Analyzing Group Behaviors of Humans Using Spatiotemporal Data from Sensor Networks

Recent advances in technology have made practical the use of sensor networks for a wide range of problems, ranging from security applications to assistive environments and urban sensing. However, sensor systems today have limited capabilities in understanding human behaviors, thus requiring the presence of human operators, canceling, therefore, their main advantages which are large numbers and access to huge amounts of information. Of particular importance (e.g., airport security) is the ability of sensor systems to autonomously detect collective human behaviors (group behaviors) spanning over time. In the proposed thesis we will attempt to develop the theory and the tools required to allow sensor systems to autonomously detect groups of humans, assign roles to their members, learn their “habits” and detect sets of predefined behaviors.