Extended Model-Based Automatic Gait Recognition of Walking and Running

Gait is an emerging biometric. Current systems are either holistic or feature based and have been demonstrated to be able to recognise people by the way they walk. This paper describes a new system that extends the feature based approach to recognise people by the way they walk and run. A bilateral symmetric and coupled oscillator is the key concept that underlies this model, which includes both the upper and the lower leg. The gait signature is created from the phase-weighted magnitude of the lower order Fourier components of both the thigh and knee rotation. This technique has proved to be capable of recognising people when walking or running and future work intends to develop invariance attributes of walking or running for the new description.