TABLE 7.

Tracking human pose.

Reference	AI Algorithm Best Achieved accuracy	Data Acquisition	Purpose
Vandersmissen et al. (2018)	Deep CNN. Error 21.54%	Low-power Radar. IDRad dataset made publicly available	Indoor PI invariant to the exact radar placement, room setup, and walking direction
Tariq et al. (2017)	Weka collection ML classifiers. 0.05 localization error. Accuracy > 93%	4 Capacitive Sensors in load mode	Indoor Person Localization
(Li et al., 2015)	Improved PDR algorithm The best achieved accuracy is within 2 m	Samsung Galaxy Note3 and Bluetooth beacons	PDR algorithm integrated with Bluetooth beacons for indoor positioning without additional infrastructure
Robertson et al. (2013)	MagSLAM Achieves a position accuracy of 9–22 cm	Foot mounted IMU sensors. Low-power radar device. No a priori map	Dynamic positioning (SLAM) of indoor pedestrians derives a multi-floor indoor map