Semantic communication (SC) is a communication paradigm that has gained significant attention, as it offers a potential solution to move beyond Shannon’s formulation in bandwidth-limited communication channels by delivering the semantic meaning of the message rather than its exact form. In this paper, we propose an autoencoder-based SC system for transmitting images between two machines over error-prone channels to support emerging applications such as VIoT, XR, M2M, and M2H communications. The proposed autoencoder architecture, with a semantically modeled encoder and decoder, transmits image data as a reduced-dimension vector (latent vector) through an error-prone channel. The decoder then reconstructs the image to determine its M2M implications. The autoencoder is trained for different noise levels under various channel conditions, and both image quality and classification accuracy are used to evaluate the system’s efficacy. A CNN image classifier measures accuracy, as no image quality metric is available for SC yet. The simulation results show that all proposed autoencoders maintain high image quality and classification accuracy at high SNRs, while the autoencoder trained with zero noise underperforms other trained autoencoders at moderate SNRs. The results further indicate that all other proposed autoencoders trained under different noise levels are highly robust against channel impairments. We compare the proposed system against a comparable JPEG transmission system, and results reveal that the proposed system outperforms the JPEG system in compression efficiency by up to $50\%$ and in received image quality with an image coding gain of up to 17 dB.

Semantic communication (SemCom) systems enhance transmission efficiency by conveying semantic information in lieu of raw data. However, challenges arise when designing these systems due to the need for robust semantic source coding for information representation extending beyond the training dataset, maintaining channel-agnostic performance, and ensuring robustness to channel and semantic noise. We propose a novel generative artificial intelligence (AI) based SemCom architecture conditioned on quantized latent. The system reduces the communication overhead of the wireless channel by transmitting the index of the quantized latent over the communication channel by mapping the quantized vector to the learned codebook vectors. The learned codebook is the shared knowledge base. The encoder is designed with a novel spatial attention mechanism based on image energy, focusing on object edges. The critic assesses the realism of generated data relative to the original distribution, with the Wasserstein distance. The model introduces novel contrastive objectives at multiple levels, including pixel, latent, perceptual, and task output, tailored for noisy wireless semantic communication. We validated the proposed model for transmission quality and robustness with low-density parity-check (LDPC), which outperforms the baselines of better portable graphics (BPG), specifically at low signal-to-noise ratio (SNR) levels ( $ {\lt }~ {5}$ dB). Additionally, it shows comparable results with joint source-channel coding (JSCC) with lower complexity and latency. The model is validated for human perception and machine perception-oriented task utility. The model effectively transmits high-resolution images without requiring additional error correction at the receiver. We propose a novel semantic-based matrix to evaluate the robustness to noise and task-specific semantic distortion.

The rapid growth of multimedia content in consumer electronics devices in edge scenarios, such as smart camera sensors, autonomous navigation, and wireless security systems, necessitates more efficient communication systems. Semantic communication (SemCom) systems enhance transmission efficiency by focusing on semantic content. However, design challenges persist, including limitations in generalizing semantic source coding, maintaining channel-agnostic performance, and ensuring robustness against channel noise. We propose a novel SemCom architecture leveraging vector quantization(VQ) and edge inference to enhance semantic content representation. This architecture models the underlying data distribution, enabling efficient and accurate predictions of unseen data. By transmitting only the indices of relevant codebook vectors extracted at the edge device, our SemCom architecture minimizes communication overhead while maintaining robust semantic content representation. The system focuses on two receiver tasks, reconstruction and classification. Evaluations demonstrate the system’s superior performance over better portable graphics (BPG) and joint source-channel coding (JSCC) in terms of perceptual quality, with low-density parity-check (LDPC) coding across both additive white Gaussian noise (AWGN) and Rayleigh flat fading channels, as well as classification accuracy with polar coding over the AWGN channel. Experimental results reveal a 25% reduction in transmission overhead compared to JSCC with LDPC coding with a code rate of 0.5 for varying signal-to-noise ratio (SNR)s, while maintaining lower overall computational complexity. Additionally, the model is validated using out-of-distribution and high-resolution datasets, validating its adaptability and scalability to diverse consumer electronics scenarios. The classification model achieved 85.52% training accuracy, with 72.25% and 50.88% accuracy at high and low SNR levels, respectively.

