Prompt-Guided Multi-Task Semantic Communication for Image Transmission

IF 4.4 3区计算机科学 Q2 TELECOMMUNICATIONS

IEEE Communications Letters Pub Date : 2026-02-10 DOI:10.1109/LCOMM.2026.3663440

Hua Zhang;Zijian Cao;Le Liang;Hao Ye;Shi Jin;Geoffrey Ye Li

引用次数: 0

Abstract

To enable multi-task semantic communication (SC) for image transmission, this letter proposes a prompt-guided multi-task SC system, termed PGMT-SC. PGMT-SC introduces a task-oriented hypernetwork that transforms natural-language description into prompt generator parameters, enabling flexible task adaptation without retraining the backbone. To enhance zero-shot generalization, we design a mixture-of-prompts (MOP) mechanism with Mixup training, which synthesizes prompts for unseen tasks by interpolating trained task embeddings. Simulation results under Rayleigh fading channels demonstrate that PGMT-SC outperforms existing schemes in multi-task performance and exhibits superior adaptability to unseen tasks.

查看原文本刊更多论文

基于提示引导的多任务语义通信图像传输

为了实现图像传输的多任务语义通信（SC），本文提出了一种快速引导的多任务语义通信系统，称为PGMT-SC。PGMT-SC引入了一个面向任务的超网络，将自然语言描述转换为提示生成器参数，无需重新训练主干即可实现灵活的任务适应。为了增强零概率泛化，我们设计了一个带有Mixup训练的混合提示（MOP）机制，该机制通过插值训练好的任务嵌入来合成未知任务的提示。在瑞利衰落信道下的仿真结果表明，PGMT-SC在多任务性能上优于现有方案，并且对未知任务具有较强的适应性。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

IEEE Communications Letters 工程技术-电信学

CiteScore

8.10

自引率

7.30%

发文量

590

审稿时长

2.8 months

期刊介绍： The IEEE Communications Letters publishes short papers in a rapid publication cycle on advances in the state-of-the-art of communication over different media and channels including wire, underground, waveguide, optical fiber, and storage channels. Both theoretical contributions (including new techniques, concepts, and analyses) and practical contributions (including system experiments and prototypes, and new applications) are encouraged. This journal focuses on the physical layer and the link layer of communication systems.