An improved multiuser detector with gradient adaptive step size for CDMA systems

In code-division multiple-access (CDMA) systems, parallel interference cancellation (PIC) schemes are attractive to the practical implementation of multiuser detection due to the lower complexity. However, a wrong interference subtraction as a result of the lack of exact knowledge of the interference signal induces a quadruple increasing interference power compared to those without any cancellation. The previous work of an adaptive multistage receiver proposed by Xue et al. (1999) using a set of weights through a normalized least mean square (NLMS) algorithm simultaneously removes from the desired user a proper part of the interference based on its reliability information in the previous stage. However, the gradient noise arises due to the improper normalized step size to the time-varying environmental parameters and hence results in worse performance by incorrect cancellation. This paper presents an improved adaptive parallel interference cancellation approach with a gradient adaptive step size, which is optimized in the sense of minimizing the Euclidean distance between the received signal and its estimate. The computational complexity of the approach is linear with the number of active users. Simulation results show that the proposed scheme significantly outperforms the original scheme.