Robust Deep Joint Source-Channel Coding for Video Transmission over Multipath Fading Channel

Bin Xiao, Jian Zou, Fanyang Meng, Wei Liu, Yibo Liang2025 · DOI

TL;DR

This paper introduces a novel DeepJSCC framework designed for robust video transmission over challenging multipath fading channels. By integrating OFDM for modulation, a conditional contextual coding mechanism for temporal redundancy, and a denoising module in the decoder, the proposed method significantly improves video reconstruction quality and convergence speed compared to existing state-of-the-art approaches.

What Problem Does This Paper Address?

The increasing demand for high-quality video streaming in future wireless systems (like 6G) faces significant hurdles. Traditional separate source and channel coding (SSCC) methods are hitting their performance limits and suffer from a "cliff effect" where performance degrades sharply under poor channel conditions. Emerging semantic communication, particularly DeepJSCC, offers a promising solution by jointly optimizing source compression and channel coding for better efficiency and robustness. However, existing DeepJSCC frameworks often struggle with complex environments like multipath fading channels, which introduce frequency-selective fading and signal distortion. This paper aims to bridge this gap by proposing a DeepJSCC framework specifically engineered to be robust against these challenging channel conditions, ensuring reliable video transmission even when bandwidth and power are constrained.

Methods

The proposed framework introduces several key innovations:

  • OFDM Modulation: Orthogonal Frequency Division Multiplexing (OFDM) is employed at the modulation stage. This technique decomposes the wideband signal into multiple orthogonal sub-carriers, each experiencing relatively flat fading. This effectively mitigates frequency-selective fading inherent in multipath channels and implicitly supports channel state learning through embedded pilot symbols, eliminating the need for explicit channel state information (CSI).
  • Conditional Contextual Coding: To better exploit temporal redundancy in video, the paper moves beyond traditional pixel-domain residual coding. Instead, it uses a conditional contextual coding mechanism that leverages multi-scale Gaussian warped features from reference frames. This approach captures richer temporal and semantic dependencies, leading to more efficient compression of inter-frame redundancy, especially under strict bandwidth constraints.
  • Denoising-Aided Joint Decoding: A significant bottleneck in conventional DeepJSCC decoders is the simultaneous handling of channel estimation, equalization, denoising, and semantic reconstruction. This paper proposes decoupling these tasks by introducing a lightweight denoising module. This module preprocesses the received signal using pilot-derived channel information, simplifying the subsequent decoding process, accelerating convergence, and improving reconstruction quality.

Datasets

The framework is trained and evaluated using the UCF101 dataset, a widely-used benchmark for video action recognition, containing over 13,000 videos across 101 action categories.

Key Insights and Ideas

  • OFDM's Dual Role: OFDM isn't just for mitigating fading; its pilot symbols are cleverly used for implicit channel estimation, simplifying the decoder's job.
  • Beyond Pixel Differences: The shift from pixel-based residual coding to feature-domain contextual coding is crucial for capturing complex temporal relationships in video, leading to better compression.
  • Modular Decoding for Efficiency: Separating denoising from the main decoding pipeline significantly reduces complexity and speeds up convergence, a critical factor for real-time video transmission.
  • Cumulative Gains: Each proposed module (OFDM, contextual coding, denoising) contributes independently to the overall performance improvement, demonstrating a well-designed, synergistic system.

Key Results

The proposed framework achieves significant performance gains:

  • Overall Reconstruction Quality: An average reconstruction quality gain of 5.13 dB is reported compared to state-of-the-art video DeepJSCC methods under challenging multipath fading conditions.
  • Module Contributions:
    • OFDM modulation provides an average gain of 2.76 dB.
  • The conditional contextual coding mechanism adds an improvement of 1.8 dB.
  • The lightweight denoising module further enhances performance by 0.57 dB.
  • Convergence Speed: The integrated denoising module significantly accelerates convergence, as evidenced by the MSE loss curves.
  • Robustness: The framework demonstrates superior performance over baseline methods (like DeepWiVe) in multipath fading environments, with PSNR increasing from 22.03 dB to 27.16 dB.

Limitations and Future Work

  • Limitations: The paper acknowledges that while the proposed framework is robust, it still relies on a fixed compression rate. The computational complexity, though improved by the denoising module, remains a factor.
  • Future Work: The authors suggest exploring advanced entropy models for fine-grained, data-driven bandwidth allocation. This could further enhance system robustness and efficiency by dynamically adapting to inter-frame information and channel conditions.

Summary updated

Abstract

To address the challenges of wireless video transmission over multipath fading channels, we propose a robust deep joint source-channel coding (DeepJSCC) framework by effectively exploiting temporal redundancy and incorporating robust innovations at the modulation, coding, and decoding stages. At the modulation stage, tailored orthogonal frequency division multiplexing (OFDM) for robust video transmission is employed, decomposing wideband signals into orthogonal frequency-flat sub-channels to effectively mitigate frequency-selective fading. At the coding stage, conditional contextual coding with multi-scale Gaussian warped features is introduced to efficiently model temporal redundancy, significantly improving reconstruction quality under strict bandwidth constraints. At the decoding stage, a lightweight denoising module is integrated to robustly simplify signal restoration and accelerate convergence, addressing the suboptimality and slow convergence typically associated with simultaneously performing channel estimation, equalization, and semantic reconstruction. Experimental results demonstrate that the proposed robust framework significantly outperforms state-of-the-art video DeepJSCC methods, which achieves an average reconstruction quality gain of 5.13 dB under challenging multipath fading channel conditions<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup>.
BibTeX
@misc{xiao2025robust,
  author = {Bin Xiao and Jian Zou and Fanyang Meng and Wei Liu and Yibo Liang},
  title = {{Robust Deep Joint Source-Channel Coding for Video Transmission over Multipath Fading Channel}},
  year = {2025},
  doi = {10.1109/globecom59602.2025.11432569},
}

Loading related content...

Figure