Novel Device-Free Indoor Human Localization using Wireless Radio-Frequency Fingerprinting

We propose a novel device-free localization technique for human-object tracking indoors using wireless radio-frequency fingerprint. The received signal-strength indicators (RSSIs) are measured by the receiving antennae as the features for machine learning subject to the Random Forest model. In our proposed approach, both transmitters and receivers are fixed within the room, thus making human(s) free from carrying any transceiver. The placement of receivers will impact on the localization accuracy, and therefore we investigate the effect of receiver placement in this work. We will introduce an optimal receiver placement strategy such that the average communication-link distance can be minimized. Our proposed method is verified through simulations supported by a popular channel-propagation software, Feko. The pertinent experimental results demonstrate that the localization accuracy of 77.50% can be attained by our proposed Random Forest learning system for a 20 mx10 m indoor area divided into eight equi-sized zones where sixteen receiving antennae are placed at their optimal locations along the perimeter of the room.