Generative Semantic Communication via Alternating Dual-Domain Posterior Sampling

2026-04-17

AI Summary

This work addresses the limitations of existing generative semantic communication receivers, which rely on maximum a posteriori estimation and struggle to preserve data distribution. Single-domain guidance in diffusion-based methods is susceptible to noise in latent space and decoder bias in image space. To overcome these issues, the paper formulates semantic decoding as a Bayesian inverse problem and proposes Alternating Dual-Domain Posterior Sampling (ADDPS), which alternately enforces consistency constraints in both latent and image spaces during the diffusion process. This approach is the first to theoretically guarantee that posterior sampling can simultaneously maintain the original data distribution and achieve optimal perceptual quality. By alternating guidance between domains, ADDPS effectively avoids gradient conflicts and leverages the complementary strengths of both spaces. Experiments on the FFHQ dataset demonstrate significant improvements over state-of-the-art methods in perceptual quality for image transmission.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Sampling/Simulation-based Search
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Security and Privacy: Data transparency and provenance

Abstract

Generative semantic communication (SemCom) harnesses pretrained generative priors to improve the perceptual quality of wireless image transmission. Existing generative SemCom receivers, however, rely on maximum a posteriori (MAP) estimation, which fundamentally cannot preserve the data distribution and thus limits achievable perceptual quality. Moreover, current diffusion-based approaches using single-domain guidance face significant limitations: latent-domain guidance is sensitive to channel noise, while image-domain guidance inherits decoder bias. Simply combining both domains simultaneously yields an overconfident pseudo-posterior. In this paper, we formulate semantic decoding as a Bayesian inverse problem and prove that posterior sampling achieves optimal perceptual quality by preserving the data distribution. Building on this insight, we propose alternating dual-domain posterior sampling (ADDPS), a diffusion-based SemCom receiver that alternately enforces latent-domain and image-domain consistency during the sampling process. This alternating strategy decomposes joint posterior sampling into simpler subproblems, avoiding gradient conflicts while retaining the complementary strengths of both domains. Experiments on FFHQ demonstrate that the proposed ADDPS achieves superior perceptual quality compared with existing methods.

Problem

• generative semantic communication
• posterior sampling
• perceptual quality
• diffusion-based
• dual-domain guidance

Innovation

• generative semantic communication
• posterior sampling
• dual-domain guidance
• diffusion-based decoding
• Bayesian inverse problem