Channel Estimation of OFDM System Based on BR-SRGAN Network
Fuhui Yu, Jun Chang, Jiaqi Wei, Yunxiao Wu, Dong Li · 2023
In the existing channel estimation methods of OFDM systems, a naive deep learning-based algorithm cannot fulfill the expected fidelity and high-frequency details are varnished at higher resolutions. A Generative Adversarial Network algorithm based on super resolution techniques and bilinear residual is proposed in this paper, to estimate traditional channel through the image processing. Firstly, the training dataset is generated by the Physical Downlink Shared Channel of the 5G New Radio standard. Then, converting the generated channel matrix of the pilot frequency position to a low-resolution image. And, utilizing neural network model to substitute the channel estimation interpolation processing. The MSE of our method is able to consistently reach 10−4, indicating that the algorithm we proposed has outstanding performance and prominent robustness than the traditional algorithms, such as LS, Practical Channel Estimation, ChannelNet and ReEsNet network.