Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Optimal preprocessing of WiFi CSI for sensing applications

Vishnu V. Ratnam, Hao Chen, Hao Hsuan Chang, Abhishek Sehgal, Jianzhong, Zhang

TL;DR

This work tackles the problem of WiFi CSI-based sensing being degraded by RX gain and synchronization-induced phase errors. It develops a mathematical model separating gain into a slow large-scale term $g^{(1)}_{p}$ and a fast AGC term $g^{(2)}_{p} \in \mathcal{G}$, along with timing error $\tau_{p}$ and common phase error $\psi_{p}$, and proposes theoretically justified preprocessing algorithms to estimate and remove these impairments. The contributions include ML-based and clustering-based gain estimators for arbitrary and uniformly spaced AGC grids, and two timing/phase estimation methods tailored to strongly LoS and strongly static channels, all validated on simulated data and a respiration-rate sensing test bed. Results show substantial improvements in sensing performance, with up to $\approx 40$–$200\%$ higher post-cleaning SNR in simulations and around $20\%$ SNR gains in real respiration-rate experiments, illustrating the practical impact for robust WiFi sensing. The work enables more accurate recovery of the true CSI, enabling broader, contact-free sensing applications in smart environments and beyond.

Abstract

Due to its ubiquitous and contact-free nature, the use of WiFi infrastructure for performing sensing tasks has tremendous potential. However, the channel state information (CSI) measured by a WiFi receiver suffers from errors in both its gain and phase, which can significantly hinder sensing tasks. By analyzing these errors from different WiFi receivers, a mathematical model for these gain and phase errors is developed in this work. Based on these models, several theoretically justified preprocessing algorithms for correcting such errors at a receiver and, thus, obtaining clean CSI are presented. Simulation results show that at typical system parameters, the developed algorithms for cleaning CSI can reduce noise by $40$% and $200$%, respectively, compared to baseline methods for gain correction and phase correction, without significantly impacting computational cost. The superiority of the proposed methods is also validated in a real-world test bed for respiration rate monitoring (an example sensing task), where they improve the estimation signal-to-noise ratio by $20$% compared to baseline methods.

Optimal preprocessing of WiFi CSI for sensing applications

TL;DR

This work tackles the problem of WiFi CSI-based sensing being degraded by RX gain and synchronization-induced phase errors. It develops a mathematical model separating gain into a slow large-scale term and a fast AGC term , along with timing error and common phase error , and proposes theoretically justified preprocessing algorithms to estimate and remove these impairments. The contributions include ML-based and clustering-based gain estimators for arbitrary and uniformly spaced AGC grids, and two timing/phase estimation methods tailored to strongly LoS and strongly static channels, all validated on simulated data and a respiration-rate sensing test bed. Results show substantial improvements in sensing performance, with up to – higher post-cleaning SNR in simulations and around SNR gains in real respiration-rate experiments, illustrating the practical impact for robust WiFi sensing. The work enables more accurate recovery of the true CSI, enabling broader, contact-free sensing applications in smart environments and beyond.

Abstract

Due to its ubiquitous and contact-free nature, the use of WiFi infrastructure for performing sensing tasks has tremendous potential. However, the channel state information (CSI) measured by a WiFi receiver suffers from errors in both its gain and phase, which can significantly hinder sensing tasks. By analyzing these errors from different WiFi receivers, a mathematical model for these gain and phase errors is developed in this work. Based on these models, several theoretically justified preprocessing algorithms for correcting such errors at a receiver and, thus, obtaining clean CSI are presented. Simulation results show that at typical system parameters, the developed algorithms for cleaning CSI can reduce noise by % and %, respectively, compared to baseline methods for gain correction and phase correction, without significantly impacting computational cost. The superiority of the proposed methods is also validated in a real-world test bed for respiration rate monitoring (an example sensing task), where they improve the estimation signal-to-noise ratio by % compared to baseline methods.
Paper Structure (22 sections, 7 theorems, 55 equations, 8 figures, 3 tables, 6 algorithms)

This paper contains 22 sections, 7 theorems, 55 equations, 8 figures, 3 tables, 6 algorithms.

Table of Contents

  1. Introduction
  2. Accounting for gain errors
  3. Accounting for phase errors
  4. System model
  5. Estimation of gain errors
  6. Distribution of gain error
  7. Baseline methods
  8. Gain estimation for arbitrary $\mathcal{G}$
  9. Gain estimation for uniformly-spaced $\mathcal{G}$
  10. Estimation of large-scale gain $g^{(1)}_{p}$
  11. Estimation of $\widehat{g}_{p}$ for each $p$
  12. Estimation of $\lambda$
  13. Estimation of timing error and CPE
  14. Distribution of timing error and CPE
  15. Baseline methods
  16. ...and 7 more sections

Key Result

Lemma 1

When $\gamma \approx 1$, the true channel gain ${\Gamma}_{p}$ is i.i.d. Gaussian distributed for each $p \in \mathcal{P}$ with a zero mean and a small variance of $\sigma_{\Gamma}^2 = 100 (1 - \gamma^2)/K$.

Figures (8)

  • Figure 1: An illustration of system model depicting an WiFi access-point, a WiFi station and the CSI acquisition frame structure.
  • Figure 2: A real-world example of CSI power (in decibels) and $\angle \widetilde{h}_{p,k}$ depicting gain and phase errors.
  • Figure 3: An illustration of ${\Gamma}_{p}$, its ML estimate $\widehat{\Gamma}_{p}$, and the distortion introduced by the ML estimation.
  • Figure 4: Scatter plot of CSI power $\Gamma_{p}$ and incremental power $\Delta \Gamma_{p}$ for the two WiFi RXs in an example static channel.
  • Figure 5: Marginal probability density function and auto-correlation function of $\widehat{\tau}_p$ and $\widehat{\psi}_p$ (using Algorithm \ref{['Algo2']} with \ref{['eqn_CPE_timing_simpl']}), for an example high SNR, static channel.
  • ...and 3 more figures

Theorems & Definitions (23)

  • Lemma 1
  • proof
  • Lemma 2
  • proof
  • Lemma 3
  • proof
  • Lemma 4
  • proof
  • Remark 1
  • Lemma 5
  • ...and 13 more