Person Identification through Harvesting Kinetic Energy

Abstract

Energy-based devices made this possible to recognize the need for batteryless wearables. The batteryless wearable notion created an opportunity for continuous and ubiquitous human identification. Traditionally, securing device passwords, PINs, and fingerprints based on the accelerometer to sample the acceleration traces for identification, but the accelerometer's energy consumption has been a critical issue for the existing ubiquitous self-enabled devices. In this paper, a novel method harvesting kinetic energy for identification improves energy efficiency and reduces energy demand to provide the identification. The idea of utilizing harvested power for personal identification is actuated by the phenomena that people walk distinctly and generate different kinetic energy levels leaving their signs with a harvested power signal. The statistical evaluation of experimental results proves that power traces contain sufficient information for person identification. The experimental analysis is conducted on 85 persons walking data for kinetic power signal-based person identification. We select five different classifiers that provide exemplary performance for identifying an individual for their generated power traces, namely NaiveBayes, OneR, and Meta Bagging. The experimental outcomes demonstrate the classifier's accuracy of 90%, 97%, and 98%, respectively. The Dataset used is publicly available for the gait acceleration series.