Activity Recognition with Fuzzy Finite Automata

Offering context aware services to users is one of the main objectives of pervasive computing. A context aware system needs to know the activities being performed by the user. Deciding what a user is doing at a given time poses a number of challenges. One significant challenge is dealing with the variation in the number, order and duration of the constituent steps of an activity. There happens to be considerable variation in these parameters even if the same user is performing the same activity at different times. Though fuzzy finite automata have been used by researchers to overcome this challenge, manual construction of the automata for daily life activities becomes onerous. This paper illustrates how finite automata can be constructed automatically and fuzziness incorporated into it, to recognize user activities in a smart environment. The proposed method is tested with a publicly available dataset and is found to give promising results.