Abstract
A wireless body sensor network (WBSN) facilitates on-body networking applications from physiological information acquisition, e.g. health monitoring, to human motion detection, e.g. fall detection. In this thesis, we focus on human motion detection applications with deterministic movements, e.g. rehabilitation exercise. A set of motion sensors and a sink are mounted on the rehabilitant to collect the motion data. However, the success rate of packet transmission from the sensor nodes to the sink might be reduced in the presence of body obstruction. A novel approach to relay node placement is hence proposed to preserve network connectivity. Our relay node placement scheme is divided into two phases. The first phase is to deploy the first phase relay nodes (FPRNs) on the torso to cover every on-body sensor node at any given time. We model the deployment of the FPRNs as a minimum set cover problem and a greedy algorithm is developed to solve it. The second phase places the second phase relay nodes (SPRNs) to provide connectivity from the FPRNs to the sink node. We model the deployment of the SPRNs as a minimum weight spanning tree problem and apply Prim's algorithm to solve it. The solution has been integrated into a rehabilitation assistant system, and apparent improvements in packet success rates are shown through authentic experiments in rehabilitation applications.