Paper | Subjects (number, age) | Sensors | Stressing procedure | Classification methods | Window length (s) | Results |
---|---|---|---|---|---|---|
Ham 2017 [23] | 6, 28.67 ± 3.3y | PPG (HRV) | Arithmetic stress, environmental stress, both while in Virtual Reality | LDA | 240 s, non-overlapping | 74% (baseline), 81% (mild stress), 82% (severe stress) Accuracy |
Pereira 2017 [19] | 14, 20-26y | ECG (HRV) | Trier Social Stress Test | N/A | 50-220 s, non-overlapping | N/A |
Zangróniz 2018 [24] | 45, 20-28y | PPG (PRV) | IAPS | Decision Tree | 70 s | 82,35% Accuracy |
Lawanont 2019 [10] | 10, 20-26y | Activity tracker (FitBit) | Real-life work shifts | KNN, SVM, Decision Tree | 1 h | 84.1% (decision tree) Accuracy |
Moridani 2020 [20] | 20, 25 ± 5y | ECG (HRV) | Physical activity, arithmetic task, emotional stress (horror movie) | CNN | 300 s (40 s for “mini emotional stress”) | 97.9% (cognitive), 94.4% (emotional) F1 score |
Pourmohammadi 2020 [34] | 34, 25.4 ± 4.2y | EMG, ECG (HRV) | Arithmetic task, Stroop, environmental stressing | SVM | 60 s, non-overlapping | 100% (2 levels), 97.6% (3 levels), 96.2% (4 levels) Accuracy |
Rodriguez-Arce 2020 [29] | 21, 18-21y | GSR, PPG, body temp., breathing | Arithmetic task, electrocutaneous stimulation | KNN, SVM | ~ 50 s (2500 samples per window) | 90% (KNN), 95% (SVM) Accuracy |
Sánchez-Reolid 2020 [27] | 147, 31.4 ± 8.03y (w), 36.3 ± 4.99y (m) | EDA | IAPS | SVM, D-SVM | 1-40 s, overlapping and non-overlapping | 83% (SVM), 92% (D-SVM) F1 score |
Zalabarria 2020 [30] | 42, 22.88 ± 3.1y | ECG (HR), GSR, breathing | 3D puzzle | Fuzzy algorithm | 20 s sliding window | 91.15% (stress), 96.61% (relax) F1 score |
Zubair 2020 [26] | 14, 25-30y | PPG (HRV) | Arithmetic tasks, Stroop | QDA, SVM | 60 s, non-overlapping | 94.33% (SVM), 89.73% (QDA) Accuracy |