Fusing Face & Gait for Non-Intrusive Person Identification in CCTV

This project is funded by a Royal Society and Natural Science Foundation of China (NSFC) International Joint Grant Scheme to develop models for the fusion of multiple biometrics (human face and gait) in robust and efficient non-intrusive person identification from a distance. Much progress has been made in the past decade on visual based automatic person identification through utilising different biometrics, including face recognition, gait analysis, iris and fingerprint recognition. However, each of these techniques in isolation has its inherent weakness and limitations. 

Current state-of-the-art face recognition algorithms cannot cope with CCTV data in which face images are captured unconstrained from a distance. Variations in viewing conditions such as pose and lighting can cause far more changes in facial appearance than variation between different people. The ability of the human vision system to recognise unfamiliar faces based on individual facial images in isolation is inherently poor. On the other hand, human walking gait provides perhaps a unique form of non-intrusive contextual information that can be readily captured and synchronised in the same CCTV images with the facial data. However, gait alone is highly unreliable and perhaps the least discriminative among all visually based biometrics. This joint project will investigate novel algorithms to combine facial image identification with both typical and abnormal human action and gait patterns in order to perform non-intrusive person identification from a distance in public space CCTV data.


  1. C. Shan, S. Gong and P. McOwan. Fusing gait and face cues for human gender recognition. Neurocomputing, to appear in 2007.

  2. S. Gong, C. Shan and T. Xiang. Visual inference of human emotion and behaviour. In Proc. ACM International Conference on Multimodal Interfaces, Nagoya, Japan, November 2007.

  3. C. Shan, S. Gong and P. McOwan. Beyond facial expressions: Learning human emotion from body gestures. In Proc. British Machine Vision Conference, Warwick, September 2007.

  4. C. Shan, S. Gong and P. McOwan. Learning gender from human gaits and faces. In Proc. IEEE International Conference on Advanced Video and Signal based Surveillance, London, September 2007.