bols detection with LRAD and correction are explaAbstract ined, simulation results will be provided in Section 5. Some concluding remarks are offered in Section 6. MLD (Maximum Likelihood Detection) has shown optimum performance in MIMO systems. However, its 2. System model complexity generally makes it impractical. To solve this problem, various techniques were proposed. We consider the often used Rayleigh fading MIMO LRAD (Lattice-Reduction-Aided Detection) was also systems model proposed to improve the performance of MIMO systems[1]. In this paper, we suggest error symbol dete-YC = Hcxc + nc (1) ction and correction in symbols estimated by LRAD. In other words, the proposed system detects the first Where usually the vectors xc, nc, yc and the matrix symbols by LRAD and searches error terms in the first Hc are obtained from the equivalent baseband model, detection symbols. And correct the error symbols by and hence are complex-valued. Hc is the channel optimal or sub-optimal systems. As a result, its matrix, xc is the vector of transmitted symbols, each performance approaches more nearly to that of MLD chosen from some X and fnc is the additive white which shows optimum performance, but it has much Gaussian noise (AWGN) vector. We assume all less complexity than MLD[1]. vectors to be column vectors of dimension K, and Hc 1. Introduction is a Kx K matrix of complex transfer coefficients between transmitter and receiver. We can equivalently write MIMO (Multiple-Input Multiple-Output) systems have recently become very popular since it can improve link reliability without sacrificing bandwidth 9YC 9i-H, -KMc 91c, 9 n ( efficiency. For the system designer, the goal is toLyc LHC ]iR-3xc i n L2) achieve an attractive performance-complexity tradeoff. At one end of the spectrum, MLD (Maximum Like-(where the 9i and 3 prefix denote the real and imalihood Detection) is optimal, but its complexity generally makes it impractical. A variety of other detectors, ginary parts), which gives an equivalent 2K-dimenboth linear and nonlinear, require substantially less sional real model of the form complexity, but sacrifice performance significantly. Recently, Yao and Wornell [2] introduced the idea of y = Hx + n (3) LRAD (Lattice-Reduction-Aided Detection). The striking property of this scheme is that it results in error