TY - GEN
T1 - Resilient Distributed Joint Source-Channel Coding with Non-Stationary Side Information
AU - Ma, Kangning
AU - Shao, Shuo
AU - Dai, Jincheng
AU - Zhong, Zhou
AU - Dai, Wenrui
AU - Xiong, Hongkai
N1 - Publisher Copyright:
© 2026 IEEE.
PY - 2026
Y1 - 2026
N2 - Transmission with side information (SI) delivery has become increasingly crucial in modern communications. Current distributed coding schemes, however, predominantly rely on noise-free or static SI link conditions, exhibiting limited robustness against dynamic SI transmission states prevalent in practical networks. Hence, we consider a Wyner-Ziv (WZ) transmission scenario with non-stationary SI delivery induced by adversarial jamming attacks. A resilient deep joint source and channel coding (DeepJSCC) framework is developed to guarantee reliable image transmissions. Under this method, we adopt a hierarchical coding approach to allow robust semantic transmission under unpredictable interference conditions. Complementarily, an intelligent discriminator capable of recognizing attack patterns is designed to guide adaptive image recovery at the receiver site after transmission. We train the DeepJSCC model using a customized multi-stage training strategy. Experimental results show that the learned scheme adapts well to the SI transmission dynamics and outperforms conventional schemes in the non-stationary WZ scenarios.
AB - Transmission with side information (SI) delivery has become increasingly crucial in modern communications. Current distributed coding schemes, however, predominantly rely on noise-free or static SI link conditions, exhibiting limited robustness against dynamic SI transmission states prevalent in practical networks. Hence, we consider a Wyner-Ziv (WZ) transmission scenario with non-stationary SI delivery induced by adversarial jamming attacks. A resilient deep joint source and channel coding (DeepJSCC) framework is developed to guarantee reliable image transmissions. Under this method, we adopt a hierarchical coding approach to allow robust semantic transmission under unpredictable interference conditions. Complementarily, an intelligent discriminator capable of recognizing attack patterns is designed to guide adaptive image recovery at the receiver site after transmission. We train the DeepJSCC model using a customized multi-stage training strategy. Experimental results show that the learned scheme adapts well to the SI transmission dynamics and outperforms conventional schemes in the non-stationary WZ scenarios.
UR - https://www.scopus.com/pages/publications/105049100705
U2 - 10.1109/ISIT62367.2026.11653986
DO - 10.1109/ISIT62367.2026.11653986
M3 - 会议稿件
AN - SCOPUS:105049100705
T3 - IEEE International Symposium on Information Theory - Proceedings
BT - ISIT 2026 - 2026 IEEE International Symposium on Information Theory, Symposium Proceedings
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2026 IEEE International Symposium on Information Theory, ISIT 2026
Y2 - 28 June 2026 through 3 July 2026
ER -