Towards modelling behaviour from long-term ubiquitous deployments

Prof. Dr. Kristof Van Laerhoven

Institut fuer Informatik, Univ. Freiburg

Tuesday, May 05, 2015, 11:00 - 12:00

Room 01-012, Georges-Köhler-Allee 102, Freiburg 79110, Germany

Whether it is to help psychiatric patients in leading a structured life, or to give smokers an insight in what triggers them to light a cigarette, the automatic capturing of human behaviours has many uses. Physical activities, and sequences and rhythms thereof, are one of the most well-known indicators for describing such behaviours. In this talk, I will highlight a type of wearable activity recognition system that aims at being unobtrusive enough, so that it can be worn for weeks at a time without much intervention of the wearer. This is important, since it allows the system to detect the user’s physical activities in the most natural situations, often without the user being aware that she is monitored. Although this comes at the cost of accuracy in the detections, such a system still holds many promises and forces us to think about how activity recognition systems should present their data.