Abstract
We show a hand shape recognition technique using conductive fiber included in a glove for a data glove. This technique estimates the bend of each finger and detects fingers' contact as the hand shape. For estimating the bend, the technique measures the electrical resistance of conductive fiber; the resistance decreases as the finger bends because the surface of the glove short-circuits. For detecting fingers' contact, the technique applies alternating currents of different frequencies on each finger and observes the signal propagation between the fingers. We conducted experiments to clarify the relationship between bend angles and the resistance of conductive fiber. Moreover, we compared the resistance of conductive fiber with that of a bend sensor. We also evaluated the accuracy of fingers' contact detection implemented using machine learning. Finally, we show that the technique also detects the contact of a pair of gloves, which are electrically independent, by observing propagated signals.
