Table 1.
Summary of Related Work
| Situation | Subjects (#) | Types of Data Used | Devices Used for Data Collection | Prediction Metrics Reported | |
|---|---|---|---|---|---|
| Choi et al. [13] | Lab | 10 | HRV, RIP, GSR, and EMG | Custom chest-strapped sensor suite | Binary classification between stressed and not stressed with 81% accuracy |
| Healey and Picard [29] | Driving tasks | 9 | EKG, EMG, respiration, GSR | Custom sensors | Classification between low, medium, or high stress with 97% accuracy |
| Hernandez et al. [30] | Call center | 9 | GSR | Affectiva Q Sensor | Personalized model: 78.03% accuracy Generalized model: 73.41% accuracy |
| Muaremi et al. [40] | Sleeping | 10 | ECG, respiration, body temperature, GSR, upper body posture | Empatica E3, Zephyr BioHarness 3.0 | Classification between low, moderate, or high stress with 73% accuracy |
| Egilmez et al. [18] | Lab | 9 | Heart rate, GSR, gyroscope | Custom GSR sensor with LG smartwatch | Binary classification: F1-score of 0.888 |
| Sano and Picard [47] | Field | 18 | GSR and smartphone usage | Affectiva Q Sensor, and smartphones | Binary classification, 10-fold cross validation: 75% |
| Plarre et al. [42] | Lab, field | 21 | ECG and RIP | Custom sensor suite, AutoSense | Lab: Binary classification of stress with 90.17% accuracy, Field: High correlation (r = 0.71) with self-reports |
| Hovsepian et al. [32] | Lab, field | Lab train data: 21 participants Lab test data: 26 participants Field test data: 20 participants |
ECG and RIP | Custom sensor suite, AutoSense | Binary classification of stress: Lab train LOSO CV F1-score: 0.81 Lab test F1-score: 0.9 Field self-report prediction F1-score: 0.72 |
| Sarkar et al. [48] | Field | 38 | ECG and RIP | Custom sensor suite, AutoSense | Using models generated with cStress, field self-report prediction F1-score: 0.717 |
| Sun et al. [50] | Lab | 20 | ECG and GSR | Custom chest and wrist-based sensor suite | Binary classification of stress, accuracy by 10-fold cross validation: 92.4%; accuracy for cross-subject classification: 80.9% |
| Gjoreski et al. [25] | Lab, field | Lab: 21 Field: 5 | BVP, GSR, HRV, skin temperature, accelerometer | Empatica E3 and E4 | Lab: Classification between no stress, low stress, and high stress achieved 72% LOSO accuracy Field: Binary classification for detecting stress with F1-score of 0.81 |