Review: Independent component analysis for multiple-input multiple-output wireless communication systems

  • Authors:

  • J. Gao;X. Zhu;A. K. Nandi

  • Affiliations:

  • Department of Electrical Engineering and Electronics, The University of Liverpool, Brownlow Hill, Liverpool L69 3GJ, UK;Department of Electrical Engineering and Electronics, The University of Liverpool, Brownlow Hill, Liverpool L69 3GJ, UK;Department of Electrical Engineering and Electronics, The University of Liverpool, Brownlow Hill, Liverpool L69 3GJ, UK

  • Venue:
  • Signal Processing
  • Year:
  • 2011

Quantified Score

Hi-index 0.08

Visualization

Abstract

Independent component analysis (ICA), an efficient higher order statistics (HOS) based blind source separation technique, has been successfully applied in various fields. In this paper, we provide an overview of the applications of ICA in multiple-input multiple-output (MIMO) wireless communication systems, and introduce some of the important issues surrounding them. First, we present an ICA based blind equalization scheme for MIMO orthogonal frequency division multiplexing (OFDM) systems, with linear precoding for ambiguity elimination. Second, we discuss three peak-to-average power ratio (PAPR) reduction schemes, which do not introduce any spectral overhead. Third, we investigate the application of ICA to blind compensation for inphase/quadrature (I/Q) imbalance in MIMO OFDM systems. Finally, we present an ICA based semi-blind layer space-frequency equalization (LSFE) structure for single-carrier (SC) MIMO systems. Simulation results show that the ICA based equalization approach provides a much better performance than the subspace method, with significant PAPR reduction. The ICA based I/Q compensation approach outperforms not only the previous compensation methods, but also the case with perfect channel state information (CSI) and no I/Q imbalance, due to additional frequency diversity obtained. The ICA based semi-blind LSFE receiver outperforms its OFDM counterpart significantly with a training overhead of only 0.05%.