Synchronization and blind signal processing in CDMA systems

Abstract

In this thesis we consider signal processing in the Direct-Sequence (DS) Code Division Multiple Access (CDMA) systems. We pay special attention to the downlink (e.g. mobile phone) environment, where the knowledge of the system parameters as well as the signal processing capacity is limited, and where the near-far problem caused by dissimilar signal energies is difficult to mitigate. Given this setting, we propose algorithms for both synchronization, and symbol estimation. Timing estimation of the propagation delay, also called synchronization, is a prerequisite task for conventional receivers. Once the delays are estimated, one can estimate the channel, possibly also the number of users in the system, and finally the symbols itself. We propose methods based on differential correlation, which can efficiently utilize the time correlation of the varying channel, unlike conventional approaches. Desired features like low computational complexity, and capability of suppressing strong interference are achieved. Competitiveness among existing methods are verified by numerical simulations. For symbol estimation we consider the use of statistical technique called Independent Component Analysis (ICA). Blind multiuser detection is performed by separating the desired source signal blindly from the received signal. Term blind relates to the fact that the separation is performed based on the observations only. We propose ICA-based blind multiuser detectors, in which ICA is considered as an additional element attached to either conventional or MMSE-based detectors. The main benefit of using ICA in CDMA is that it offers a near-far resistant receiver which is not explicitly subject to code timing nor channel estimation errors, which both may limit the performance of conventional and other near-far resistant receivers. Also, the statistical independence of the source signals can be utilized.