This paper investigates distributed source-channel coding for correlated image semantic communication, where different but correlated image sources from distributed transmitters are separately encoded and transmitted through dedicated wireless channels to a common receiver. We propose a novel distributed nonlinear transform coding (NTC)-based approach to explicitly model source correlation from both probabilistic and geometric perspectives. The probabilistic perspective is achieved by a joint entropy model that approximates the joint distribution of latent representations and guides adaptive rate allocation, while the geometric perspective is achieved by a feature alignment module that aligns latent features for maximal correlation learning at the decoder.We implement this distributed NTC framework for both source coding only, referred to as D-NTSC, and joint source-channel coding (JSCC), referred to as D-NTSCC. Variational inference is employed to derive principled loss functions that jointly optimize encoding, decoding, and joint entropy modeling. Extensive experiments on real-world multi-view datasets demonstrate that D-NTSC and D-NTSCC outperform existing distributed source coding and distributed JSCC baselines, respectively, achieving state-of-the-art performance in both pixel-level and perceptual quality metrics.

We consider multi-user semantic communications over broadcast channels. While most existing works consider that each receiver requires either the same or independent semantic information, this paper explores the scenario where the semantic information desired by different receivers is different but correlated. In particular, we investigate semantic communications over Gaussian broadcast channels where the transmitter has a common observable source but the receivers wish to recover hierarchical semantic information in adaptation to their channel conditions. Inspired by the capacity achieving property of superposition coding, we propose a deep learning-based superposition coded modulation (DeepSCM) scheme. Specifically, the hierarchical semantic information is first extracted and encoded into basic and enhanced feature vectors. A linear minimum mean square error (LMMSE) decorrelator is then developed to obtain a refinement from the enhanced features that is uncorrelated with the basic features. Finally, the basic features and their refinement are superposed for broadcasting after probabilistic modulation. Extensive experiments are conducted for two-receiver image semantic broadcasting with coarse and fine classification as hierarchical semantic tasks. DeepSCM outperforms the benchmarking coded-modulation scheme without a superposition structure as well as the classic separate source-channel coding baselines, especially with large channel disparity and high order modulation. It also approaches the performance upperbound as if there were only one receiver.

Semantic communications have emerged as a new paradigm for improving communication efficiency by transmitting the semantic information of a source message that is most relevant to a desired task at the receiver. Most existing approaches typically utilize neural networks (NNs) to design end-to-end semantic communication systems, where NN-based semantic encoders output continuously distributed signals to be sent directly to the channel in an analog fashion. In this work, we propose a joint coding-modulation (JCM) framework for digital semantic communications by using variational autoencoder (VAE). Our approach learns the transition probability from source data to discrete constellation symbols, thereby avoiding the non-differentiability problem of digital modulation. Meanwhile, by jointly designing the coding and modulation process together, we can match the obtained modulation strategy with the operating channel condition. We also derive a matching loss function with information-theoretic meaning for end-to-end training. Experiments on image semantic communication validate the superiority of our proposed JCM framework over the state-of-the-art quantization-based digital semantic coding-modulation methods across a wide range of channel conditions, transmission rates, and modulation orders. Furthermore, its performance gap to analog semantic communication reduces as the modulation order increases while enjoying the hardware implementation convenience.

This paper investigates digital semantic communications over broadcast channels, where the transmitter has a common observable source to transmit, but the receivers desire to recover different levels of semantic information in adaptation to their respective channel conditions. Inspired by the capacity-achieving superposition coding structure for degraded broadcast channel, we propose a superposition coded modulation (SCM) scheme for semantic information transmission over a two-receiver Gaussian broadcast channel. In specific, the transmitter consists of two separate neural network (NN)-based semantic encoders, one NN-based linear minimum mean square error (LMMSE) decorrelator, and one superposition-based digital modulator. The SCM scheme enables the receiver with lower signal-to-noise ratio (SNR) to decode basic semantic information while the receiver with larger SNR can decode enhanced semantic information. Experiments are conducted on image transmission with coarse and fine classifications as semantics. Results show that our proposed SCM scheme outperforms the basic coded-modulation scheme without a superposition structure for both receivers, especially when channel conditions of the two receivers differ significantly. It also approaches the performance upperbound as if there were only one receiver.