Movement Detection of Human Body Segments: Passive radio-frequency identification and machine-learning technologies.

Movement detection of human body segments is a fertile research topic in human-computer interaction, as well as in medical and entertainment applications. In spite of the fact that most of the current methods to track motion are based on optoelectronic systems and wearable inertial sensors, promising solutions could spring from the application of passive radio-frequency identification (RFID) technology. When the human body´s limbs move within an electromagnetic field radiated by an interrogating antenna, a movement-dependent modulation of the backscattered field is sensed by the remote receiver. The collected signals, properly conditioned by wearable electromagnetic markers (tags), may therefore carry intrinsic information about human motion. This article investigates the potentiality of the synergy between electromagnetics and machine-learning technologies to classify gestures of arms and legs by using only passive and sensorless transponders. The electromagnetic signals, backscattered from the tags during gestures, are collected by a fixed reader antenna and processed by the support vector machine (SVM) algorithm to recognize periodic limb movements and classify more complex random motion patterns. Experimental sessions demonstrated a classification accuracy higher than 80-90%, which is comparable to more complex systems involving active wearable transceivers. The results further indicate that the achievable bit rate is 48 b/min, suggesting that the platform could be used to input coded controls to a gesture-oriented user interface.