The increasing dominance of video traffic in intelligent sensing and control applications introduces a major challenge to the capacity limits of modern wireless networks. Classical information theory defines fixed physical boundaries on channel capacity, beyond which further improvement requires rethinking what information is transmitted. Semantic communication (SemCom) bridges this by only sending the semantics of the intended message. This paper presents a SemCom framework that leverages latent-space procedural video prediction with world-model-guided temporal dynamics. Instead of transmitting pixel data, the transmitter encodes high-level semantic representations of context frames and sends them through the physical channel. A temporal transformer predicts future latent states at the receiver. The framework jointly optimizes perceptual, adversarial, and temporal objectives to preserve both visual quality and trajectory consistency under channel impairments. Experiments conducted on video sequences of the BAIR robot pushing dataset demonstrate that the proposed method achieves lower normalized endpoint and velocity errors compared to learned baseline real-time intermediate flow estimation (RIFE) + better portable graphics (BPG) with reduced bit-rate compared to traditional codecs. The results indicate that incorporating temporal dynamics as semantics into the communication process enables efficient and anticipatory video transmission suitable for applications such as tele-robotic, underwater, and autonomous systems.

Semantic communication (SemCom) systems enhance transmission efficiency by conveying semantic information in lieu of raw data. However, challenges arise when designing these systems due to the need for robust semantic source coding for information representation extending beyond the training dataset, maintaining channel-agnostic performance, and ensuring robustness to channel and semantic noise. We propose a novel generative artificial intelligence (AI) based SemCom architecture conditioned on quantized latent. The system reduces the communication overhead of the wireless channel by transmitting the index of the quantized latent over the communication channel by mapping the quantized vector to the learned codebook vectors. The learned codebook is the shared knowledge base. The encoder is designed with a novel spatial attention mechanism based on image energy, focusing on object edges. The critic assesses the realism of generated data relative to the original distribution, with the Wasserstein distance-enhancing training stability. We trained the model on a novel dataset consisting of images from Flickr8K, DIV2K, and CelebA-HQ, with data augmentation applied to enhance generalization across diverse distributions. The proposed SemCom system outperforms the traditional scheme with low-density parity-check (LDPC) channel coding across various channel coding rates and noise levels. Notably, at low Signal-to-noise ratio (SNR) levels (<5 dB) SemCom system outperforms better portable graphics (BPG). Additionally, it shows comparable performance with joint source-channel coding (JSCC). The model effectively transmits high-resolution images without additional error correction at the receiver and achieves superior performance in human-face image transmission. We propose a novel semantic-based matrix to evaluate the robustness to noise and task-specific semantic distortion, as traditional matrices lack meaningful evaluation for SemCom systems, which the novel SemCom system outperforms conventional methods.

Semantic communications focus on transmitting information that encapsulates meaning, enabling both machines and humans to understand the intended message with greater accuracy. Unlike traditional communication systems, which send data without considering its semantic value, this approach prioritises the content's meaning and requires a novel metric to gauge semantic quality. Our framework integrates a specialised Vision Transformer (ViT) for semantic segmentation, named SemExT, at the transmission end and a pre-trained Generative Adversarial Network (GAN) for image reconstruction at the receiving end. The system's effectiveness is evaluated by comparing the semantic content of the reconstructed image with the original, using Deceptron2, an advanced object detection model. This comparison establishes a new metric for assessing the quality of semantic transmission. Empirical evidence shows that the semantic quality metric ranges from 90% to 100% for images containing fewer objects and 80% to 98% for those with more objects. In comparison, an autoencoder-based communication system exhibits a range of 80% to 100% for simpler images and 75% to 95% for more complex ones. These findings highlight the robustness of our proposed metric across different semantic communication frameworks, contributing to the advancement of semantic information transmission and setting a foundation for future research in this field.