Deviation detection in automated home care using CommonSens

Abstract

Automated home care uses sensors to report about the well-being of monitored persons. Our complex event processing (CEP) system CommonSens simplifies the work of the application programmers, i.e., those who write the complex queries, by facilitating reuse of queries, sensor instances and environments. We have observed that many automated home care systems focus on detecting alarming behaviour, e.g., falls and heart attacks. However, it is impossible to predict and describe everything that can go wrong. In this paper we define deviation detection in CommonSens, which means that the application programmer only needs to state queries that define correct behaviour. If something happens that does not correspond to the query, this is interpreted as a deviation and a notification or alarm is sent. We believe that this approach further simplifies the work of the application programmer. Deviation detection is implemented in our CommonSens prototype, and we show that our prototype detects deviations as expected by running a set of functionality tests and an experiment based on real-world trace files.