A novel real-time channel prediction algorithm in high-speed scenario using convolutional neural network

The accurate channel state information (CSI) is important to realize the high-reliability and high-efficiency transmission. So far, most of the conventional methods reconstruct CSI on non-RS position by the interpolation. However, high-speed mobility leads to the significant nonlinear channel variation and the performance of conventional methods is so poor that the reliability of transmission deteriorates a lot. In this paper, a novel real-time channel prediction algorithm based on convolutional neural network (CNN) is proposed, which uses the latest reference signal (RS) for online training and extracts the temporal features of channel, followed by prediction employing the optimal model. For high-speed moving scenario, the proposed algorithm is conducted in Orthogonal Frequency Division Multiplexing (OFDM) systems, e.g., long-term evolution (LTE) and fifth generation (5G) systems to track the fast time-varying and non-stationary channel via the real-time RS-based training algorithm, and obtains the accurate CSI without modification of the radio frame. Evaluated by experiments, the proposed algorithm outperforms conventional methods a lot, and more improvement could be achieved in the higher speed.