Department of Exercise Sciences

Gait Identification

The Department of Exercise Sciences Biomechanics Laboratory conducts numerous research projects.

Biomechanics Laboratory Project

Gait identification. Sharon Walt and Yanxin Zhang.

The study of personal identification technologies is required for various security applications, but also for medical studies, eg, separating between common and individual behaviour, or, for example, for the unique animation of characters in the movie industry.

While there are known limits of present biometrics technologies, multi-modal solutions and biometric technologies are research directions to overcome those limits. Until recently, identification mostly relied on fingerprint, facial and iris biometric techniques but because humans have an apparent inherent ability to recognise individuals by their walking gait attempts have been made in recent years to use certain kinematic gait parameters in developing unique identification protocols.

Studies have focused on either 2D representations of the lower body, temporal-spatial parameters such as walking period, or the analysis of image silhouettes. The field of clinical gait analysis has shown that gait kinematics (joint angles and derivatives) are highly repetitive and unique, and are consistent enough from day to day for making significant, clinical decisions. However, due to the subtle nature of the individual differences, unique identification is more challenging and requires a more advanced biomechanical model-based approach which incorporates whole-body 3D kinematic gait parameters, measures of movement coordination and image parameters.

Kinematic gait analysis is a relatively routine endeavor in a dedicated motion capture laboratory, but to date, accurate 3D kinematic representations of walking gait in public spaces (known as markerless tracking), using common surveillance cameras, has not yet been possible. This is despite good progress in markerless pose recognition, also supported by progress in neighbouring fields in computer vision [stereo analysis, video processing, Human Machine Infterface (HMI), and so on].

My proposed research will initially consist of two parallel (but closely interlinked) lines of inquiry:

  1. development of a minimum set of parameters, including kinematic, movement coordination, and/or image parameters, to produce a unique gait identifier
  2. development of a markerless tracking system that can create accurate body segmentation for the production of joint kinematic data. Both lines will then be combined for demonstrating a system for 3D-gait-based biometrics, at a distance.