This project focuses on data mining methods and sensors to model human behavior in home environments and techniques to infer knowledge from such models.
A demographic change is occurring in many areas of the world. The population share in which people in age over 60 years has been increasing for the last decades and estimations predict that this group of elderly population will near quadruple in the year 2050 ref1. This change will bring exponentially increasing costs of health care ref3, which will be supported by the decreasing share of younger people. One solution to this challenge is through technological developments aiming at reducing the costs of health care. Smart environments, ref2, targeted for ambient assisted living, enable people to remain independent at their own home and to live in a decent way longer. Key functions of such environments are: * Answering queries (where is the person, for example). * Activity recognition (what the person is doing). * Detection of specific behaviour and potentially dangerous situations. * Fall monitoring. Camera sensors have been used for the detection of human activities of daily living (ADL). However, the privacy issues of such camera-based solutions motivates the usage of other sensors such as wearable inertial sensors and accelerometers. A wearable sensor is dependent on several aspects of human behavior such as remembering to put on the sensors and doing so properly. Other, often used, sensors in ubiquitous computing are switches, motion detectors and electromechanical sensors, which do not, at the same extent, breach the privacy of individuals. Because of the large variety of sensor types and settings, information processing approaches, and individuals living in the environments, finding an accurate, robust and economically efficient solution to the problem is a hard task. This project focuses on data mining methods and sensors to model human behavior in home environments and techniques to infer knowledge from such models.