Singular value decomposition techniques for multiuser detection receivers

Designing multiuser detection (MUD) receivers with low complexity and near single-user performance has generated a great deal of interest in code division multiple access (CDMA) systems. This thesis proposes novel approaches, primarily based on the Singular Value Decomposition (SVD), for designing multiuser detection receivers for CDMA systems with and without the use of multiple antennas. A new transmitter precoding scheme is proposed for the uplink of a single-input single-output (SISO) direct spread (DS) CDMA Rayleigh fading channel where non-orthogonal spreading codes are used. This technique is based on the SVD and can be analytically shown to achieve the single user bound. The application of the SVD to a multiple-input multiple-output (MIMO) DS-CDMA Rayleigh fading channel with forward error correction (FEC) codes is subsequently investigated. An iterative (turbo) multiuser detection algorithm is proposed for this MIMO multiple access system. The proposed MUD receiver is of low complexity when compared to the optimal MUD receiver and has near single-user performance in the presence of long-term power control. The salient feature of our iterative MIMO MUD receiver is that it combines SVD filtering, FEC coding and spreading codes. It is demonstrated that the proposed iterative MIMO MUD receiver can also be derived by applying the sum-product algorithm to the corresponding factor graph. The extrinsic information transfer (EXIT) chart analysis is shown to be accurate in studying the convergence behaviour and estimating the coded bit error rate (BER) for the iterative MIMO MUD receiver. A gear-shift MUD receiver is then introduced where the receiver is allowed to iteratively select a particular algorithm from a group of MUD algorithms in order to achieve a target performance level in the least amount of time.