Iterative Multi-User Detection Using Reduced-Complexity Equalization

Inhalt:
In this paper, we consider a reduced-complexity equalization method for turbo Multi-User Detection (MUD) in a frequency-selective Code Division Multiple Access (CDMA) uplink scenario. The non-linear trellis-based detector which is normally used in turbo receivers, is replaced by the Matrix Wiener Filter (MWF), the optimal linear filter based on the Mean Square Error (MSE) criterion. To further reduce the complexity, the Block Conjugate Gradient (BCG) algorithm is used to approximate the MWF in a low-dimensional Krylov subspace. Additionally, second order statistics of non-stationary random processes are approximated by their time-invariant averages. Simulation results show that embedding the suboptimal MWF in a turbo receiver does not lead to significant performance loss, while reducing the computational complexity enormously.