People in office environments are exposed to a variety of interruptions. Some interruptions might support the office worker, some interruptions are very disruptive, such work gets more and more segmented resulting in lower productivity.
The goal of this project is to predict the interruptability of a subject in its office. Although the interruptability of a person is a inner state we achieved amazing prediction power. We showed that motion detection based on activity regions (the 5 regions separated by the bold lines in the Figure on the right) is very robust and outperforms audio which has been the best predictor so far. Additionally, we are able to determine one's presence in the office.
After having finished the feasability study, we continue this experiment for a larger number of offices. At the very end we are planning to obtain a productive system supporting collaborative work.