Context aware approach for activity recognition based on precondition-effect rules

Context awareness plays an essential role in systems dealing with activity recognition. The context information present to the system, and the way in which it is modelled, shape the performance of the system during activity inference. In this paper we present a novel approach for modelling human behaviour based on preconditions and effects and employing it for generating training-free probabilistic models, that are later used for recognizing the user activities. Furthermore, we use our approach to recognize the activities in a three-person meeting and compare the results from our generated models with those of hand crafted and trained models. Finally, we show that we are able to successfully infer the user state even in models with huge state space.