Template Ageing in Gait Biometrics

Monday, 15 June 2015

In a gait biometric system or any biometric system in general evaluates the similarity between the stored reference templates and the probe template. The matching decision is fundamental element of any biometric system. At some stage person own template may start to fail due to the factor like ageing. The biometric systems deployed for security purpose is reliant on datasets. With ageing the biometric features tend to degrade.

Problem Being Solved

Template ageing is a common issue in biometrics data and refers to increase in the error rates caused by changes in the biometric pattern and its presentation with time. Longer time intervals usually makes it more difficult to match the individual’s sample to the stored templates i.e. the given biometric measurements tend to have a large intra-class variability. Therefore, it is possible for the stored templates to be significantly different over time. In gait analysis where the person is captured from a distance it is difficult to identify the ageing parameters. Even though the captured gait features may accurately identify the person, however the deterioration happening in these features due to ageing may not be visible. Therefore, a set of parameters that directly relates to the ageing factor can be determined if the lower body part is studied.



In the disclosed embodiment the changes in gait patterns are studied through the lower body part analysis. The disclosed embodiment proposes a set of features to analyse the variations occurring due to ageing in the gait information of an individual. Each individual person has a unique walking style which they usually adhere to during a normal walk. To masquerade someone else walking, the person needs to have similar body type and follow the same steps in a similar manner. Based on this concept it is then possible to create a different gait profile which distinguishes one person from another. Although individual’s gait vary due to factors such as physical build, shoe heel height, clothing, body weight and the emotional state of the walker, at a coarse level the basic pattern of bipedal motion is the same across healthy adults, and each person's body passes through the same sequence of canonical poses while walking. Thus, even if an individual’s walk will never be same and will always vary slightly (intra-class variance), it is still expected that lower body-part walking pattern will still expected to be same for all the instances of an individual’s gait samples. With ageing occurring in an individual the observed pattern will start to change. In a normal walking when a person’s legs are at a full stretch it is known as double-stance or double-support phase. If the resulting image at the double-stance phase is analysed, a virtual triangle is formed by both the legs and the ground, this is referred as a stride image. On an average as a subject walks the resulting stride image formed will be unique to the subject. This stride image provides information regarding the characteristics of the subject's gait and the possible deterioration occurring due to ageing can be analysed.

We have investigated and evaluated the set of features that can help in studying the ageing parameters in gait features using CASIA-B gait database. However, the approach requires gait dataset of individuals over a longer period of time to actually validate the concept.

Our method can be applied as an additional cue for multi modal biometrics, where other biometric features ageing can be analysed if the changes in the gait patterns is detected. In addition, the disclosed embodiment can be studied for health deterioration, where the subjects are reluctant to wear any kind of body worn sensors in clinical settings.



This invention results from advanced research into verifying individuals in surveillance videos (e.g. CCTV) using gait behavioural biometrics. Biometrics offers a reliable solution to certain aspects of identity management by utilizing fully automated or semi-automated schemes to recognize individuals based on their biological characteristics. Gait analysis and recognition form the basis of unobtrusive technologies for the detection of individuals who represent a security threat or behave suspiciously from surveillance videos.

The stored gait features used for verification and identification purposes are affected by different parameters such as ageing, injuries, fatigue and psychological condition. The changes in gait due to these parameters may make the gait of a lawful individual match the gait of a suspicious individual resulting in lawful person held on charges.


Opportunity/Partnership Sought


School of Computing and Intelligent Systems
University of Ulster
Magee campus