 | |
|
Wristband sensor platform with RFID Antenna, 3-axis IMU
|
Mobile assistence systems are technical appliances, which proactively support humans while performing specific activities. A main challenge in the design of those systems is activity recognition – the identification of the user's current action by the system. If also the surrounding context is regarded, the framework is called situation-aware.
In our work we are combining sensors, allowing a partial observation of different szenarios, and models, which provide rules for inferring a full image of the situation from partial and incomplete data.
One aim of our research is to analyze the advantages of the model based approach for accomplishing situation-awareness. In such an approach the system contains an idealized model of the considered spectrum of activities. In this case the activity recognition determines that model activity which explains the real observations best. We are focusing in particular on the most promising probabilistic models. Therefor it is of special importance that there is at least a partial sensorically recognizable connection between abstract activities and concrete physical actions. Because of the use of model based recognition methods this connection doesn't necessarily have to be distinct, deterministic or complete – for missing, inconsistent or noisy sensor data the corresponding activities can be extrapolated from the model's prediction.
