Blind and semi-blind equalisation for Multiple-Input Multiple-Output wireless communication systems using Independent Component Analysis

Sarperi, Luciano Blind and semi-blind equalisation for Multiple-Input Multiple-Output wireless communication systems using Independent Component Analysis. Doctor of Philosophy thesis, University of Liverpool.

	PDF thesis_sarperi.pdf - Author Accepted Manuscript Access to this file is embargoed until Unspecified. Available under License Creative Commons Attribution No Derivatives. Download (1MB)
	PDF (renamed file) SarperiLuc_Jun2007_429.pdf - Author Accepted Manuscript Access to this file is embargoed until Unspecified. Available under License Creative Commons Attribution No Derivatives. Download (1MB)

Abstract

This thesis proposes novel blind and semi-blind receivers for Multiple-Input Multiple-Output (MIMO) wireless communication systems employing Orthogonal Frequency Division Multiplexing (OFDM) or Single-Carrier Frequency Domain Equalisation (SCFDE) modulation. MIMO systems can achieve higher data rates than Single-Input Single-Output (SISO) systems while the use of OFDM and SC-FDE is motivated by their simple equalisation of frequency selective channels. Furthermore, the blind or semi-blind operation increases the bandwidth efciency compared to traditional training based receivers, which is particularly interesting in MIMO systems since they require more training data than SISO systems. Blind and semi-blind equalisation in this work is performed using the statistical tool of Independent Component Analysis (ICA), which relies on the independence of the source streams. First, novel blind receivers for MIMO OFDM systems are proposed. Two linear receivers, which have different complexity / performance tradeoffs are presented. The indeterminacies of ICA estimates are handled by post-processing of the estimates, which relies on precoding of the source streams. Subsequently, these receivers are combined with Layered Space-Frequency Equalisation (LSFE), which results in performance close to the case with perfect Channel State Information (CSI) at the receiver. Reduced-complexity versions are also considered, which are essential for OFDM systems employing a large number of subcarriers. Furthermore, performance and computational complexity comparisons with other published blind receivers show the superiority of the proposed methods. Next, semi-blind receivers for MIMO OFDM systems are proposed, which are derived from the blind methods above. The advantage of operating semi-blindly is that the indeterminacies of the ICA estimates are avoided. Furthermore, better performance than in the blind case is obtained at low to moderate Signal to Noise Ratios (SNRs) and Doppler frequency shifts, using a training overhead of 2%. However, for high SNRs or Doppler frequency shifts as well as with reduced-complexity versions convergence to a solution extracting the sources in the wrong order can occur, which results in an error floor, since no post-processing to handle these ICA indeterminacies is used. Finally, semi-blind receivers for MIMO SC-FDE systems using linear equalisation and LSFE are investigated. To the author's knowledge this is the first published application of ICA to MIMO SC-FDE systems. Since in SC-FDE systems the frequency domain signals are Gaussian, ICA is applied in time domain while a computationally efficent frequency domain equalisation is employed. The resulting receiver is shown to achieve performance close to the perfect CSI case for moderate to high SNRs, using a training overhead of only 0:05%. Furthermore, convergence to a wrong solution as in the semi-blind OFDM case has not been observed, which can be explained by the larger solution space in SC-FDE systems than in OFDM systems.

Item Type:	Thesis (Doctor of Philosophy)
Additional Information:	Date: 2007-06 (completed)
Subjects:	?? TK ??
Divisions:	Faculty of Science and Engineering > School of Electrical Engineering, Electronics and Computer Science
Depositing User:	Symplectic Admin
Date Deposited:	24 Nov 2010 10:37
Last Modified:	16 Dec 2022 04:32
DOI:	10.17638/00000429
Supervisors:	Nandi, Asoke
URI:	https://livrepository.liverpool.ac.uk/id/eprint/429