Automated Generation of Models of Activities of Daily Living

In order to increase the safety of autonomous elderly people in their home, Ambient Assisted Living technologies are currently emerging. Namely, the recognition of their activities might be a way to detect eventual health problems, and can be performed in a Smarthome equipped with binary sensors. Hence, this communication aims at providing means to automatically generate a formal model of the Activities of Daily Living. A data mining approach in order to discover frequent habits of the observed inhabitant from a database of sequences of sensor events is proposed. Those frequent habits are then formally modelled using finite automata, leading to the construction of a map of habits mirroring the behaviour of the inhabitant. Such a model could then be used for online identification of habits, and even predictions of the upcoming behaviour. Results obtained on a case study are also presented.