Abstract
AbstractTurbo codes can provide surprising bit error rate performance on digital communications. They have been considered for use in third-generation wireless mobile communication systems, such as direct-sequence code-division-multiple-access (DS-CDMA) systems. In order to provide better performance in DS-CDMA systems, an appropriate joint design of the turbo code decoder and the multiuser detector is highly desirable. In this thesis, an efficient iterative multiuser receiver is proposed for turbo-coded CDMA systems. The proposed structure consists of a likelihood calculator, a metric generator, and a bank of K turbo decoders, where K is the number of users and one multiuser detection iteration includes only one turbo decoding iteration. The metric generator can make use of the reliability information fed back from the turbo decoders to provide the a priori information for use in the turbo decoders. The reliability information passed from the turbo decoders to the metric generator in the conventional iterative scheme is either the extrinsic information or the a posteriori information. In this thesis, we propose a simple rule to choose the type of reliability information delivery; if the component decoders inside a turbo decoder have the same decision for the concerned symbol at an iteration, the extrinsic information generated at the turbo decoder is passed to the metric generator, or the a posteriori information is passed. Computer simulation results show that the iterative scheme with the proposed rule outperforms the conventional schemes with no increase in computational complexity. In addition, we propose a reduced-complexity metric generator for iterative coded CDMA systems. As compared to the conventional method, the proposed metric generator achieves a significant complexity reduction. For the optimal iterative multiuser receiver, the complexity per user bit per iteration is reduced from O(2KK) to O(2K/K). For the suboptimal iterative multiuser receiver proposed by Reed et al., the corresponding complexity is further reduced to be O(1). The good-performance and low-complexity features make the proposed approaches attractive for use in practical applications.