17 November 2025 in IEEE Wireless Communications Letters
This letter presents a novel generative semantic communications framework for wireless image transmission, designed to excel in dynamic channel conditions and at low signal-to-noise ratios (SNRs). The proposed framework, dubbed DiffMoECom (Diffusion Mixture-of-Experts Communications), leverages a Mixture-of-Experts (MoE) architecture integrated with prompt learning to dynamically select optimal channel adaptation strategies by activating the most relevant prompt experts. A Joint Source and Channel Coding (JSCC) backbone is employed to transmit the low-frequency components of the source image. Furthermore, we introduce a range-space inversion mechanism, which projects the coarse JSCC output into a noisy latent space. This facilitates faithful semantic reconstruction of high-frequency details through conditional diffusion-based null space compensation. Extensive experiments across various fading channels and SNR conditions demonstrate that our method consistently surpasses state-of-the-art DeepJSCC and generative approaches, achieving superior performance in distortion metrics with competitive inference efficiency.
8 December 2025 in IEEE Transactions on Cognitive Communications and Networking
Emerging wireless video services demand efficient transmission schemes resilient to dynamic channel conditions while preserving semantic integrity. Conventional deep joint source-channel coding (DeepJSCC) faces challenges in jointly modeling spatio-temporal dependencies and adapting to bandwidth constraints. We propose Spatio-Temporal Video DeepJSCC (ST-VJSCC) that addresses these limitations through three key innovations. 1) A decomposed spatio-temporal Transformer architecture that separates 3D feature processing into 2D spatial and 1D temporal domains, reducing over 17% computational complexity compared to 3D ViT baselines; 2) A content-aware dynamic token selection mechanism enabling adaptive coding rates via tunable token retention ratio γ, achieving 40-60% bandwidth reduction while achieving only 1-2 dB decay in motion-related features; 3) A hierarchical feature fusion strategy combining multi-scale representations for enhanced reconstruction fidelity. Extensive evaluation demonstrate ST-VJSCC’s enables real-time deployment and is superior over H.264/H.265 with LDPC and state-of-the-art DeepJSCC methods. We further propose a Temporal Switching Smoothness (TSS) metric quantifying inter-frame consistency. The framework’s adaptability to AWGN and fading channels is validated through rigorous ablation studies, highlighting its potential for next-generation semantic video communications.
1 January 2026 in IEEE Communications Letters