Channel Estimation Method Based on Transformer in High Dynamic Environment

Abstract

As the relative movement speeds of the communication parties increase, the Doppler frequency offset gradually increases, and the speed of channel state information(CSI) change also increases, which limits the performance of traditional channel estimation algorithms. To solve the above problems, we propose a channel estimation structure based on Transformer. Convolutional Neural Network (CNN) is used to extract the feature vectors of channel response and Transformer is used for channel estimation. Utilize Transformer's deep learning capabilities to better track channel variation characteristics in highly dynamic environments. By simulation, we get results that compared with traditional channel estimation methods, the performance of the proposed channel estimation method is improved significantly under high dynamic environment.