Sensing Line-of-Sight Perturbations in 6-GHz Wi-Fi Using Channel Model-Based Features

A new standard, IEEE 802.11bf, has been created to offer Wi-Fi sensing capabilities across sub-7 GHz and millimeter-wave bands. Applications of Wi-Fi sensing rely on Channel State Information (CSI) to enable various applications such as motion detection, activity recognition, and gesture recognition. Within this context, this paper investigates part of the 6-GHz spectrum for use in Wi-Fi sensing, with the aim of recognizing different types of line-of-sight (LOS) perturbation. To achieve this, a novel feature extraction methodology is presented, along with innovative features designed to comprehensively capture information from CSI. More precisely, a novel Random Forest-based algorithm is introduced that automatically selects optimal features and constructs accurate decision trees for the classification of various human interactions with the LOS link between two Wi-Fi devices. The proposed feature extraction and selection methodology leverages variations in the channel, which are manifested by the changes in the characteristics of signal propagation caused by movements in proximity of the LOS link. Using statistical channel metrics, which can be directly linked to the physical channel, enhances the efficiency and accuracy of LOS perturbation classification. A detailed set of experiments are used to demonstrate the accuracy of our approach, which we call Channel Model based Features-Random Forest (CMF-RF). CMF-RF has been shown to outperform existing methods when used to classify human interactions with the LOS link.