Multiuser detection techniques for improper signals

Abstract

In this dissertation, we propose new techniques of three practical and important topics in DS-CDMA systems. The first one is blind multiuser detection for improper received signal of DS-CDMA system, in which we assume that the asynchronous DS-CDMA system and multipath fading channels. The second is reduced-complexity maximum-likelihood (ML) multiuser detection. The last is optimal spreading sequence design utilizing minimum mean-squared-error (MMSE) multiuser detection. First, we consider a new linear blind multiuser detection based on widely linear (WL) signal processing. The received signal and its complex conjugate are separately filtered and the results are linearly combined. The WL maximum/minimum (max/min) mean-output-energy (MOE) receiver is derived by applying the modified cost function. It is shown that a performance gain can be attained for the improper DS-CDMA signal, due to the additional information contained in pseudo-covariance matrix of observations. The adaptive implementation with acceptable complexity is also developed. Computer simulation results show that a significant performance gain is obtained over the other classical methods. Second, the reduced-complexity maximum-likelihood (ML) multiuser detection is proposed for improper DS-CDMA signals in multi-cell environment. It is shown that the inter-cell interference and the background noise can be modelled as an improper random vector with the nonzero pseudo-covariance matrix. Based on this model, a computationally efficient multiuser detection is developed and its bit error probability is analyzed. The computer simulations demonstrate the superiority of the proposed scheme to the various conventional detection schemes and validate the bit error probability analysis result. Finally, the design of optimum spreading sequences is considered over multipath fading channel employing the widely linear (WL) minimum mean square error (MMSE) receiver. The optimum spreading sequences...