Brownian Bridge Diffusion-Assisted Joint Channel Estimation and Data Detection for Jamming-Resilient OFDM Receivers

arXiv:2606.28778 · cs.IT, cs.AI, eess.SP, math.IT · Submitted 2026-06-27 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Computer Science. The day in computer science: software, systems and theory on arXiv.

Tom: I'm Tom.

Jane: And I'm Jane, here with Lu, Meng and our in-house large language model, Lalam.

Tom: We're looking at the paper: "Brownian Bridge Diffusion-Assisted Joint Channel Estimation and Data Detection for Jamming-Resilient OFDM Receivers".

Paper summary: Tom: Alright, so to recap from our first part, this paper is about tackling jamming in next-generation wireless networks where device density is high and spectrum is limited one <ref:2606.28778#pg1,in next-generation wireless networks>. The core thesis is that when jamming overlaps with both pilot and data symbols in time and frequency, it severely bottlenecks joint channel estimation and detection one <ref:2606.28778#pg1,pilot and data symbols in>.

Jane: So the authors propose a framework called the BBD-JCED framework to fight this. They have two modules: one extracts jamming features in the short time Fourier transform domain to suppress those bad samples, improving the signal-to-jamming-plus-noise ratio or SJNR one <ref:2606.28778#pg1,extracts jamming features in the short time Fourier transform>.

Lu: And then they pair that with a Brownian bridge diffusion process to model how the suppressed signal and the encoded bits behave, even when channel estimation has errors, which enables better joint estimation and detection one <ref:2606.28778#pg1>.

Meng: So if I’m thinking about this in terms of deployment, it’s essentially two stages: first clean the noise out of the signal domain, then use a diffusion model to intelligently recover the data despite imperfect channel knowledge one <ref:2606.28778#pg1>. That sounds like a lot of steps for a chip.

Tom: It is, but they managed to make that second part efficient by deriving a fast ordinary differential equation solver for the Brownian bridge process one <ref:2606.28778#pg1>. They also designed their training algorithm in two stages: first train the jamming suppression part, and then jointly train both modules using a cosine annealing strategy three <ref:2606.28778#pg3>.

Jane: So what does this actually mean for us? It means they aren't just throwing a big neural network at the problem hoping it works; they are building a structured system where one part cleans and the other recovers intelligently one <ref:2606.28778#pg1>.

Lu: What’s interesting is how they handle the channel estimation refinement in their BB-JR module by using a residual channel attention network to refine the initial linear estimate into an interpolated channel state information or CSI two <ref:2606.28778#pg2>.

Meng: And what does that interpolation do? Does it just smooth out the errors, or does it actually help recovery significantly? Because I’m worried about adding extra dependencies.

Tom: The authors show that this joint approach, using their BBD-JCED framework, significantly reduces jamming components in the received signal and offers superior bit recovery performance compared to baseline schemes three <ref:2606.28778#pg3,jamming components in the received signal and offers superior bit recovery performance>.

Jane: So the numbers they provide show that this method keeps the model parameters low and computational complexity competitive with existing methods three <ref:2606.28778#pg3>. That’s a big win for practical implementation.

Conclusion: Tom: So, looking at this paper, Brownian Bridge Diffusion-Assisted Joint Channel Estimation and Data Detection for Jamming-Resilient OFDM Receivers by She et al., it seems like the authors successfully combined two distinct techniques to solve a hard problem one <ref:2606.28778#pg1>.

Jane: Exactly. They took a signal processing idea—jamming suppression via STFT masking—and combined it with a statistical modeling idea—the Brownian bridge diffusion process for joint recovery one <ref:2606.28778#pg1>.

Lu: The main implication, from my side, is that when you have overlapping time and frequency jamming, treating the channel estimation and data detection as two separate steps just doesn't work reliably anymore two <ref:2606.28778#pg2>.

Meng: So what’s the real-world change? For an engineer deploying this on a network, it means we can expect better reliability even in really noisy environments where we have to deal with both pilot and data symbols being hit simultaneously one <ref:2606.28778#pg1>.

Tom: Right. And the authors proved that they can achieve this without ballooning their model size or making the device too slow, which is what makes it practical for real-world hardware three <ref:2606.28778#pg3>.

