On evaluating the scalability aspect of gait-based biometric systems for larger population

The use of biometric technologies becomes ubiquitous in modern society due to the proliferating number of crimes and terror attacks as well as the vital need to provide safer environment. Gait biometrics is considered one of the emerging key research areas due to its potential use in a plethora of applications such as forensics and visual surveillance. As the evaluation of biometric-based evidence in investigation plays a pivotal role for its admissibility in court, we investigate in this research the performance of a model-based approach for people identification over an increasing population. A model-based markerless approach is described to derive the joints positions of walking subjects from uncalibrated single cameras through the use of a haar-like template matching approach. Matching of subjects is performed through posture comparison through a window of frames. Experimental results have shown that increasing the database size, the similarity scores for inter- and intra-class comparisons tend to converge to fixed values that are well separated. This suggests that for larger population, gait analysis can be still deployed and the size of the database should not be a factor impacting on the analysis.