TABLE 3.

Summary of fatigued gait studies.

Reference	Algorithm/best accuracy reported	How is data collected	Task
Russell et al. (2021)	CNN 97.8%	accelerometer worn around the chest, GPS watch for location tracking, 1 person	HAR: Climb Gate/Lay/Sit/Walk/Run. Variations in terrain and fatigue
Baghdadi et al. (2021)	MHTSCA with DTW as a dissimilarity measure	IMU worn at the right ankle 15 subjects	fatigue development over time
Sedighi Maman et al. (2020)	RF with BSS 85.5%	one sensor in the torso, 15 subjects	4-phase fatigue management framework in the workplace (1) detection (2) Identification (3) diagnosis: whole-body vs. localized (4) recovery
Maghded et al. (2020)	CNN/RNN	smartphone sensors, images and videos from the camera	detection of fatigue due to Covid-19
Karvekar, (2019)	2-class SVM 91%	24 subjects, smartphone attached to the shank	detection of fatigue: baseline, low, medium, and strong fatigued states
	3-class SVM 76%
	4-class SVM 61%
Baghdadi et al. (2018)	SVM 90%	one IMU in the ankle, 20 subjects	detection of fatigue after MMH tasks
Zhang et al. (2014a)	SVM 96%	17 subjects, IMU at sternum level	recognition of localized fatigued/non-fatigued state
Janssen et al. (2011)	SVM and SOM with PCA 98.1%	9 subjects GRFs	inter and intra-personal gait classification before, during, and after leg exhaustion

Legend: Best Subset Selection (BSS), Manual Material Handling (MMH), Multivariate Hierarchical Time Series Clustering Algorithm (MHTSCA).