Channel estimation improvement in massive MIMO systems based on pilot decontamination

Abstract

This paper considers multi-cell multi-user multi-input multi-output (MIMO) systems with a very large antenna array at base station. In massive MIMO systems, pilot contamination degrades the performance of uplink (UL) channel estimation. It has been shown that data-aided channel estimation method can achieve a better performance than pilot-aided channel estimation method. In this paper, to mitigate the effects of pilot contamination, we propose a data-aided semi-blind channel estimation method based on independent component analysis (ICA) that does not require cell cooperation or statistical information of the channels. Based on demodulated UL data of all users in the target cell, we finally obtain the least square (LS) estimate of the channel. This demodulated data contains a short training sequence just for ambiguity correction. In order to obtain a better performance compared to the previous works, we use mutual information as a new criterion in ICA. Simulation results show that the proposed method is more reliable and more robust against outliers and has less bit error probability compared to the previous works.