WBAN Node Topologies for Reliable Posture Detection from On-Body UWB RSS Measurements


Robert Heyn and Armin Wittneben


IEEE International Conference on Communications (ICC), Seoul, South Korea, May 2022.

[BibTeX, LaTeX, and HTML Reference]


In the growing domain of e-health and technology-assisted living, posture monitoring is an essential component. Wearable systems are favorable for ubiquitous operation, and radio signals between on-body nodes have shown to be a suitable means for detecting human body postures and enabling low-cost and low-complexity systems. In order to minimize overall power consumption and increase the comfort for the user, reducing the number of nodes and optimizing their placement on the body is essential. As an extensive analysis on the topology of the wireless body area network (WBAN) has not been performed yet due to a lack of comprehensive data, this work takes a measurement-based approach to the selection of suitable on-body links for posture detection. We analyze and compare the posture classification performance based on low-complexity UWB RSS measurements using Random Forest classifiers. Ranking links and on-body nodes according to their respective importance for the classification task provides a basis for selecting topologies for three different architectures, which consist of (i) the most important links, (ii) selected TRX nodes and (iii) hierarchical TX-RX topologies. With proper selection of only 8 ultra-low-complexity TX and 4 simple RX nodes, a classification accuracy of 91 % can be achieved on a very diverse posture set. The influence of measurement time and system bandwidth is analyzed for selected topologies of all three architectures and discussed in the context of implementation with UWB Impulse Radio.


posture detection, UWB, WBAN, classification

This publication is currently not available for download.