International Joint Conference on Ambient Intelligence edition:9 location:Pisa, Italy date:13-15 November 2012
Activity recognition mandates complex sensor fusion processing. Many contributions in the literature focus on improving the recognition accuracy of a limited set of activities or the efficiency of the algorithms. However, there is little work on how to dynamically adapt the activity recognition techniques when human activity goes beyond the observational and functional borders of one application-specific personalized assistance system. We present tool support to model transitions between activities, and a modular distributed framework of human activity recognition components with support for analyzing resource and recognition trade-offs for different deployments and configurations.