Experimental Demonstration of Contact-free Localisation Using Real-time Backscatter Sensing

Wireless technology has been used to locate and track people in real time using active sensors attached to the subject. Camera systems can be contact-free, but they are also invasive in terms of privacy. Backscattering and deep learning algorithms have enabled low-cost sustainable tracking, but this also needs to be worn by the subject. We present a novel solution that utilises backscattering from radio frequency identification (RFID) tags placed in the environment, not on the person, to detect the underlying location. These battery-less, sub-dollar energy harvesting tags are sustainable and enable passive contactfree localisation, acting as multiple sensors. To come up with a timely, real-time green, noninvasive wireless localisation system, we use a pre-trained linear machine learning algorithm to predict a person’s location based on a Received Signal Strength Indicator (RSSI) measured from an array of tags placed in the target environment. Our experiments demonstrate that we can accurately predict an individual’s location with high precision on commodity hardware, achieving a resolution of $0.4 \times 0.4 \mathbf{m}^{2}$ in a $6 \mathbf{m}^{2}$ area, with an accuracy of 81.26% in real-time.