Optimization of the positioning of sensors and energy harvesters for the design of portable physical activity recognition systems

The analysis of human motion offers the possibility of evaluating and improving sports performance, for example. Wearable systems have been developed using inertial measurement units (IMU). However, there are still issues, the determination of the number or the placement of these systems on the human body is not trivial and represents a common challenge to all applications. This thesis aims at investigating the sources of information on the human body in order to optimize the placement of these systems. The methods employed are based on the use of supervised learning algorithms such as SVM and on a biomechanical model. In the context of wearable sensors, the notions of embedded system and autonomy are also addressed. An energetic aspect is treated through the simulation of piezoelectric generators. Thus, an analysis of energy sources for different placements and orientations of harvesters on the human body is conducted. Finally, a multi-objective optimization method is implemented to meet the objectives of this thesis.