Deep Learning-Based channel estimation with SRGAN in OFDM Systems
Siqiang Zhao, Yuan Fang, Ling Qiu
Abstract
In this paper, we propose a novel deep learning-based channel estimation scheme in an orthogonal frequency division multiplexing (OFDM) system. The channel response with known pilot positions can be treated as a low-resolution image. Then, we explore a generative adversarial network (GAN) for channel super-resolution (SR) to estimate the whole channel state information (CSI). For previous deep learning-based channel estimators recovered by a single model, high-frequency details are missing and they fail to match the fidelity expected at the higher resolution. The scheme we proposed is more consistent with the real channel by adding a discriminator to recover more details of the channel. The simulation results show that our scheme is superior to other SR-based channel estimation methods and close to the linear minimum mean square error (LMMSE) performance.