Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Neuro-Wideband WiFi Sensing via Self-Conditioned CSI Extrapolation

Sijie Ji, Weiying Hou, Chenshu Wu

TL;DR

This work tackles the fundamental bandwidth bottleneck in WiFi sensing by introducing Neuro-Wideband (NWB), a paradigm that extrapolates continuous wideband CSI (eCSI) from a single narrowband CSI on commodity hardware. The WuKong framework (FreDiT) combines Relative Frequency Embedding, Self-conditioned Diffusion, and a Frequency-aware Transformer to perform self-conditioned CSI extrapolation, producing physically meaningful eCSI across expanded bandwidths $\Pi^e(f_i,k)$. The method is trained without wideband ground truth via self-supervision and demonstrates high-fidelity extrapolation (e.g., 20$\rightarrow$160 MHz) with strong multipath-resolving power, validated through localization and multi-person breathing case studies. Results show NWB can substantially improve ToF-based localization and vital-sign sensing while remaining hardware-compatible, opening a practical pathway for high-resolution, real-time RF sensing on commodity WiFi devices.

Abstract

WiFi sensing has suffered from the limited bandwidths designated for its original communication purpose, leading to fundamental limits in multipath resolution and thus multi-user sensing. Unfortunately, it is practically prohibitive to obtain large bandwidths on commercial WiFi, considering the conflict between the limited spectrum and the crowded networks. In this paper, we present Neuro-Wideband (NWB), a completely different paradigm that enables wideband WiFi sensing without specialized hardware or extra channel measurements. Our key insight is that any physical measurement of channel state information (CSI) inherently encapsulates multipath parameters, which, while unsolvable in isolation, can be transformed into an expanded form of CSI (eCSI) approximating measurements over a broader bandwidth. To ground this insight, we propose WUKONG to address NWB as a unique self-conditioned learning problem that can be trained by using any existing CSI data as self-labeled samples. WUKONG introduces a novel deep learning framework by integrating Transformer and Diffusion models, which captures sample-specific multipath parameters and transfers this sample-level knowledge to the outcome eCSI. We conduct real-world experiments to evaluate WUKONG on diverse WiFi signals across protocols and bandwidths. The results show the promising effectiveness of NWB, which is further demonstrated through case studies on localization and multi-person breathing monitoring using eCSI. Overall, the proposed NWB promises a practical pathway toward realizing wideband WiFi sensing on commodity hardware, expanding the design space of wireless sensing systems.

Neuro-Wideband WiFi Sensing via Self-Conditioned CSI Extrapolation

TL;DR

This work tackles the fundamental bandwidth bottleneck in WiFi sensing by introducing Neuro-Wideband (NWB), a paradigm that extrapolates continuous wideband CSI (eCSI) from a single narrowband CSI on commodity hardware. The WuKong framework (FreDiT) combines Relative Frequency Embedding, Self-conditioned Diffusion, and a Frequency-aware Transformer to perform self-conditioned CSI extrapolation, producing physically meaningful eCSI across expanded bandwidths . The method is trained without wideband ground truth via self-supervision and demonstrates high-fidelity extrapolation (e.g., 20160 MHz) with strong multipath-resolving power, validated through localization and multi-person breathing case studies. Results show NWB can substantially improve ToF-based localization and vital-sign sensing while remaining hardware-compatible, opening a practical pathway for high-resolution, real-time RF sensing on commodity WiFi devices.

Abstract

WiFi sensing has suffered from the limited bandwidths designated for its original communication purpose, leading to fundamental limits in multipath resolution and thus multi-user sensing. Unfortunately, it is practically prohibitive to obtain large bandwidths on commercial WiFi, considering the conflict between the limited spectrum and the crowded networks. In this paper, we present Neuro-Wideband (NWB), a completely different paradigm that enables wideband WiFi sensing without specialized hardware or extra channel measurements. Our key insight is that any physical measurement of channel state information (CSI) inherently encapsulates multipath parameters, which, while unsolvable in isolation, can be transformed into an expanded form of CSI (eCSI) approximating measurements over a broader bandwidth. To ground this insight, we propose WUKONG to address NWB as a unique self-conditioned learning problem that can be trained by using any existing CSI data as self-labeled samples. WUKONG introduces a novel deep learning framework by integrating Transformer and Diffusion models, which captures sample-specific multipath parameters and transfers this sample-level knowledge to the outcome eCSI. We conduct real-world experiments to evaluate WUKONG on diverse WiFi signals across protocols and bandwidths. The results show the promising effectiveness of NWB, which is further demonstrated through case studies on localization and multi-person breathing monitoring using eCSI. Overall, the proposed NWB promises a practical pathway toward realizing wideband WiFi sensing on commodity hardware, expanding the design space of wireless sensing systems.
Paper Structure (24 sections, 1 theorem, 16 equations, 11 figures, 2 tables, 1 algorithm)

This paper contains 24 sections, 1 theorem, 16 equations, 11 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background and Context
  3. The Wireless Channel and CSI
  4. Limitations of Existing Arts
  5. Rationale of NWB
  6. WuKong Design
  7. Relative Frequency Embedding
  8. Self-conditioned Diffusion
  9. Frequency-aware Transformer Model
  10. One-step CSI Extroplation
  11. One-step CSI Extroplation
  12. Experiments and Evaluation
  13. Evaluation Setup
  14. Baseline Comparison
  15. Multipath Resolution
  16. ...and 9 more sections

Key Result

Proposition 1

Given the narrowband CSI $\boldsymbol{H}^s$ observed over a narrow bandwidth $\Pi(f_i)$, and a continuous frequency band $\Pi^e(f_i,k)$ expanded by a factor of $k$, there exists a deterministic mapping for CSI extrapolation: that is uniquely determined by the parameters of the underlying physical environment $\Omega_i$ (including the geometry, materials, antenna positions, etc.).

Figures (11)

  • Figure 1: Limited WiFi bandwidth: 160MHz traffic (ch36) will degrade to 80/40/20 MHz bandwidths due to interference from overlapping 20MHz channels (ch40, ch56).
  • Figure 2: (i) Existing approaches assume a determined multipath mapping from current(extensive) data samples, whereas WuKong learns to extrapolate the multipath information inherent in the sub-samples of each sample (self-conditioned). (ii) Unlike conventional methods rely on stitching to achieve large bandwidth, WuKong attains substantial bandwidth directly.
  • Figure 3: Architecture of FreDiT.
  • Figure 4: Detailed mechanism of relative frequency embedding (RFE) of WuKong.
  • Figure 5: WuKong Data Collection Hardware: (a)(b); Scenario: (c) a natural hallway (d) a highly dynamic classroom (e) a natural office (f) a static meeting room.
  • ...and 6 more figures

Theorems & Definitions (2)

  • Proposition 1
  • proof