Towards Collaborative Group Activity Recognition Using Mobile Devices

In this paper, we present a novel approach for distributed recognition of collaborative group activities using only mobile devices and their sensors. Information must be exchanged between nodes for effective group activity recognition (GAR). Here we investigated the effects of exchanging that information at different data abstraction levels with respect to recognition rates, power consumption, and wireless communication volumes. The goal is to identify the tradeoff between energy consumption and recognition accuracy for GAR problems. For the given set of activities, using locally extracted features for global, group activity recognition is advantageous as energy consumption was reduced by 10 % without experiencing any significant loss in recognition rates. Using locally classified single-user activities, however, caused a 47% loss in recognition capabilities, making this approach unattractive. Local clustering proved to be effective for recognizing group activities, by greatly reducing power consumption while incurring a loss of only 2.8 % in recognition accuracy.