Gait recognition using acceleration from MEMS

Abstract

This paper presents an approach on recognising individuals based on 3D acceleration data from walking, which are collected using MEMS. Unlike most other gait recognition methods, which are based on video source, our approach uses walking acceleration in three directions: vertical, backward-forward and sideways. Using gait samples from 21 individuals and applying two methods, histogram similarity and cycle length, the equal error rates of 5% and 9% are achieved, respectively.