Behavioral pattern detection from Personalized Ambient Monitoring

Bipolar disorder (BD) is a serious psychiatric condition that affects a large number of people. Many people with BD self-monitor their condition in order to try and keep the disturbances from affective episodes to a minimum. The Personalized Ambient Monitoring (PAM) project has developed a system that performs behavioral monitoring in an unobtrusive manner and can detect changes in a person´s behavior. The system uses a variety of discreet sensors to gather data on the parson´s behavior and this data is processed to extract behavioral patterns and detect changes in those patterns. In this paper we present one method of data processing that takes 24hr long data-streams from the sensors, pre-processes them and uses the Continuous Profile Model to align and extract the underlying patterns from the data-streams. We present some preliminary results from a technical trial.