Application of improved P-BEM in time varying channel prediction in 5G high-speed mobile communication system

Abstract To improve people’s communication experience on high-speed trains, this study proposed a time-varying channel prediction (TVCP) method based on an improved polynomial basis extension model and backpropagation neural network. The improved polynomial-basis expansion model proposed in this study extracts the optimal basis function by constructing a channel correlation matrix and performing singular value decomposition to adapt to high-speed mobile channel changes. By using these basis functions and pilot signals to estimate historical basis coefficients as training data for back propagation neural network, future channel states can be predicted to improve the accuracy of TVCP in high-speed mobile communication systems. The results showed that when the training sample of the prediction method was 2,000, the maximum mean square error before improvement was close to 0.01, the maximum after improvement was 4.0 × 10 −4 . After increasing the normalized Doppler frequency shift to 0.5, the mean square error of the improved method was around 0.0001, while other methods were above 0.001. The improvement of the TVCP method effectively reduced the mean square error of TVCP, the prediction accuracy of the improved prediction method was much superior than that of traditional channel prediction methods. The designed method can greatly perfect the prediction accuracy of time-varying channels and enhance people’s communication experience on high-speed trains in high-speed mobile scenarios.