In Germany, the majority of people already owns a smartphone or another mobile device – and carries, unconsciously, a well-equipped sensor system with them. Just a few of these sensors are sufficient to draw conclusions about current and future activities or the context of the owner. Even non-physical data sources may provide relevant information for this purpose. User-sensitive information such as logs of cloud services, SMS or call history, calendar data or data from social networks provide further information on the current status and the activity of the user.

Trained behavior models can be used to autonomously detect the current user activity and trigger an response of the system. Such responses may include various reactions such as the predictive pre-caching of data, automatic initialization of services or the adaptation of services to the current activity.

Such models can be improved by applying so-called crowdsourcing. Data from many users (the “crowd”) is collected and analyzed. In the vast amount of data common behavior patterns and correlations between sensor data and the underlying activity or the user context can be identified.

The aim of the MARC² project is to develop a framework that supports existing mobile applications with new features in terms of activity recognition and user analysis.

Start/End

• 03/2014 — 02/2016

Partners

• SAP AG

Research Topics

• Human Activity Recognition, Mobile Computing, Crowdsourcing, Cloud Services

Contact

• Anja Bachmann (email: bachmann@teco.edu)