Human Centric Technology and Service in Smart Space vol:182 pages:173-180
Lecture Notes in Electrical Engineering
HumanCom 2012 - 5th FTRA International Conference on Human-centric Computing edition:5 location:Gwangju, South Korea date:6-8 September 2012
Mobile based activity recognition is gaining importance with the ever increasing sensing, communication and computational power of smart phones. Our work addresses the current key challenges in the field of mobile sensing - how to make continuous sensing mobile applications both efficient and scalable. We motivate our work with a smart context-aware e-health application. Activities of daily living are modeled using a hidden Markov model and the classified activities are used to adapt sensing and feature extraction to decrease the energy consumption on the mobile. We present a federated context-management framework for activity recognition which implements a smart service adapter to offload execution to the cloud to achieve scalability and efficiency.