EUVIMEDEUVIMED entdecken
Crossref

VQ-DSC-R: Robust Vector Quantized-Enabled Digital Semantic Communication With OFDM Transmission

Jianqiao Chen, Nan Ma, Xiaodong Xu, Tingting Zhu, Huishi Song, Chen Dong, Wenkai Liu, Rui Meng, Ping Zhang

IEEE Transactions on Cognitive Communications and Networking · 2026 · 12 · 10060-10077

DOI: 10.1109/tccn.2026.3712380 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Cognitive Communications and Networking, 2026. Verfasst von Jianqiao Chen und 8 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 49 Literaturangaben. Gefördert von: National Key Research and Development Program of China, Lenovo-XNET Cooperative Project, 6G Key Technologies R&D and Ecosystem Cultivation Project.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Digital semantic communication promises seamless integration with existing infrastructure, yet faces critical challenges in reconciling continuous semantic features with discrete transmission constraints over multipath fading channels. Existing vector quantization (VQ) schemes suffer from codebook collapse and gradient mismatch, while channel estimation errors induce catastrophic decoding failures. To address these, we propose VQ-DSC-R, a robust digital semantic communication framework built upon orthogonal frequency division multiplexing (OFDM). First, we design a hierarchical semantic codec utilizing a Swin Transformer backbone, coupled with a novel differentiable VQ with adaptive noise-variance (ANDVQ) scheme. ANDVQ dynamically calibrates quantization boundaries using K-nearest neighbor statistics and stabilizes training via exponential moving average, resolving the long-standing codebook under-utilization issue. Second, we develop a conditional diffusion model (CDM) for channel state information (CSI) refinement. By incorporating SNR-adaptive noise scheduling and a gated fusion mechanism, the CDM accurately recovers CSI from sparse pilots within five sampling steps, effectively mitigating multipath distortion. Third, a three-stage training strategy is introduced to decouple semantic quantization, channel adaptation, and joint optimization, ensuring stable convergence. Extensive experiments on the 3GPP EPA channel demonstrate that VQ-DSC-R achieves superior performance, attaining a 29.22dB peak signal-to-noise ratio (PSNR) and 0.959 multiscale structural similarity (MS-SSIM) at a 0.02 bit compression ratio (BCR). Compared to conventional JPEG+LDPC schemes, our framework reduces bandwidth consumption by 80% while avoiding the cliff effect.

Förderung

  • National Key Research and Development Program of China
  • Lenovo-XNET Cooperative Project
  • 6G Key Technologies R&D and Ecosystem Cultivation Project
  • China Postdoctoral Science Foundation

Zitieren

APA

Chen, J., Ma, N., Xu, X., Zhu, T., Song, H., Dong, C., Liu, W., Meng, R., & Zhang, P. (2026). VQ-DSC-R: Robust Vector Quantized-Enabled Digital Semantic Communication With OFDM Transmission. IEEE Transactions on Cognitive Communications and Networking, 12, 10060-10077. https://doi.org/10.1109/tccn.2026.3712380

Vancouver

Chen J, Ma N, Xu X, Zhu T, Song H, Dong C, et al. VQ-DSC-R: Robust Vector Quantized-Enabled Digital Semantic Communication With OFDM Transmission. IEEE Transactions on Cognitive Communications and Networking. 2026;12:10060-10077. doi:10.1109/tccn.2026.3712380

BibTeX
@article{Chen2026,
  author = {Chen, Jianqiao and Ma, Nan and Xu, Xiaodong and Zhu, Tingting and Song, Huishi and Dong, Chen and Liu, Wenkai and Meng, Rui and Zhang, Ping},
  title = {VQ-DSC-R: Robust Vector Quantized-Enabled Digital Semantic Communication With OFDM Transmission},
  journal = {IEEE Transactions on Cognitive Communications and Networking},
  year = {2026},
  volume = {12},
  pages = {10060--10077},
  doi = {10.1109/tccn.2026.3712380}
}
Kostenlos starten →