Wireless Channel for 6th Generation Networks Sensing

Abstract

This thesis investigates the potential of using S21 parameter data for human presence detection in indoor wireless environments, as part of early research related to 6th generation (6G) communication systems. By combining physical measurements from a vector network analyzer with simulations generated using a digital twin of Chalmers’ antenna lab in Ansys HFSS, the study evaluates signal behavior across the 4–40 GHz frequency range. Power-Angle-Delay Profiles (PADPs) were created to visualize signal propagation and identify human-induced reflections. A feed-forward neural network was trained on both measured and simulated S21 data to classify human presence. The network achieved high classification accuracy when trained and tested on data from the same environment and on combined data, while cross-environment testing revealed a significant drop in performance, likely due to differences between simulated and real environments. Additionally, frequency slicing and link budget analysis were used to recommend suitable frequency bands for indoor sensing. The findings indicate that a lower-range band around 5 GHz offers a good signalto- noise ratio and maintains the main features of the signal. The study provides a proof of concept for RF-based human sensing using communication signals and offers insights into simulation accuracy, neural network training, and future development of 6G human-aware systems.