WTSCNet: A CSI feedback network based on wavelet transform convolution and attention mechanism

Yong-Li An; Shuo-Yang Lu; Yi-Xiao Liu; Ying-Chao Liu; Qiang Kong

doi:10.1016/j.jnlest.2026.100367

Journal of Electronic Science and Technology ›› 2026, Vol. 24 ›› Issue (3) :100367 DOI: 10.1016/j.jnlest.2026.100367

Regular Papers

research-article

WTSCNet: A CSI feedback network based on wavelet transform convolution and attention mechanism

Yong-Li An ¹^,²^,^*
, Shuo-Yang Lu ¹^,²
, Yi-Xiao Liu ¹^,²
, Ying-Chao Liu ¹^,²
, Qiang Kong ¹^,²

Author information +

¹ College of Artificial Intelligence, North China University of Science and Technology, Tangshan, 063210, China

² Hebei Key Laboratory of Industrial Intelligent Perception, Tangshan, 063210, China

^* E-mail addresses: anyl@ncst.edu.cn (Y.-L. An),

lsyleaon@stu.ncst.edu.cn (S.-Y. Lu),

yxliu@stu.ncst.edu.cn (Y.-X. Liu),

liuyc@ncst.edu.cn (Y.-C. Liu),

kq3515@stu.ncst.edu.cn (Q. Kong).

Yong-Li An received the Ph.D. degree from Beijing Jiaotong University, Beijing, China in 2015. She is currently a Professor with North China University of Science and Technology, Tangshan, China. She was a Visiting Scholar at Oakland University, Rochester, USA in 2011 and Griffith University, Brisbane, Australia in 2018. Her current research interests include key technologies of wireless communication physical layer, intelligent communication network, applications in the field of unmanned aerial vehicle (UAV) communication and authentication, and key technologies for UAV system channels.

Shuo-Yang Lu was born in Shijiazhuang, China in 2001. He obtained the M.S. degree from North China University of Science and Technology in 2026. His research interest mainly focuses on channel state information (CSI) feedback for massive multiple-input multiple-output (MIMO) systems.

Yi-Xiao Liu was born in Shijiazhuang, China in 2001. She received the B.S. degree from Qiqihar University, Qiqihar, China in 2019 and the M.S. degree from North China University of Science and Technology in 2026. Her main research interests include end-to-end communication systems and channel modeling.

Ying-Chao Liu received the Ph.D. degree from Yanshan University, Qinhuangdao, China in 2020. She is currently a Professor with North China University of Science and Technology. She was a Visiting Scholar at Northumbria University, Newcastle, UK in 2025. Her current research interest mainly focuses on intelligent optical fiber sensing technology integrating physical information neural networks.

Qiang Kong was born in Jining, China in 2002. He received the B.S. degree from Linyi University, Linyi, China in 2024. He is currently pursuing the M.S. degree at North China University of Science and Technology. His research interest mainly focuses on the radio frequency fingerprint security authentication of UAVs.

Show less

History +

PDF (10986KB)

Abstract

In order to enhance the performance of massive multiple-input multiple-output (MIMO) systems, the downlink channel state information (CSI) must be efficiently compressed and fed back to the base station in a frequency-division-duplex mode. The deep-learning-based CSI feedback networks can learn more complicated patterns in the data as compared with the traditional compressed-sensing techniques. Nevertheless, many of the current methods are founded on the convolutional neural network (CNN) modules that were initially developed to act on images and might not make full use of relevant channel and spatial information. The suggested model, wavelet transform convolution and spatial channel self-attention-channel network (WTSCNet), is a combination of wavelet-transform convolution and attention mechanisms. The encoder also has access to wavelet-transform convolution to generate multi-resolution features, enabling it to perform better in both space information and local details at a reduced computational cost. On the decoder part, it is presented as a spatial channel self-attention-channel reconstruction block (CARBlock) which combines multi-scale residual reconstruction with spatial-channel attention to improve the recovery of CSI. Experiments prove that WTSCNet enhances the indoor reconstruction precision by 6.41 dB, compared with CRNet, at a compression ratio of 1/4. In comparison with the Transformer-based TransNet+, WTSCNet requires about 9.9 million floating-point operations less than TransNet+ but offers a 0.28-dB improvement in the indoor normalized mean squared error (NMSE) at the same compression ratio. The simulation results indicate that WTSCNet offers a good accuracy-complexity trade-off between CSI compression and feedback.

Keywords

CSI feedback / Frequency division duplex / Massive MIMO systems / WTConv

Cite this article

Download citation ▾

Yong-Li An, Shuo-Yang Lu, Yi-Xiao Liu, Ying-Chao Liu, Qiang Kong. WTSCNet: A CSI feedback network based on wavelet transform convolution and attention mechanism. Journal of Electronic Science and Technology, 2026, 24 (3) : 100367 DOI:10.1016/j.jnlest.2026.100367

登录浏览全文

4963

注册一个新账户忘记密码

References

Publishing order | Descend order by publishing year | Descend order by cited within

[1]	P.-H. Kuo, H.T. Kung, P.-A. Ting, Compressive sensing based channel feedback protocols for spatially-correlated massive antenna arrays, in: Proc. of the IEEE Wireless Communications and Networking Conf., Paris, France, 2012, pp. 492-497.

