Channel Estimation in RIS-Assisted Wireless Communication System Based on Attention-Based Convolutional Neural Network

In RIS-assisted wireless communication systems, accurate channel estimation is vital for achieving efficient communication and precise sensing. Traditional methods like least squares and minimum mean square error often exhibit limited accuracy in complex environments. Deep learning approaches, particularly convolutional neural networks (CNNs), offer a significant advantage by automatically extracting intricate features from channel data, thereby enhancing estimation performance. Moreover, incorporating attention mechanisms allows the network to prioritize critical channel features, leading to further improvements in estimation accuracy. In this study, we develop and train an attention-based network model. By inputting actual pilot signals and their corresponding received signals into this trained model, we obtain accurate channel estimates for the RIS-assisted wireless communication system. Our results demonstrate that this scheme achieves rapid convergence and yields a lower average MSE loss compared to existing methods, indicating superior estimation accuracy.