Attention-Based Semantic Communication Systems for Artificial Intelligence of Things

With the rapid growth in demand for intelligent services and the corresponding surge in massive data, semantic communication has emerged as a promising paradigm for intelligent service applications. However, existing approaches are limited by their capability to effectively extract semantic features. To address this challenge and improve the quality of transmitted image, this paper proposes an attention-based semantic communication system for image transmission. Specifically, by adaptively weighting spatial and channel features through the Convolutional Block Attention Module (CBAM), the proposed approach enhances the representation of crucial information. Furthermore, focusing on Artificial Intelligence of Things (AIoT) applications, where intelligent tasks require autonomous decision-making based on perceived information, the system is trained on two receiver tasks: reconstruction and classification, enabling it to adapt to diverse application. Experimental results demonstrate that our proposed method outperforms both deep learning based joint source and channel coding (JSCC) and the combination of joint photographic experts group(JPEG) and low-density parity-check (LDPC) in terms of image reconstruction quality. Furthermore, the proposed system achieves high classification accuracy and exhibits robustness across varying compression ratios.