Rabia Jafri, Hamid R. Arabnia and Syed Abid Ali. Manuscript submitted to Computer Vision and Image Understanding (published by Elsevier).
Abstract: The performance of most current face recognition techniques degrades rapidly when presented with poor quality and low resolution face images; unfortunately, this is the only data available in many application scenarios. One solution is to use the gait in conjunction with the face since this biometric offers many of the same advantages, but is not impeded by the same challenges, as the face in these scenarios. In our previous work [Jafri et al. 2008], a system that integrated the face, a physiological biometric, with gait, a behavioral biometric, for automatically recognizing human beings was proposed. A decision-level fusion approach was adopted where the top matches of the face classifier were passed on to the gait classifier which then determined the identity of the unknown person. For gait recognition, a novel system was implemented, which utilized various gait features identified as being the most pertinent for recognition based on data collected using an optoelectronic motion capture system. To further develop this approach and to realistically evaluate its application potential and its scalability, we have now utilized data from a much larger number of individuals. Our results indicate that though the utility of certain gait variables decreases as the number of subjects increases, however, in general, the integrated face-gait system does continue to outperform the individual face and gait classifiers that it is composed of, demonstrating the viability of this approach even for larger data sets.