ConvLSTM based Channel Prediction for 5G Network

With the aim to obtain accurate and timely channel state information (CSI) in the 5G system, a convolutional long short term memory (ConvLSTM) based channel prediction method is proposed. From the measured channel characteristics, it is found that the CSI on the operating frequency of 5G system has low correlation in the time domain, but the correlation in the frequency domain is high. Therefore, a joint time-frequency channel prediction method is proposed to improve the accuracy of channel prediction. Experiments on the measured channel data show that, compared to the existing channel prediction methods, the proposed method has a great improvement in the prediction accuracy, and the normalized mean square error (NMSE) is reduced to half of the existing methods.