Attention Guided Dynamic D-JSCC Framework for Semantic Communication System

The integration of semantic communication systems with advanced attention mechanisms marks a significant evolution in wireless communication, particularly in challenging environments like underwater and satellite systems. This paper proposes an Attention-Guided Dynamic Deep Joint Source-Channel Coding (D-JSCC) framework that leverages Convolutional Block Attention Modules (CBAM) to enhance the efficiency and reliability of data transmission. By dynamically prioritizing semantically meaningful features, the proposed framework addresses bandwidth constraints and noise challenges in low Signal-to-Noise Ratio (SNR) conditions. Evaluations on high-resolution datasets, such as DIV2K and Places365, demonstrate significant improvements in image reconstruction quality, achieving superior Peak Signal-to-Noise Ratio (PSNR) performance, especially under adverse conditions. The system’s adaptability to varying SNR levels and its ability to optimize resource utilization make it suitable for real-world applications, including environmental monitoring, disaster response, and maritime safety. Future directions include exploring dynamic switching mechanisms and expanding the framework’s applicability to other domains, such as intelligent transportation systems and medical imaging. This study sets a benchmark for robust, scalable, and semantically aware communication systems tailored for the next generation of wireless networks.

• Supervision