Individual and Group Activity Recognition in Moving Object Trajectories.

The knowledge about which activities people do at certain places is useful for several application domains. Existing works for activity recognition from trajectory data assume that only one activity is performed at each place, and do not identify the objects involved in each activity. We claim that several activities may be performed at certain places, such as shopping centers, universities, and others. In this paper, we propose a new method to recognize multiple activities performed at a place by integrating GPS trajectories and social media data, labeling trajectories with activities and the individuals involved in each activity. Experiments show that the proposed solution achieved good results in labeling and recognizing both individual and group activities.