CSI feedback algorithm for massive MIMO systems based on SFNet
To address the issues of high computational complexity, low feedback accuracy, and neglect of quantization loss in existing deep learning-based channel state information (CSI) feedback methods for frequency-division duplex massive multiple-input multiple-output (MIMO) systems, the deep learning algo...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | Chinese |
| Published: |
Editorial Department of Journal on Communications
2025-06-01
|
| Series: | Tongxin xuebao |
| Subjects: | |
| Online Access: | http://www.joconline.com.cn/zh/article/doi/10.11959/j.issn.1000-436x.2025097/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | To address the issues of high computational complexity, low feedback accuracy, and neglect of quantization loss in existing deep learning-based channel state information (CSI) feedback methods for frequency-division duplex massive multiple-input multiple-output (MIMO) systems, the deep learning algorithm SFNet for CSI feedback was proposed. SFNet integrated a traditional convolutional neural network (CNN) and Transformer architecture, incorporating a spatial-frequency block designed to leverage global information and a multi-scale adaptive spatial attention gate for fusing local and global features. Fast Fourier convolution and a dynamic feature fusion mechanism were utilized to activate more input information, adjust the receptive field, selectively highlight spatially correlated features, suppress interference, and allow the network to achieve advanced performance with extremely low computational complexity. The experimental results show that the proposed algorithm achieves advanced estimation performance with significantly low computational complexity. Furthermore, the trained model exhibits strong robustness across various environments. |
|---|---|
| ISSN: | 1000-436X |