We are interested in the recognition and prediction of human activity while people are mobile in order to provide the right services at the right time (or in the right situation). By applying machine learning and activity theory to sensor data available on a mobile phone and from the environment, we can recognize a user's current activity with contextual information, and then predict the next activity or behavior.
Based on the prediction, we might provide appropriate services to a user. While most researchers are focused on the first problem of recognizing current activity, almost no one is looking at the other problem of predicting future activity, which from a commercial perspective is more interesting, as it is tied inherently to the provision of services.
Following up on your request, I can provide my research abstract in more details.