University of Southampton


Biometrics is about recognising people automatically. Essentially we derive a set of numbers that are unique to a particular person. Our biometrics are usually based on using computer vision, so the set of numbers is derived each time we see the person. We seek to derive the same set of numbers each time we see the same person, and a set of numbers which is different from those of other people. Our work in biometrics started long ago, with work on automatic face recognition. We then came to pioneer recognition by the way a subject walks, and joined the early workers in recognition by the ears.  More recently we’re working on using human descriptions for recognition, which is akin with relating computer vision properties with those of human vision.

Walking subject extraction Biometric Feature Space (4 samples of each of 10 subjects) 3D Model

Essentially, we apply computer vision techniques to analyse sequences of images. By this we derive a set of numbers which represents a subject’s identity – a signature if you may. We recognise people by these sets of numbers, since each person has a different set. When viewed as a 3D plot, each time someone is seen by the cameras, a slightly different set of numbers is derived and these form clumps of numbers for each person.

Gait Biometrics | Ear Biometrics | Soft Biometrics | Books | Media | Publications

- Computer Vision -