Jane: It’s about moving away from single-task solutions toward integrated systems that handle these complex interference scenarios more robustly, and this paper provides a clear way to do that using diffusion modeling one <ref:2606.28778#pg1>.

Honghan She, Yufan Cheng, Tieming Sun, Pengyu Wang, Siya Huang, Kaikai Yang

University of Electronic Science and Technology of China · School of Artificial Intelligence, University of Science and Technology Beijing

cs.IT, cs.AI, eess.SP, math.IT

Submitted: 2026-06-27

Updated: 2026-11-18

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 83/100

The gist: In next-generation wireless networks, the growing density of devices and limited spectrum resources pose severe jamming challenges to fragile legitimate communication links in the wireless…

Key concepts

Brownian Bridge Diffusion (BBD)
This is a mathematical process used to model how a signal evolves over time when errors, like channel estimation mistakes, are present. It helps in tracking the signal's path more accurately than traditional methods, allowing for better joint estimation of the channel and data.
Jamming Suppression Front-end (JSF)
This module uses a U-Net based mask estimator to analyze the signal in the time-frequency domain. It identifies and suppresses specific time-frequency bins where jamming is present, effectively cleaning up the received signal before further processing to improve performance.
Joint Channel Estimation and Data Detection
Instead of estimating the channel and detecting data separately, this framework treats them as coupled problems. The BB-JR module uses a residual channel attention network to refine channel estimates while simultaneously using diffusion modeling to recover the original encoded bits.
Fast ODE Solver
Since modeling the BBD process iteratively is computationally expensive, the authors developed a fast ordinary differential equation (ODE) solver. This allows the complex diffusion evolution to be calculated quickly, keeping the overall framework computationally efficient.

Terminology

Summary

In next-generation wireless networks, the growing density of devices and limited spectrum resources pose severe jamming challenges to fragile legitimate communication links in the wireless electromagnetic environment <ref:2606. p. 028778#pg2> Crucially, when jamming overlaps with pilot and data symbols in both time and frequency domains, it inflicts a severe bottleneck on receiverside joint estimation and detection <ref:2606. p. 028778#pg2> To address this issue, we propose a Brownian bridge diffusion-based joint channel estimation and data detection framework (BBD-JCED) for jamming-resilient receivers <ref:2606. p. 028778#pg2> Simulation results demonstrate that the proposed framework achieves superior bit recovery performance compared with baseline schemes while maintaining a lower number of model parameters and competitive computational complexity <ref:2606. p. 028778#pg2>

Framework Architecture

The proposed BBD-JCED framework comprises two core modules: the first extracts jamming features in the short-time Fourier transform (STFT) domain and suppresses jamming samples, thereby improving the signalto-jamming-plus-noise ratio (SJNR) of the received signal <ref:2606.28778#pg2> The second introduces a Brownian bridge diffusion (BBD) process to model the evolution of the suppressed signal and the encoded bits in the presence of channel estimation errors, thereby enabling enhanced joint channel estimation and data detection <ref:2606. p. 028778#pg2> This overall architecture integrates two core modules: the jamming suppression front-end (JSF) module and the Brownian bridge-joint recovery (BB-JR) module <ref:2606. p. 328778#pg2> The JSF module employs a U-Net-based mask estimator to identify jamming-contaminated time-frequency bins and suppresses malicious jamming, thereby significantly enhancing the SJNR for subsequent processing <ref:2606. p. 328778#pg2>

Core Modules Details

The JSF module is designed to capture the time-frequency characteristics of jamming signals in the STFT domain, making it particularly suitable for handling narrow-band or non-stationary jamming <ref:2606.28778#pg2> It utilizes a fully trained U-Net based mask estimator to identify the positions of jamming samples in the STFT domain and output the mask Mˆ for notching <ref:2606. p. 328778#pg2> The BB-JR module tightly couples channel estimation and data detection, utilizing a residual channel attention network (RCAN)-based interpolator to refine the LS estimate into the interpolated CSI <ref:2606. p. 328778#pg2> Subsequently, a bilinear residual layer (BRL)-based diffusion origin estimator fuses multi-dimensional features and iteratively performs the evolution of the BBD process via an embedded ODE solver, thereby achieving data recovery <ref:2606. p. 328778#pg2>

