Simulating human-centered pervasive systems requires accurate assumptions on the behavior of human groups. Recent models consider this behavior as a combination of both social and spatial factors. Yet, establishing accurate traces of human groups is difficult: current techniques capture either positions, or contacts, with a limited accuracy. In this paper, we introduce a new technique to capture such behaviors. The interest of this approach lies in the unprecedented accuracy at which both positions and orientations of humans, even gathered in a crowd, are captured. The open-source software pipeline we developed to exploit captured data allows extraction of several metrics on movement and social contacts, and permits study of their respective interrelationship. From the mobility to the topological connectivity, this framework offers a layered approach that can be tailored, allowing to compare and reason about models and traces. We demonstrate the accuracy and validity of our approach on social events and calibration runs in which we captured the motions of humans. In particular, we introduce an open-access trace of 50 individuals and compare it against random waypoint models that have the same global characteristics. Our fine-grain analyses, that take into account social interactions between users, show that the random way point model does not provide accurate predictions for socially-induced motion; to model human kinetics, new group-and interaction-based models should be developed. From the computer science point of view, these models are required to fully exploit the power of human-centered mobile computing, crucial for ubiquitous computing, and referred to as Short Range Communication Systems, Mobile Opportunistic Networking, or Mobile Networking in Proximity.