testing (%) | Model Size (KB) | ||||
|---|---|---|---|---|---|---|---|
Acc | Gmean | Sens | Spec | Acc | Gmean | ||
e-Glass | \(75.8 \pm 4.7\) | \(75.1\pm 5.0\) | 82.9 | 66.1 | 74.5 | 74.0 | 113.5 |
Full-Cap | \(82.0 \pm 5.2\) | \(81.3\pm 5.8\) | 73.3 | 87.7 | 80.5 | 80.2 | 645.7 |