Computational Efficiency and Training

To alleviate the computational burden of the BBD process in the second module, we derive a fast ordinary differential equation (ODE) solver that enables its low-complexity iterative evolution <ref:2606.28778#pg2> The training algorithm is designed to improve data recovery performance by first training the JSF module and then jointly training the JSF and BB-JR modules <ref:2606. p. 728778#pg2> This joint training employs a cosine annealing strategy to balance the trade-off between channel estimation accuracy and data detection performance <ref:2606. p. 128778#pg2>

Performance Validation

Simulation results demonstrate that the proposed framework ensures effective data recovery under various types of jamming <ref:2606.28778#pg2> Specifically, the proposed framework significantly reduces the jamming components in the received signal and offers superior bit recovery performance compared to baseline schemes while maintaining a lower number of model parameters and competitive computational complexity <ref:2606. p. 228778#pg2> The BBD-JCED consistently achieves the best BER performance across all tested jamming scenarios, demonstrating superior robustness of the proposed framework for jamming-resilient wireless reception <ref:2606. p. 928778#pg2>

Ablation Study Results

The ablation study shows that neither the baseline receiver (“None”) nor the standalone JSF module achieves reliable data recovery under both severe CSN and LFM jamming, resulting in an almost flat data BER curve <ref:2606.28778#pg2> While the standalone BB-JR module provides limited robustness, its performance degrades significantly without JSF-based preprocessing <ref:2606.28778#pg2> The complete BBD-JCED framework consistently achieves the lowest error probabilities across all settings, indicating strong complementarity between the two modules <ref:2606. p. 928778#pg2> Using a target data BER of 10−5, the proposed BBD-JCED achieves an SJR gain of approximately 9–10 dB over the standalone BB-JR module in both TDL-A and TDL-D channels under CSN and LFM jamming <ref:2606. p. 928778#pg2>

Conclusion

In this paper, we propose a Brownian bridge diffusion-based joint channel estimation and data detection framework for jamming-resilient receivers <ref:2606.28778#pg2> The framework consists of two modules: (i) the JSF module, which employs a masking estimator to identify jamming-contaminated time-frequency bins and suppress jamming; and (ii) the BB-JR module, which leverages a BBD process to model the evolution of the suppressed signal and encoded bits in the presence of channel estimation errors <ref:2606. p. 028778#pg2> In addition, we derive a fast ODE solver that enables low-complexity iterative evolution of the BBD process to reconstruct the coded bits <ref:2606. p. 028778#pg2> Finally, we design a dedicated training algorithm that combines multi-stage training with cosine-annealed weighting to further improve data recovery performance <ref:2606. p. 028778#pg2> Extensive simulation results demonstrate that the proposed framework achieves lower channel BER and data BER than baseline schemes under both CSN and LFM jamming, while maintaining a lower number of model parameters and competitive computational complexity <ref:2606. p. 0

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Improved AI system can perform joint channel estimation and data detection in jamming-contested environments by integrating a U-Net based mask estimator to identify jamming samples in the Short-Time Fourier Transform domain and adaptively applying spectral notching to suppress malicious jamming, thereby improving the signalto-jamming-plus-noise ratio (SJNR).

  2. Improved AI system can model the evolution of suppressed signals and encoded bits in the presence of channel estimation errors using a Brownian bridge diffusion process anchored by fixed endpoints, enabling enhanced joint channel estimation and data detection.

  3. Improved AI system can achieve low computational complexity during iterative evolution by deriving a fast ordinary differential equation (ODE) solver for the Brownian bridge diffusion process, which enables its low-complexity iterative evolution in real-time deployment.

  4. Improved AI system can improve data recovery capability by utilizing a multi-module training algorithm that combines training the jamming suppression front-end with cosine annealing strategy to balance trade-offs between channel estimation accuracy and data detection performance.

  5. Improved AI system can provide superior bit recovery performance compared to baseline schemes under CSN and LFM jamming, achieving significant SJR gains over existing methods in both TDL-A and TDL-D channels when integrated with the proposed JSF module.

Related papers