Wiener-based smoother and predictor for massive-MIMO downlink system under pilot contamination

Abstract

One of the challenges in massive-MIMO system is pilot contamination during the channel estimation process. Pilot contamination can cause error or inaccurate channel estimation process for future fifth generation (5G) downlink transmissions. This paper considers using a Wiener-based filter to smooth and predict the channel estimation to reduce the pilot contamination for more accurate CSI during channel estimation. The simulation results show that the Wiener-based smoothing and predicting technique reduces the effect of pilot contamination and increases the accuracy of CSI during channel estimation process. Wiener smoother (WS) is implemented based on Wiener-based filtering technique. The previous estimated CSI and weight coefficient vector are used to smooth the current estimated CSI by using block data formulation to reduce the effect of pilot contamination. However, WS technique suffers from pilot contamination due to pilot training. This motivates the development of two Wiener predictors (WP), known as WP1 and WP2. The WP1 and WP2 run a prediction technique for CSI and number of pilot training during the prediction period, which is missing from the original WS. Comparison results show that the proposed WS and WP outperforms the conventional minimum mean square error and least square, in terms of channel estimation error and per-cell rate. WP2 perform better than WS and WP1 because of the algorithm complexity that required more information to be updated, stored and processed for prediction. Thus, WP2 requires large computation and matrix operation compared to WS and WP1. The results indicate that the channel estimation error due to pilot contamination can be reduced by using the Wiener-based approaches.