Channel estimation of 5G system based on deep learning and residual connectivity network
Xiuping Song, Youlong Gong, Zengchun Yang · 2025
This paper proposes a 5G wireless link channel estimation network model based on deep learning and residual connection network, to address the problem of reduced channel estimation accuracy caused by noise interference and the difficulty of traditional channel estimation algorithms in meeting practical application requirements. The time-frequency response of the 5G communication channel is treated as a two-dimensional image. Firstly, set the parameters to generate a channel dataset for the physical downlink shared channel based on the 5G NR standard, and then convert the generated channel dataset into a two-dimensional image; Then, a channel estimation model based on convolutional neural networks is constructed, introducing residual connections to solve the degradation problem in deep networks, in order to improve network performance. After batch normalization of the data, the impact of gradient vanishing is reduced; Finally, use the trained network model for channel estimation. The simulation results show that under a wide range of SNR testing conditions, the proposed channel estimation algorithm has significantly improved performance compared to practical estimation algorithms and interpolation estimation algorithms, and has good generalization ability.