Long Short-Term Memory Network for Wireless Channel Prediction

In modern wireless systems, channel prediction is an effective way to overcome the feedback delay of channel state information (CSI). When the receiver performs adaptive transmission based on the feedback CSI, the channel prediction algorithm can reduce the system overhead by predicting the future CSI. In this paper, we provide a long short-term memory (LSTM) network for wireless channel prediction. This method can get a smaller prediction error than other intelligence methods. Experiments show that the LSTM model has a lower normalized mean square error (NMSE) and less running time than support vector machine, artificial neural network, and recurrent neural network prediction approaches.