Combined passive radiofrequency identification and machine learning technique to recognize human motion

Abstract

Moving limbs within an electromagnetic field radiated by an interrogating antenna will generate a modulation of the backscattered field sensed by a receiver. The measured signals may therefore carry raw information about the human motion. Moreover, the proper placement of UHF passive Radiofrequency Identification (RFID) tags over body segments will increase the amount of collected signals. This paper investigate the potentiality of a possible synergy between Electromagnetics and Machine Learning technology at the purpose to recognize and classify, for the first time, the gestures of arms and legs by using only passive transponders. Electromagnetic signals backscattered from the tags during limb motion are collected by a fixed reader antenna and then processed by the Support Vector Machine (SVM) algorithm. Experimental results demonstrated a degree of accuracy in the classification of periodic movements that is fully comparable with that of more complex systems involving active wearable transponders.