Generative Semantic Communication via Alternating Dual-Domain Posterior Sampling

TL;DR

This paper introduces ADDPS, a novel diffusion-based semantic communication receiver that alternates between latent and image domain guidance to enhance perceptual quality in wireless image transmission.

Contribution

It formulates semantic decoding as a Bayesian inverse problem and proposes an alternating dual-domain posterior sampling method to improve data distribution preservation.

Findings

ADDPS outperforms existing methods in perceptual quality on FFHQ dataset.

Alternating guidance avoids gradient conflicts and combines strengths of both domains.

Posterior sampling achieves optimal perceptual quality by preserving data distribution.

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.