Detection Technique of Individual Characteristic Appearing in Walking Motion

Abstract

In this study, we develop a technique to measure walking under noncontact conditions by using an ultrahigh sensitivity electrostatic induction technique. The results of walking measurements using this technique indicate that the detected electrostatic-induced current waveform exhibits a peak at the time of foot contact or detachment owing to walking. Based on these results, we construct a theoretical model in which an induced current is generated, and the correspondence relationship between walking and electrostatic-induced current is revealed. Furthermore, when comparing the walking signals of each participant, we used a scalogram obtained by performing a wavelet transformation on the walking signal. Person identification was attempted by learning the participant's scalogram using a convolutional neural network (CNN).

Keywords

• Electrostatic induction
• walking motion
• person identification
• deep learning
• convolutional neural network