Errors in channel prediction based on linear prediction in the frequency domain: A combination of frequency-domain and time-domain techniques

Channel prediction is one of the most important issues in dealing with a fading environment. If a future channel response is predicted, adaptive techniques, such as pre-equalization and beamforming, can be employed before transmission in order to avoid degradation of the quality of the communication. This paper proposes a highly accurate method for predicting time-varying channels by using a linear prediction of Fourier-domain channel parameters. The method combines frequency-domain and time-domain techniques. Simulation assuming a time-division duplex (TDD) system demonstrates that the proposed method very accurately predicts varying channels. The paper also presents the dependence of the prediction error on the length of the observation window, and on the distance between scatterers. It is found that our method reduces the prediction error even when the scatterers are located very closely together.