Wi-Fi Radar via Over-the-Air Referencing: Bridging Wi-Fi Sensing and Bistatic Radar

TL;DR

This paper introduces LoSRef, a novel Wi-Fi sensing scheme that uses the direct line-of-sight path as an over-the-air reference to enable phase-coherent radar-like delay-Doppler analysis with commodity Wi-Fi devices.

Contribution

It proposes a practical over-the-air LoS path referencing method for phase calibration, enabling Wi-Fi radar functionalities without wired references or dedicated antennas.

Findings

Achieved phase-coherent delay-Doppler responses using commodity Wi-Fi devices.

Enabled human gait and respiration sensing with sub-wavelength accuracy.

Demonstrated detection of target dynamics up to 20 dB weaker than static multipath.

Abstract

Wi-Fi channel state information (CSI), which is originally acquired for communication purposes, has recently been reused for sensing and radar-like functionalities. However, in practical Wi-Fi systems with independent clocks at the transmitter and receiver, the lack of a common delay and phase reference fundamentally precludes phase-coherent radar-like delay--Doppler analysis. By exploiting the line-of-sight (LoS) path component, i.e., the earliest-arriving direct path, as an over-the-air (OTA) reference for delay and phase, we propose an OTA LoS-path referencing scheme, termed LoSRef, that enables delay calibration and phase alignment under this practical constraint. Unlike conventional Wi-Fi bistatic radar systems that rely on wired reference signals or dedicated reference antennas, the proposed LoSRef-based framework enables phase-coherent bistatic radar-like operation that can be integrated into typically deployed Wi-Fi systems. Through human gait and respiration experiments in indoor environments, we demonstrate that phase-coherent channel impulse responses and corresponding delay--Doppler responses can be obtained using only commodity Wi-Fi devices. This enables physically interpretable human motion sensing, including gait-induced range variation and respiration-induced sub-wavelength displacement, as well as the extraction of target-induced dynamics up to 20 dB weaker than dominant static multipath components.