CNN-based Doppler Shift Estimation for Low Earth Orbit Satellites

This paper proposes a convolutional neural network (CNN)-based Doppler shift estimation model for low earth orbit (LEO) satellite communication systems. We propose a deep learning model which estimates the Doppler shift caused by the high orbital velocity of the satellite. The proposed model extracts the channel feature with the CNN and is trained to minimize the error with the ground truth Doppler shift value. In numerical results, it is shown that the proposed model can accurately estimate the Doppler shift that the satellite communication channel experienced.