Performance Bounds Analysis for Semi-Blind Channel Estimation with Pilot Contamination in Massive MIMO-OFDM Systems Ouahbi Rekik * , Abdelhamid Ladaycia * , Karim Abed-Meraim † and Anissa Mokraoui * * L2TI, Institut Galilée, Université Paris 13 Sorbonne Paris Cité, France {ouahbi.rekik, abdelhamid.ladaycia, anissa.mokraoui}@univ-paris13.fr † PRISME, Université d’Orléans, France karim.abed-meraim@univ-orleans.fr Abstract—Pilot contamination, in a massive Multiple-Input Multiple-Output (MIMO) system, is an undeniable challenging issue severely affecting the performance of the system by including channel estimation errors. The aim of this paper is to investigate the effectiveness of semi-blind channel estimation approaches in massive MIMO-OFDM (Orthogonal-Frequency Division-Multiplexing) systems. For an estimator-independent study, the performance analysis is carried out using the Cramér Rao Bound (CRB) derivation for pilot-based and semi-blind channel estimation strategies. This analysis demonstrates in particular that: (i) when considering the finite alphabet nature of communication signals, it is possible to efficiently solve the pilot contamination problem with semi-blind channel estima- tion approach; and (ii) the Second Order Statistics (SOS) only are not sufficient to address the full channel identifiability even if the semi-blind approach is considered. 1. Introduction Massive Multiple-Input Multiple-Output (MIMO) is a promising technology for the next generation cellular net- works [1]. With a higher number of Base Station (BS) antennas (beyond 100 antennas), compared to the classical MIMO systems, massive MIMO technology has proven its ability to improve the spectral and power efficiency [2]. So that, both throughput and reliability will be highly enhanced for the future cellular networks. In order to fully exploit all of the potentials offered by a massive MIMO system, accurate Channel State Information (CSI) is necessary. It is obtained only during the uplink transmission, thanks to the channel reciprocity property and according to the widely accepted Time Division Duplexing (TDD) protocol. The traditional methods to get the CSI rely on the pilot-based channel estimation (see e.g. [1]). How- ever, due to the non-orthogonality of the pilot sequences, these methods are severely affected by what is called pilot contamination [3]. It is one of the major issues of massive MIMO systems that must be addressed because its effect cannot be reduced by increasing the number of BS antennas. Many pilot contamination mitigation strategies have been proposed: creating more orthogonal pilots by slicing the time and frequency resources [4], subspace projection-based interference suppressing [5], data-aided channel estimation [6], or by designing appropriate inter-cell communication protocols and resource allocation. In recent works (e.g. [7], [8]), a particular attention has been drawn to blind and semi- blind methods. The former is fully based on the statistical properties of the transmitted data, whereas the latter depends on the joint use of pilots and data. The focus of this paper falls into the scope of perfor- mance analysis of semi-blind channel estimation with pilot contamination in the context of multi-cell massive MIMO- OFDM systems. For an estimator-independent performance analysis, the Cramér Rao Bound (CRB) is derived for both pilot-based and semi-blind channel estimation. Note that a thorough study has been conducted in [9] where the achievable performance of semi-blind approaches compared to pilot-based ones has been quantified for channel es- timation in a MIMO-OFDM system. This study will be extended to a massive MIMO-OFDM system, by taking into account the multi-cell context and the phenomenon of pilot contamination. The rest of this paper is organized as follows. Section 2 describes the massive MIMO-OFDM system model. Section 3 illustrates the pilot contamination effect. The derivation of the CRB for pilot-based and semi-blind approaches are detailed in section 4. Simulation results are discussed in section 5. Finally, Section 6 concludes the work. 2. Massive MIMO-OFDM system model This section presents the massive MIMO-OFDM wire- less system adopted in this paper. An uplink transmission is considered. The system is composed of N c cells where each cell contains one BS with N r antennas and N t users using each a single antenna. The received signals from the adjacent cells are initially ignored. Therefore the received signal at the r-th BS antenna of the l-th cell, assumed to be a K sub-carriers OFDM signal, is given by [9]: y l,r = Nt  i=1 F T (h l,i,r ) F H K x l,i + v l,r , (1) where the superscript (.) H denotes the Hermitian operator; K is the OFDM symbol length; F represents a K-point 2018 26th European Signal Processing Conference (EUSIPCO) ISBN 978-90-827970-1-5 © EURASIP 2018 1277