DD-JSCC: Dynamic Deep Joint Source-Channel Coding for Semantic Communications

TL;DR

DD-JSCC introduces a flexible, adaptive deep joint source-channel coding architecture for semantic image transmission that dynamically adjusts to varying conditions, improving performance and reducing training costs.

Contribution

It proposes a novel hierarchical, dynamically adjustable encoder-decoder architecture for Deep-JSCC, enabling real-time adaptation to device and channel variations.

Findings

Achieves up to 2 dB PSNR improvement over fixed models.

Reduces training costs by over 40%.

Eliminates the need for multiple specialized models.

Abstract

Deep Joint Source-Channel Coding (Deep-JSCC) has emerged as a promising semantic communication approach for wireless image transmission by jointly optimizing source and channel coding using deep learning techniques. However, traditional Deep-JSCC architectures employ fixed encoder-decoder structures, limiting their adaptability to varying device capabilities, real-time performance optimization, power constraints and channel conditions. To address these limitations, we propose DD-JSCC: Dynamic Deep Joint Source-Channel Coding for Semantic Communications, a novel encoder-decoder architecture designed for semantic communication systems. Unlike traditional Deep-JSCC models, DD-JSCC is flexible for dynamically adjusting its layer structures in real-time based on transmitter and receiver capabilities, power constraints, compression ratios, and current channel conditions. This adaptability is achieved through a hierarchical layer activation mechanism combined with implicit regularization via sequential randomized training, effectively reducing combinatorial complexity, preventing overfitting, and ensuring consistent feature representations across varying configurations. Simulation results demonstrate that DD-JSCC enhances the performance of image reconstruction in semantic communications, achieving up to 2 dB improvement in Peak Signal-to-Noise Ratio (PSNR) over fixed Deep-JSCC architectures, while reducing training costs by over 40%. The proposed unified framework eliminates the need for multiple specialized models, significantly reducing training complexity and deployment overhead.