An ICVBPNN Algorithm for Time-varying Channel Tracking and Prediction

Abstract

An improved complex-valued back propagation neural network (ICVBPNN) algorithm is proposed in this paper. In allusion to the defect of gradient descent of traditional complex-valued back propagation network (CVBPNN) algorithm, additive momentum has been introduced. It is used for time-varying channel tracking and prediction in wireless communication system and better application results are acquired. Firstly, with the use of the learning ability of the neural network, the tracking training is started based on the obtained channel state information (CSI), thus the nonlinear channel model is constructed. Secondly, the unknown channel state information is predicted using the ICVBPNN trained model. The simulation results demonstrate that the proposed method has less estimated error, and can track the channel more accurately than the traditional CVBPNN and the Kalman Filter algorithm.