Event Detection for Ambient Assisted Living
Background
Currently, rehabilitation and assisted living includes intensive human interventions. Next to the desirable social interactions, the care attendants need to frequently control the home of the patients to extract information on their clinical status. For example, to detect senility and therefore the need to intensify the care, they control the trash, toilet and several further items. Many people grade these actions as invasive. To reduce these undesired surveys and increase the capabilities of the care attendants, a constant observation can be done by several sensors. To respect the self-reliance of the patients, it is most important that these sensors do not tamper with their privacy. Hence, a direct online observation of the patients by cameras or microphones is not reasonable. For a greater acceptance, the sensors have to be virtually invisible, measure only the environment of the patients, and require as few interactions with the patients as possible. The idea behind this method is that we have influence on our environment by interacting with it (for example, opening and closing a door, opening a window or switching the light), and therefore we can measure changes in our behaviour by measuring the status of the objects we interact with. It is necessary that the interpretation of the data is done automatically. This can also increase the acceptance by the patients, and no additional personnel is needed to control the measurements. If an intervention is necessary, the system is supposed to alert the care attendants.
Project Goal
Within this frame we conduct research with the purpose of designing algorithms that optimally detect such events. We model the normal measurements and define an event as the absence of the normal case. This is convenient, because the event is unknown by default. By observing the environment of the healthy patient and controlling the health status manually for some time, we can collect data of which we know that it describes the normal case.
Project Members
External partners
Our research project is part of SmartAssist. Our partners in SmartAssist are:
