Is it possible to apply an existing gait recognition method to pathological subjects without any adaptation?