A novel approach to enable decorrelating multiuser detection without matrix inversion operations

摘要

This paper proposes a non-matrix inversion based algorithm to implement decorrelating detection (DD), namely quasi-decorrelating detector (QDD), which uses truncated matrix series expansion to overcome the problems associated with the matrix inversion in DD, such as noise enhancement, computational complexity and matrix singularity, etc. Two alternative QDD implementation schemes are presented in this paper; one is to use multi-stage feedforward filters and the other is to use an wth order single matrix filler (neither of which involves matrix inversion). In addition to significantly reduced computational complexity if compared with DD, the QDD algorithm offers a unique flexibility to trade among MAI suppression, near-far resistance and noise enhancement depending on varying system set-ups. The obtained results show that the QDD outperforms DD in either AWGN or multipath channel if a proper number of feed-forward stages can be used. We will also study the impact of correlation statistics of spreading codes on the QDD's performance with the help of a performance-determining factor derived in the paper, which offers a code-selection guideline for the optimal performance of QDD algorithm.