CAREACT: an adaptive and context aware framework for anomaly detection in ambient assisted living

Abstract

In this thesis, we analyzed current trends and challenges in ambient assisted living environments and deduced that there is a need for a monitoring system that incorporates existing telemonitoring concepts to activities of daily living recognition systems in a way that is easy to integrate, requires little or no maintenance after deployment and is easy to modify whenever needed. To address this need, we present CAREACT, an ambient assisted living system which monitors daily lives of inhabitants for anomaly detection in real-time with a set of features supported by state of the art literature, that have not been combined in a single system up to date: interoperability, rule based pattern detection, adaptivity and analyzability. CAREACT is built on complex event processing technology which has builtin support for rule based pattern detection and interoperability that works in real-time under heavy load. In this work, we also present a new methodology, called future events, that changes the paradigm of processing the time windows within patterns in a way that it takes constant time to process time windows regardless of their size. We implemented CAREACT and integrated it with one of the existing activities of daily living systems in our testbed environment to show that the proposed architecture is applicable and indeed it increases the efficiency of the anomaly detection process.