Personal Activity Logger with Hierarchical Activity Representation

Activity recognition is a key function for many context-aware applications in a smart environment. However, data collection and annotation for activity recognition is both time-consuming and costly. This paper proposes the hierarchical activity representation to enhance data reusability and introduces Personal Activity Logger (PAL), a computer aided tool with it, to reduce annotation efforts. We experimented with PAL in annotating activities within a personal space from power meters and a webcam in the office. Preliminary results show that PAL is effective in reducing the annotation efforts with only a slight loss in quality. In addition, we indicate the potential possibility to identify users from the distribution of events in their activities through the data analysis.