Joint Semantic-Channel Coding and Modulation for Generative Image Transmission
We propose SemQAM, a joint semantic-channel coding and modulation scheme for generative image transmission converting source to modulated symbols, such as standard QAM symbols.
ProofPaper ↗
Key points
- Digital image transmission typically optimizes source coding, channel coding, and modulation separately, limiting how source representations adapt to channel impairments.
- First, SemQAM represents latent features with vector quantization (VQ), maps each VQ index to a QAM symbol, and updates the codeword corresponding to the detected symbol so that channel impairments directly affect codebook learning.
- Moreover, SemQAM uses the conditional expectation of received codewords to approximate the receiver reconstruction and reduce error propagation during autoregressive coding.
- Experiments show that (i) under the same LDPC and QAM settings, SemQAM-C outperforms OneDC in LPIPS and DISTS with substantially lower complexity, showing that the unified model retains competitive capability for source coding. (ii) Compared with this separated counterpart, SemQAM reduces channel uses by 18.57% at matched LPIPS and outperforms the compared JSCC methods, demonstrating the benefit of joint coding.
Sources (1)
- [1]Joint Semantic-Channel Coding and Modulation for Generative Image TransmissionarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 21, 12:25 PM
We propose SemQAM, a joint semantic-channel coding and modulation scheme for generative image transmission converting source to modulated symbols, such as standard QAM symbols.
Digital image transmission typically optimizes source coding, channel coding, and modulation separately, limiting how source representations adapt to channel impairments.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 21, 2026PACE: Periodic and Channel-Aligned Encoding of Four-Bit Recurrent States
- Oct 20, 2026Do Diffusion LLMs Scale? Testing Mercury 3 + Decide
- Oct 20, 2026Meet Drex Max: Nace AI’s Multimodal Decision Model for Computer-Use Agents with a Score 67.59
- Oct 20, 2026AnyLanguageModel 1.0: One Year of Local Models on Apple Platforms
- Oct 20, 2026Specter: Fast MoE Inference via Quantization-Enabled Multi-Step Speculative Offloading
- Oct 19, 2026Amazon SageMaker JumpStart: 2026 year-to-date in review