Code-aided multiuser detection in multiple access wireless communication systems

As a multiuser communication technology, direct sequence (DS) code division multiple access (CDMA) becomes widely accepted and even promising to play a key role in future wireless systems. However, a DS/CDMA system is interference limited. Multiuser detection techniques are required to combat multiple access interference effectively in the presence of multipath fading.; This dissertation proposes different decentralized blind linear multiuser detection techniques with the aid of only the desired user's spreading codes. For short-code CDMA systems, higher order statistics (HOS) based approaches are first investigated. With a set of code-aided constraints on the receiver, the constant modulus (CM) algorithm is shown to provide conditional convergence. Further optimization of the constraint yields an equivalent maximum ratio combining scheme without knowing desired user's channel parameters. A kurtosis maximization/minimization technique is then proposed that ensures global convergence and yields a zero-forcing solution. However, it shows slow convergence in practical conditions. To solve these problems, a new constraint in a minimum mean-square-error (MMSE) form is imposed on the CM-based receiver. This method enjoys global convergence and provides an asymptotic MMSE solution. To achieve less complexity and even faster convergence, a second-order statistics (SOS) based approach termed as Power of R (POR) is proposed. It demonstrates superior performance to other popular SOS approaches and can achieve asymptotic performance of the MMSE receiver. It is also comparable to the HOS based ones. Therefore, by applying those proposed methods, performance gaps between existing decentralized blind receivers and the ideal MMSE receiver are bridged. Meanwhile, comprehensive performance analyses are provided in terms of convergence property, channel estimation mean-square-error, and receivers' bit error rates. The dissertation also focuses on performance study for long-code CDMA systems, where effects of finite sample size on linear receivers are analyzed when some existing channel estimation methods are employed.