[2]	C.A. Metzler, A. Maleki, R.G. Baraniuk, From denoising to compressed sensing, IEEE T. Inform. Theory 62 (9) (2016) 5117-5144.

[3]

C.-B. Li, W.-T. Yin, Y. Zhang, User’s guide for TVAL3: TV minimization by augmented Lagrangian and alternating direction algorithms, Version beta2.4 [Online], Department of Computational and Applied Mathematics, Rice University, Houston, USA (2010-11-11), https://usermanual.wiki/Document/UserGuide.2125550948.pdf.

[4]	Y.-H. Zhang, Y.-F. Mu, Y. Liu, et al., Deep learning-based beamspace channel estimation in mmWave massive MIMO systems, IEEE Wirel. Commun. Le. 9 (12) (2020) 2212-2215.

[5]	H.-T. He, S. Jin, C.-K. Wen, et al., Model-driven deep learning for physical layer communications, IEEE Wirel. Commun. 26 (5) (2019) 77-83.

[6]	J.-J. Guo, C.-K. Wen, S. Jin, et al., Overview of deep learning-based CSI feedback in massive MIMO systems, IEEE T. Commun. 70 (12) (2022) 8017-8045.

[7]	C.-K. Wen, W.-T. Shih, S. Jin, Deep learning for massive MIMO CSI feedback, IEEE Wirel. Commun. Le. 7 (5) (2018) 748-751.

[8]	J.-J. Guo, C.-K. Wen, S. Jin, et al., Convolutional neural network-based multiple-rate compressive sensing for massive MIMO CSI feedback: design, simulation, and analysis, IEEE T. Wirel. Commun. 19 (4) (2020) 2827-2840.

[9]	Z.-L. Lu, J.-T. Wang, J. Song, Multi-resolution CSI feedback with deep learning in massive MIMO system, in: Proc. of the IEEE Intl. Conf. on Communications, Dublin, Ireland, 2020, pp. 1-6.

[10]	Q.-Y. Cai, C. Dong, K. Niu, Attention model for massive MIMO CSI compression feedback and recovery, in: Proc. of the IEEE Wireless Communications and Networking Conf., Marrakesh, Morocco, 2019, pp. 1-5.

[11]	S.-J. Ji, M. Li, CLNet: complex input lightweight neural network designed for massive MIMO CSI feedback, IEEE Wirel. Commun. Le. 10 (10) (2021) 2318-2322.

[12]	B.-H. Wang, Y.-L. Teng, V. Lau, et al., CCA-Net: a lightweight network using criss-cross attention for CSI feedback, IEEE Commun. Le. 27 (7) (2023) 1879-1883.

[13]	Y. Xu, M.-Q. Yuan, M.-O. Pun, Transformer empowered CSI feedback for massive MIMO systems, in: Proc. of the 30th Wireless and Optical Communications Conf., Taipei, China, 2021, pp. 157-161.

[14]	Y.-D. Cui, A.-H. Guo, C.-L. Song, TransNet: full attention network for CSI feedback in FDD massive MIMO system, IEEE Wirel. Commun. Le. 11 (5) (2022) 903-907.

[15]	Q. Chen, A.-H. Guo, Y.-D. Cui, Multi-head transformer architecture with higher dimensional feature representation for massive MIMO CSI feedback, Applied Sciences 14 (4) (2024) 1356.

[16]	Y.-L. An, S.-Y. Lu, H.-R. Cai, et al., A deep learning-based approach to lightweight CSI feedback, Phys. Commun.-AMST 68 (2025) 102538.

[17]	J.-X. Li, Y.-J. Wang, Z. Liu, et al., MetaFormer-based lightweight neural network for massive MIMO CSI feedback, IEEE Wirel. Commun. Le. 14 (2) (2025) 275-279.

[18]	X.-Y. Cen, C.-T. Lam, B.K. Ng, et al., When Transformer meets CSI feedback in mMIMO systems: a lightweight CsiMobileViT approach, in: Proc. of the IEEE 102nd Vehicular Technology Conf., Chengdu, China, 2025, pp. 1-6.

[19]	Y. Liu, G. Yang, T. Li, et al., Nighttime flare removal via wavelet-guided and gated-enhanced spatial-frequency fusion network, in: Proc. of the 40th AAAI Conf. on Artificial Intelligence, Singapore, 2026, pp. 7404-7412.

[20]	H. Zhou, Y.-L. Wang, Q. Zhang, et al., A dual-stage residual diffusion model with perceptual decoding for remote sensing image dehazing, IEEE T. Geosci. Remote 63 (2025) 4109312.

[21]	Y. Liu, X.-R. Wang, E.-P. Hu, et al., VNDHR: variational single nighttime image dehazing for enhancing visibility in intelligent transportation systems via hybrid regularization, IEEE T. Intell. Transp. 26 (7) (2025) 10189-10203.

[22]	Y.J. Kwon, A. Lee, H. Lee, CSI feedback compression based on deep learning using wavelet transform, in: Proc. of the 14th Intl. Conf. on Information and Communication Technology Convergence, Jeju Island, Republic of Korea, 2023, pp. 1091-1094.

[23]	Y.-Z. Si, H.-Y. Xu, X.-Z. Zhu, et al., SCSA: exploring the synergistic effects between spatial and channel attention, Neurocomputing 634 (2025) 129866.