Efficient Beamforming Feedback Information-Based Wi-Fi Sensing by Feature Selection

TL;DR

This paper models the relationship between beamforming feedback information (BFI) and channel state information (CSI) in MIMO systems, proposing a feature selection algorithm that improves sensing efficiency with minimal accuracy loss.

Contribution

It derives a mathematical model of BFI, develops a CRB-based feature selection method, and demonstrates improved sensing efficiency with fewer features and parameters.

Findings

BFI and CSI have comparable sensing capabilities.

The proposed algorithm halves the number of features needed.

Parameter reduction exceeds 20% with minimal impact on positioning accuracy.

Abstract

Wi-Fi sensing leveraging plain-text beamforming feedback information (BFI) in multiple-input-multiple-output (MIMO) systems attracts increasing attention. However, due to the implicit relationship between BFI and the channel state information (CSI), quantifying the sensing capability of BFI poses a challenge in building efficient BFI-based sensing algorithms. In this letter, we first derive a mathematical model of BFI, characterizing its relationship with CSI explicitly, and then develop a closed-form expression of BFI for 2x2 MIMO systems. To enhance the efficiency of BFI-based sensing by selecting only the most informative features, we quantify the sensing capacity of BFI using the Cramer-Rao bound (CRB) and then propose an efficient CRB-based BFI feature selection algorithm. Simulation results verify that BFI and CSI exhibit comparable sensing capabilities and that the proposed algorithm halves the number of features, reducing 20% more parameters than baseline methods, at the cost of only slightly increasing positioning errors.