System-level design of energy-efficient sensor-based human activity recognition systems: a model-based approach

This thesis contributes an evaluation of state-of-the-art dataflow models of computation regarding their suitability for a model-based design and analysis of human activity recognition systems, in terms of expressiveness and analyzability, as well as model accuracy. Different aspects of state-of-the-art human activity recognition systems have been modeled and analyzed. Based on existing methods, novel analysis approaches have been developed to acquire extra-functional properties like processor utilization, data communication rates, and finally energy consumption of the system