Table 3.

Summary of characteristic of studies that investigated mental fatigue quantification using wearable devices.

No.	Type of study	# of participants	Fatiguing task	Task duration	Input data	Reference measure	Modeling approach	Output	Model performance
1	Lab	3	Mental arithmetic operations, reading professional literature	60 min	ECG, GSR, RES	Subjective self-report questionnaires, reaction time test	Decision Trees	3 levels	ACC >84%
2	Lab	6	Excavation operating simulation	60 min	Eye movement	SSS, NASA-TLX, task performance	SVM	3 levels	ACC = 85% precision = 86.6% recall = 85.9% F1 score = 85.1%
3	Lab	10	Virtual navigation	90 min	EEG	NASA-TLX	Dempster–Shafer fusion technique	4 Levels	F-score = 60–80%
4	Field	6	Not stated	Not stated	PPG, GSR, TSk	CFQ	Deep convolutional autoencoding memory network	Binary	ACC = 82.9%
5	Lab	10	Not stated	Not stated	GSR, PPG	Likert scale	Multilayer neural networks	4 levels	ACC = 71.2%
6	Lab	29	Quiz with 55 questions	54 ± 8 min	ECG	CFQ	K-nearest neighbors	Binary	ACC = 75.5% AUC = 0.74
7	Lab	28	Simulated driving	150 min	GSR, PPG, TSk, motion	Not stated	SVM	4 states (normal, stressed, fatigued and/or drowsy)	ACC = 68.31% (4 states)/ ACC = 84.46% (3 states)
8	Field	6	Not stated	Not stated	ECG, TSk	HRV metric	SVM	Binary	ACC = 94.3%*

ECG, electroencephalogram; EEG, electroencephalogram; GSR, galvanic skin response; PPG, photoplethysmogram; RES, respiration; T_Sk, skin temperature; CFQ, the Chalder fatigue scale; HRV, heart rate variability; NASA-TLX, NASA task load index; SSS, Stanford sleepiness scale; SVM, support vector machine; ACC, accuracy; AUC, area under the receiver operating characteristic curve.

^*Calculated average of classes (97.2% for alert, and 91.3% for fatigued state).