Multiscale Semantic Compression for Robust Collaborative CNN Inference in Low-SNR Environments: An Attention-Enhanced UNet Autoencoder
- 23 February 2026
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Internet of Things Journal
- Vol. 13 (10) , 21755-21770
Abstract
In collaborative inference scenarios, semantic communication replaces raw data transmission by conveying task-oriented semantic features to improve bandwidth efficiency. However, under noisy wireless channels, the combined effects of semantic compression distortion and channel noise lead to severe information loss, resulting in degraded inference accuracy. To address this issue, this article proposes a multiscale semantic compression collaborative inference (MSCCI) framework that achieves efficient, stable inference performance under high compression ratios (CRs) and elevated noise levels. Specifically, a UNet-based encoder extracts multiscale semantic features on an IoT device. These features are then integrated into a unified stream using a novel semantic fusion compression strategy, thereby substantially reducing communication overhead. The edge server decoder decompresses features and recovers image semantics via progressive upsampling and multiscale semantic restoration. For noisy wireless channels, the framework incorporates squeeze-and-excitation (SE) attention for dynamic feature channel weighting and residual connections for enhanced low-SNR robustness. Experimental results demonstrate that our collaborative inference framework for semantic communication outperforms state-of-the-art algorithms, and the approach’s effectiveness and robustness are verified across various channel conditions.Keywords
Funding Information
- Natural Science Foundation of Beijing Municipality (4262021)
- National Natural Science Foundation of China (61202076)
- National Natural Science Foundation of China (62202019)
This publication has 41 references indexed in Scilit:
- Lightweight transformer-driven multi-scale trapezoidal attention network for saliency detectionEngineering Applications of Artificial Intelligence, 2025
- NORRIS: Noise-Resilient and Resource-Efficient Semantic Encoded Point Cloud Data Transmission for Internet of Things CommunicationsIEEE Internet of Things Journal, 2025
- Bilateral Feature Fusion with hexagonal attention for robust saliency detection under uncertain environmentsInformation Fusion, 2025
- Toward AI-Powered Edge Intelligence for Object Detection in Self-Driving Cars: Enhancing IoV Efficiency and SafetyIEEE Internet of Things Journal, 2025
- Application of Convolutional Neural Networks for Parallel Multi-Scale Feature Extraction in Noise Image DenoisingIEEE Access, 2024
- Rep ViT: Revisiting Mobile CNN From ViT Perspective2024
- A ConvNet for the 2020s2022
- Deep-Reinforcement-Learning-Based Collision Avoidance in UAV EnvironmentIEEE Internet of Things Journal, 2021
- Semi-Supervised Semantic Segmentation with Cross Pseudo Supervision2021
- Coordinate Attention for Efficient Mobile Network Design2021