LSI-REC: A Link State Indicator based gesture recognition scheme in a RFID system

This paper presents a novel gesture recognition scheme that utilizes RFID system to enable gesture recognition. The gesture feature of a certain object is captured by its shadow area in an arrayed RFID tags. We use the variance of counts of a tag being successfully read to represent the state of an obstructed link. The relative collection of tags determines the shape of an obstruction. We define a link state indicator (LSI) to indicate the signal changes caused by certain obstruction. In the proposed scheme, a gesture matrix is then introduced to represent gestures of a human being intuitively and fisher's linear discriminant method is used to recognize gestures. The scheme achieves the goal of gesture recognition without requiring any additional instruments of a human body. We build an experimental system and evaluate the scheme through a group of experiments, the result shows that the scheme can identify and classify a set of twelve defined gestures with an average accuracy of 94%.