Gait Recognition By Walking and Running: A Model-Based Approach

Abstract

Gait is an emerging biometric for which some techniques, mainly holistic, have been developed to recognise people by their walking patterns. However, the possibility of recognising people by the way they run remains largely unexplored. The new analytical model presented in this paper is based on the biomechanics of walking and running, and will serve as the foundation of an automatic person recognition system that is invariant to these distinct gaits. A bilateral and dynamically coupled oscillator is the key concept underlying this work. Analysis shows that this new model can be used to automatically describe walking and running subjects without parameter selection. Temporal template matching that takes into account the whole sequence of a gait cycle is applied to extract the angles of thigh and lower leg rotation. The phase-weighted magnitudes of the lower order Fourier components of these rotations form the gait signature. Classification of walking and running subjects is performed using the k-nearest-neighbour classifier. Recognition rates are similar to that achieved by other techniques with a similarly sized database. Future work will investigate feature set selection to improve the recognition rate and will determine the invariance attributes, for inter- and intra- class, of both walking and running.

More information

Identifiers

ORCID for Mark S. Nixon: orcid.org/0000-0002-9174-5934

Catalogue record

Export record