A Framework for Distributed Activity Recognition in Ubiquitous Systems.
Petteri Nurmi, Michael Przybilski, Greger Lindén, Patrik Floréen · 2005
Abstract — The recognition of human activity has many important applications that rely on linking observed behaviour with particular actions. However, activity recognition systems are usually build for specific applications, and the used architectures and solutions are often not applicable in other problem domains. In contrast, we propose a generic component-based framework for activity recognition. The use of components allows the architecture to be potentially distributed over the network and is thus suitable for a wide class of environments. We also demonstrate the workings of our framework by using it to recognize walking and travelling by metro from 3D accelerometer data.