Sensing SNR-Based Feature Selection of Beamforming Feedback Toward Firmware-Agnostic WiFi Sensing

In this study, we analyzed the beamforming feedback (BFF) as a WiFi sensing feature. The BFF is used for eigenmode transmission and is calculated from the channel state information (CSI), which includes the fine-grained information mirroring propagation characteristics. Owing to its accessibility, BFF has attracted attention as an alternative WiFi sensing feature to CSI. However, whether the elements of a BFF can be utilized as sensing features has not been sufficiently discussed. Thus, we analyze the averaged stream SNR (AS-SNR) included in the BFF as a WiFi sensing feature by leveraging the sensing SNR (SSNR), which is a metric for evaluating the WiFi sensing capability. The analysis results reveal that the final-stream AS-SNR with the lowest SNR is the best feature among all streams. An experimental evaluation using respiration rate estimation supports this hypothesis.