Affiliations: OFFIS – Institute for Information Technology, Escherweg 2, 26121 Oldenburg, Germany. E-mail: {firstname.lastname}@offis.de | School of Medicine and Health Sciences, University of Oldenburg, Ammerländer Heerstraße 114-118, 26129 Oldenburg, Germany. E-mail: [email protected]
Abstract: A technical system for unobtrusive presence measurement and two novel models for describing user behavior in domestic environments are presented. Within the developed models user behavior is either described as the probability of being present at a certain location within an environment at a certain time on a day of the week or being present at a location for a certain number of times with a certain duration. The models are called timeslot-based and duration-based. Both models have been applied to presence information gathered by a technical system using home automation sensors. The system was installed into two flats of older people during a field trial for eight months. Results of the experiment show that the two models can be applied to describe individual user behavior. The influence of data structure and model quality on the detection of anomalies and the generation of alarms is discussed. On the long-term, the approach aims at detecting cutbacks in self-care ability and changes in health state by autonomously learning typical user behavior from presence information in a spatial model and by detecting untypical behavior, called anomalies, and generating alarms for caretakers. Such automatic assessment of self-care ability and health state is required in order to meet the increased challenges imposed to the decreasing number of care personal during the progress of the demographic change.
Keywords: Behavior modeling, ambient, home automation, assessment, field trial
