Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ICDM: Interference Cancellation Diffusion Models for Wireless Semantic Communications

Tong Wu, Zhiyong Chen, Dazhi He, Feng Yang, Meixia Tao, Xiaodong Xu, Wenjun Zhang, Ping Zhang

TL;DR

This work addresses interference in wireless semantic communications by formulating interference cancellation as a MAP estimation problem over the joint posterior of signal and interference. It introduces ICDM, a diffusion-model–based framework that uses two independent priors to model signal and interference and combines their log-gradients with a Bayes-derived channel likelihood via an efficient ODE-based sampler (p-ConJPC) to recover the clean signal. The authors prove an upper-bound on the joint estimation error and demonstrate substantial performance gains on multiple datasets and channels, notably reducing MSE and improving perceptual metrics (eg, a 4.54 dB MSE reduction and a 2.47 dB LPIPS gain on CelebA under Rayleigh fading at SINR 0 dB) while achieving faster inference than conventional diffusion methods. ICDM is compatible with various JSCC architectures, offering a practical approach for rapid, accurate interference cancellation in wireless semantic communication systems.

Abstract

Diffusion models (DMs) have recently achieved significant success in wireless communications systems due to their denoising capabilities. The broadcast nature of wireless signals makes them susceptible not only to Gaussian noise, but also to unaware interference. This raises the question of whether DMs can effectively mitigate interference in wireless semantic communication systems. In this paper, we model the interference cancellation problem as a maximum a posteriori (MAP) problem over the joint posterior probability of the signal and interference, and theoretically prove that the solution provides excellent estimates for the signal and interference. To solve this problem, we develop an interference cancellation diffusion model (ICDM), which decomposes the joint posterior into independent prior probabilities of the signal and interference, along with the channel transition probablity. The log-gradients of these distributions at each time step are learned separately by DMs and accurately estimated through deriving. ICDM further integrates these gradients with advanced numerical iteration method, achieving accurate and rapid interference cancellation. Extensive experiments demonstrate that ICDM significantly reduces the mean square error (MSE) and enhances perceptual quality compared to schemes without ICDM. For example, on the CelebA dataset under the Rayleigh fading channel with a signal-to-noise ratio (SNR) of $20$ dB and signal to interference plus noise ratio (SINR) of 0 dB, ICDM reduces the MSE by 4.54 dB and improves the learned perceptual image patch similarity (LPIPS) by 2.47 dB.

ICDM: Interference Cancellation Diffusion Models for Wireless Semantic Communications

TL;DR

This work addresses interference in wireless semantic communications by formulating interference cancellation as a MAP estimation problem over the joint posterior of signal and interference. It introduces ICDM, a diffusion-model–based framework that uses two independent priors to model signal and interference and combines their log-gradients with a Bayes-derived channel likelihood via an efficient ODE-based sampler (p-ConJPC) to recover the clean signal. The authors prove an upper-bound on the joint estimation error and demonstrate substantial performance gains on multiple datasets and channels, notably reducing MSE and improving perceptual metrics (eg, a 4.54 dB MSE reduction and a 2.47 dB LPIPS gain on CelebA under Rayleigh fading at SINR 0 dB) while achieving faster inference than conventional diffusion methods. ICDM is compatible with various JSCC architectures, offering a practical approach for rapid, accurate interference cancellation in wireless semantic communication systems.

Abstract

Diffusion models (DMs) have recently achieved significant success in wireless communications systems due to their denoising capabilities. The broadcast nature of wireless signals makes them susceptible not only to Gaussian noise, but also to unaware interference. This raises the question of whether DMs can effectively mitigate interference in wireless semantic communication systems. In this paper, we model the interference cancellation problem as a maximum a posteriori (MAP) problem over the joint posterior probability of the signal and interference, and theoretically prove that the solution provides excellent estimates for the signal and interference. To solve this problem, we develop an interference cancellation diffusion model (ICDM), which decomposes the joint posterior into independent prior probabilities of the signal and interference, along with the channel transition probablity. The log-gradients of these distributions at each time step are learned separately by DMs and accurately estimated through deriving. ICDM further integrates these gradients with advanced numerical iteration method, achieving accurate and rapid interference cancellation. Extensive experiments demonstrate that ICDM significantly reduces the mean square error (MSE) and enhances perceptual quality compared to schemes without ICDM. For example, on the CelebA dataset under the Rayleigh fading channel with a signal-to-noise ratio (SNR) of dB and signal to interference plus noise ratio (SINR) of 0 dB, ICDM reduces the MSE by 4.54 dB and improves the learned perceptual image patch similarity (LPIPS) by 2.47 dB.

Paper Structure

This paper contains 14 sections, 3 theorems, 52 equations, 10 figures, 2 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. System Model and Theorem Formulation
  3. An overview of the system
  4. Theorem for Interference Cancellation
  5. Challenges of Solving the MAP Problem
  6. Interference Cancellation Diffusion Models
  7. Training Algorithm of ICDM for Prior Gradient Estimation
  8. Deriving Expression for Guidance Gradient Estimation
  9. Interference Cancellation Algorithm of ICDM
  10. Experimental Results
  11. Experimental Setup
  12. Performance Evaluation
  13. Ablation Study
  14. Conclusion

Key Result

Lemma 1

The real-valued vectors $\mathbf{x}$, $\mathbf{y}$ and $\mathbf{z}$ satisfy where $\mathbf{n}\sim \mathcal{N}(0,\frac{\sigma^2}{2}\mathbf{I}_{2k})$. For the Rayleigh fading channel with MMSE equalizer, $\mathbf{H}=diag()$, $\mathbf{H_x^r}=diag(Re(\mathbf{h_x}))$, $\mathbf{H_x^I}=diag(Im(\mathbf{h_x}))$, $\mathbf{W_z}=(\mathbf{H}^2+\sigma^2\mathbf{I})^{-1}$, $\mathbf{W_s}=\m

Figures (10)

  • Figure 1: The overall architecture of wireless semantic communications system.
  • Figure 2: The overall structure of the proposed ICDM.
  • Figure 3: The overall sampling process of ICDM. $\mathbf{x}_t$ and $\mathbf{z}_t$ are invisible signals and we decode them into images here for illustration.
  • Figure 4: The illustration of the JCG module.
  • Figure 5: MSE, LPIPS and CLIP performance of the MambaJSCC-based schemes on the CelebA dataset under both the AWGN and Rayleigh fading channels.
  • ...and 5 more figures

Theorems & Definitions (7)

  • Lemma 1
  • proof
  • Theorem 1
  • proof
  • Proposition 1
  • proof
  • Remark 1