Robustness of biometric gait authentication against impersonation attack

Abstract

This paper presents a gait authentication based on time-normalized gait cycles. Unlike most of the previous works in gait recognition, using machine vision techniques, in our approach gait patterns are obtained from a physical sensor attached to the hip. Acceleration in 3 directions: up-down, forward-backward and sideways of the hip movement, which is obtained by the sensor, is used for authentication. Furthermore, we also present a study on the security strength of gait biometric against imitating or mimicking attacks, which has not been addressed in biometric gait recognition so far. © Springer-Verlag Berlin Heidelberg 2006.