Classifying Human Behavior in Indoor Daily Living Environment for Predicting “Where to Go and What to Do Next?”

This article discusses how to predict the behaviors of the inhabitants in an indoor daily living environment such as a home or an office after their movements in the environment together with the destinations without predetermining the predictable behaviors or movements. The behaviors of the inhabitants in each place of the environment are characterized by frequency distributions of different postures possible to be taken by the inhabitants in the place. If a large amount of co occurrence is observed for a certain pair of the behaviors in different places right before and after every movement between those places, the behavior in one of the places can be predicted from the behavior in the other place. Furthermore, by choosing the varieties of postures and behaviors distinguished as different ones based on the co occurrence between those behaviors in different places, the postures and behaviors described with grain sizes appropriate for detecting the movements with predictable destinations and behaviors there are obtained.