Table 2.
Summary of papers with the classification of cognitive states.
| Ref. | Year | Cognitive States | Best Performing Models | No. of Subjects (Female/ Male) | Stimulus | Data |
|---|---|---|---|---|---|---|
| [39] | 2016 | Confusion | RF, sensitivity 0.61, specificity 0.926 | 136 (75F/61M) | Data visualization software | Self-report, ET (with pupillometry), clicks |
| [36] | 2016 | Mental workload, attention | LDA, accuracy: 92% mental workload and 86% attention | 12 (3F/9M) | Virtual maze game | Self-report, EEG, keyboard, and touch behavior |
| [22] | 2016 | Mental stress | RF, click-level user-dependent f1-score 0.66; logistic classifier, session-level user-independent f1-score 0.79 | 20 (7F/13M) | Arithmetic questions software | ET (from video), clicks |
| [35] | 2016 | Engagement | SVM, f1-score 0.82 | 10 (3F/7M), 10 (3F/7M), 130 (34F/96M) | Cell phone usage | 1st and 2nd studies: EEG and usage logs; 3rd study: usage logs, context, and demographic data |
| [34] | 2018 | Mental workload | MLP, accuracy 93.7% | 61 (19F/42M) | Website browsing | EDA, Photoplethysmography (PPG), temperature, ECG, EEG, ET (with pupillometry) |
| [37] | 2019 | Confusion | RF, accuracy range 72.6–99.1% | 29 (14F/15M) | Personal data sheets | ET, age, gender |
| [40] | 2019 | Engagement (as a basis for interest detection) | kNN (k-Nearest Neighbors), average accuracy 80.3% | 4 (2F/2M) | Videos | Self-report, EEG |