A connectionist embedded agent approach for abnormal behaviour detection in intelligent health care environments

This work aims to realise the vision of ambient intelligence in health care environments. The proposed system combines the use of unobtrusive sensors and effectors with intelligent embedded-agents. This paper presents a novel embedded agent mechanism based on an adaptive neural approach which is able to recognize activities inside an environment in an on-line mode. Its ultimate goal is to learn, discriminate and react to personal behaviours and signal departures from the normal behaviour that are significant to health care applications. Experiments were performed in a pervasive computing test-bed known as the iDorm. The presented results show that the system is able to characterize a normal set of activities as well as detecting novel or abnormal activities inside an environment.