A Hybrid Method for Context-Based Gait Recognition Based on Behavioral and Social Traits

With the increasing demand for automatic security systems capable of recognizing people from a far distance and with as little cooperation as possible, gait as a behavioral biometric has recently gained a lot of attention. It is a remotely observable and unobtrusive biometric. However, the complexity and the high variability of gait patterns limit the power of gait recognition algorithms and adversely affect their recognition rates in real applications. With the goal to improve the performance of gait recognition systems without investing into costly and complex algorithms, we introduce a novel multimodal gait recognition system that combines the gait behavioral patterns of the subjects with the social patterns of their activities. For this purpose, a standard gait recognition system is implemented. A novel context matcher module is added to the system that provides a framework for modeling, learning, extracting and matching the contextual behavioral patterns. The learning of gait and behavioral patterns and clustering of results is performed. This allows grouping the subjects into similar profiles for faster recognition and enrollment. The results from two modules: context matcher and gait recognition are fused in the multi-modal decision making. The experiments on HumanID Challenge dataset are performed to validate that recognition rate improves using the combination of video context and gait recognition method even in the presence of low quality data.