The project presented here records the data flow from the
building sensors, gets to learn about the normal behaviour
of the house residents over several weeks and then deter-
mines deviations from normal behaviour. On the basis of the
sensor data collected by the system and the degree of de-
viation from the „normal state“, it can be assessed whether
a person, who is living alone, is doing as always or whether
changes – e.g. due to deterioration in health – can be recog-
nised. Regardless of the amount and types of sensors (light
switches, motion detectors, window contacts, etc.) that are
in use, the system learns the „normal state“ from the exist-
ing sensors. The resident does not notice anything when
the system is learning, because it is a „side effect“ of the
actual house network. In addition, the single person does
not get the feeling of being old, sick or dependent, as he
does not have to buy a special „health monitoring system“,
but uses the sensors for the building services.
The information flow of a KNX house bus system with its
system devices, sensors and actuators serves as the basis
for pattern recognition. A mini-PC preconfigured for the
house bus system is connected via the house bus system to
the home network via LAN in order to categorise data tel-
egrams based on their transmission and useful information
and to process them using software-supported algorithms.
Activities are stored internally through sensor detection.
On the basis of correlated influencing factors such as time,
frequency and grouped occurrence of data values, data are
calculated with statistical heuristics and finally weighted
and evaluated on the basis of a regular daily routine.