MADNet: based on multiple attention and dilated convolution for CSI feedback

Xiang Zhao; Chuansheng Yang; Chao Wang; Xiaoshuang Wang

doi:10.1117/12.2686378

10 August 2023 MADNet: based on multiple attention and dilated convolution for CSI feedback

Xiang Zhao, Chuansheng Yang, Chao Wang, Xiaoshuang Wang

Proceedings Volume 12759, International Conference on Automation Control, Algorithm, and Intelligent Bionics (ACAIB 2023); 127592Q (2023) https://doi.org/10.1117/12.2686378
Event: 2023 3rd International Conference on Automation Control, Algorithm and Intelligent Bionics (ACAIB 2023), 2023, Xiamen, China

Abstract

The massive multiple-input and multiple-output (MIMO) system based on channel state information (CSI) is the core technology of next-generation communication. As the complexity of the CSI matrix gradually increases, CSI feedback becomes more challenging. CSI feedback technology based on deep learning (DL) has been successful in frequency-division duplex (FDD) MIMO systems. In this paper, we propose a complex-valued lightweight neural network MADNet for CSI feedback. The network is based on an encoder-decoder structure, which enriches the information extraction of the CSI matrix by adopting channel information extraction modules and coordinates information extraction modules, which reduces the computational complexity of the network by using lightweight convolution. Experimental results show that our network outperforms most CNN-based network architectures at multiple compression rates and performs significantly better in both indoor and outdoor scenarios.

Citation Download Citation

Xiang Zhao, Chuansheng Yang, Chao Wang, and Xiaoshuang Wang "MADNet: based on multiple attention and dilated convolution for CSI feedback", Proc. SPIE 12759, International Conference on Automation Control, Algorithm, and Intelligent Bionics (ACAIB 2023), 127592Q (10 August 2023); https://doi.org/10.1117/12.2686378

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available