Tackling the Fidelity-Energy Trade-Off in Wireless Body Sensor Networks

Abstract

Wearable and connected health is a dominant field in the era of the Internet of Things (IoT). Indeed, Body Sensor Networks (BSNs) have been widely used for enabling many connected health applications in diverse areas including: activity recognition, elder care, sports, and rehabilitation. Although the number of transistors in an integrated circuit follows Moore's law, it does not apply to batteries. Thus, in BSNs, the sensor node's compact size means it has a strict energy constraint. At the same time, body sensor nodes have sensing requirements that should provide an acceptable level of data fidelity in order to accurately infer the amount and type of activity. This paper investigates the energy challenges associated with combining MEMS-based Inertial Measurement Units (IMU) and ball-tube motion detectors for BSNs. We examine multiple scenarios using these sensor types for body motion characterization. We then investigate the feasibility of these technologies along with collaborative sensing with the aim of reducing overall energy consumption. We conducted experiments examining sedentary (i.e. no movement) and non-sedentary activities. Our results show the possibility of 67 % reduction in energy consumption during the sedentary periods only, while maintaining the required level of data fidelity for inferring motion types.