TABLE 5.

SG in sports.

Reference	Algorithm	Data Collection/Input	AI Task/Output
Taborri et al. (2021)	Linear SVM 96%	N = 39, inertial sensors, and optoelectronic bars	ACL risk prediction in female basketball players via LESS score
Johnson et al. (2021)	CNN, not enough accuracy	Wearable accelerometer	predict near real-time GRF/Ms from kinematic data
Nguyen et al. (2020)	CNN	7 IMU’s	Gait classification: athlete vs. foot abnormalities
Guo and Wang, (2021)	TS-DBN	Public datasets of videos KTH and UCF	HAR/sports behavior recognition
Gholami et al. (2020)	CNN	shoe-mounted accelerometer	Abnormal running kinematics Activity recognition
Cronin et al. (2019)	DeepLabCut	single GoPro camera	Markerless 2D kinematic analysis of underwater running
Kang et al. (2018)	FFT	Smartphone (unconstrained)	Detects walking, counts steps, irrespective of phone placement
Onodera et al. (2017)	ANN with IG	infrared cameras and force plates	Influence of shoe midsole resilience and upper structure on running kinematics and kinetics
Sundholm et al. (2014)	KNN with DTW	pressure sensor mat	Exercise detection and exercise count

Legend: Fast Fourier Transform (FFT), Time-Space Deep Belief Network (TS-DBN), Landing Error Score System (LESS), Ground Reaction Forces and Moments (GRF/M), DeepLabCut as in (Mathis et al., 2018).

Datasets: Royal Institute of Technology (KTH) (Jaouedi et al., 2020) and University of Central Florida (UCF) (Perera et al., 2019).