Semantic Adaptive Communication Based on Double-Attention Phase and Compress Estimator for Wireless Image...

Area of Science:

• Computer Science
• Electrical Engineering
• Information Theory

Background:

• Existing semantic communication systems struggle with low image quality and high transmission rates.
• There is a need for advanced semantic communication to address these challenges in image transmission.

Purpose of the Study:

• To propose a Semantic Adaptive Communication (SAC) framework for efficient image transmission.
• To enhance image reconstruction quality and optimize transmission rates.

Main Methods:

• Developed a SAC framework with Semantic Encoder (SE), Semantic Decoder (SD), Semantic Code Generator/Restore (SCG/SCR), Compression Estimator (CE), and Channel State Information Acquisition (CSIA).
• Integrated a Double-Attention Module (DAM) into SE and SD to capture channel and spatial attention for semantic features.
• Utilized CE to predict compression rates based on channel conditions and desired recovery quality.

Main Results:

• The SAC framework achieved higher Peak Signal-to-Noise Ratio (PSNR) (0.5-2 dB increase) and accuracy (91-93%) compared to traditional and other semantic communication methods.
• Demonstrated robustness and improved performance in image transmission scenarios.
• Achieved adaptive transmission rates with minimal loss in recovery performance.

Conclusions:

• The proposed SAC framework significantly enhances image transmission quality and efficiency.
• The system offers adaptive transmission capabilities, optimizing bandwidth utilization.
• SAC represents a critical advancement in semantic communication for image transmission.