TABLE 11.

Gender recognition (GR).

Reference	AI Algorithm Best Achieved accuracy	Dataset/Input features	Task
Kwon and Lee, (2021)	KNN, SVM, NB, DT, 100%	UPCVgaitK1, UPCVgaitK2	GR
Zhang et al. (2019b)	multi-task CNN, AE: MAE = 5.47, GR: 98.1%	OULP-Age dataset GEI from video	GR and AE
El-Alfy and Binsaadoon, (2019)	LK-SVM with FLBP Normal: 96.40%	CASIA-B GEI from video	GR
	Carrying: 87.97%
	Wearing coat: 86.54%
(Jain and Kanhangad, 2018)	Bootstrap DT 94.44%	1D HG extracted from Smartphone in the front pocket	GR
(Castro et al., 2017)	CNN F:77%, M:96%	TUM-GAID: extracted from low-resolution video streams recorded with MS Kinect	automatic PI and GR
Lu et al. (2014)	AP clustering + SRML PI: 87.6% GR: 93.1%	Own dataset: ADSCAWD USF and CASIA-B C-AGI instead of GEI from MS Kinect Depth Sensor	PI and GR

Legend: Sparse Reconstruction-based Metric Learning (SRML), Cluster-based Averaged Gait Image (C-AGI), Affinity Propagation (AP), Optical Flow (OF), Person Identification (PI), Gender Recognition (GR), Age Estimation (AE), Fuzzy Local Binary Pattern (FLBP), Linear Kernel SVM (LK-SVM).

Datasets: UPCVgaitK1 (Kastaniotis et al., 2013), UPCVgaitK2 (Kastaniotis et al., 2016), OULP-Age (Iwama et al., 2012), CASIA-B (Yu et al., 2006), TUM-GAID (Hofmann et al., 2014), USF (Sarkar et al., 2005).