Wideband Beamspace Channel Estimation via Conditional GAN s

Channel estimation is a challenging task, especially for wideband beamspace systems with beam squint effect. While the conventional deep learning (DL) methods can achieve the mapping of the input to the realistic channel, the estimation performance is limited due to the common loss function. In this paper, we propose a wideband beamspace channel estimation scheme based on conditional generative adversarial networks (cGANs), which not only can achieve the mapping from measurements to the channels, but also can adaptively perform the loss function to estimate the channel more accurately. Simulation results show that the proposed scheme outperforms existing DL methods with higher robustness.