Performance Comparison of cGAN Models for Channel Estimation in One-Bit Massive MIMO System

In massive multiple-input multiple-output (MIMO) systems, deep learning (DL) methods exhibit a prominent role because they reduce the analytical complexity and improve the performance of wireless communication systems. Channel estimation becomes more challenging for a massive MIMO system in a highly unbalanced and movable environment. This paper investigates a DL-based channel estimation technique, namely conditional generative adversarial network (cGAN), with various generator model architectures to forecast more realistic and accurate channels. cGAN not only predicts the channels from quantized observations but also calculates the adaptive loss function. The simulation results confirm the accuracy of the DL-based channel estimation method. These results of this work have been compared with those of other previously reported DL methods in terms of mean square error.