Grace: Recognition of Proximity-Based Intentional Groups Using Collaborative Mobile Devices

People in social environments often appear in groups to have face-to-face interactions, automatically recognizing these groups will facilitate many applications while being unobtrusive. In this work, we focus on recognizing multiple concurrent intentional groups of people in proximity using a real-time distributed approach running on commodity mobile devices. Specifically, we study Bluetooth signal strength probability distribution for proximity estimation and model the group probability distribution under Bluetooth signal strength. Further, we develop a real-time distributed group determination algorithm. We implement a prototype system on iOS platforms which includes both the group recognition service and a mobile social application using the service. Both our simulation and prototype results show that the proposed approach can recognize proximity-based intentional groups with high accuracy.