A nearest neighbor search algorithm for LR-LD on high SNR

The faster decoding algorithm is still required in multiple-input multiple-output (MIMO) wireless communication technology. Lattice reduction (LR) technique can improve the performance of linear MIMO detection methods. LLL (Lenstra-Lenstra-Lovasz) algorithm is one of the LR techniques and it has being extensively used for better basis of channel matrix by removing noise. LR-aided detection algorithms involve two problems. The first problem is that the reduced basis is still non-orthogonal and most algorithms are devoted for finding a near orthogonal matrix. The second problem is that the feasible set of the detected symbol cannot be found without huge operation. In this paper, we propose a new algorithm which considers unknown transformed symbol to search the neighboring points on transformed basis of the channel matrix. This paper mainly focuses on the second problem to resolve the poor accuracy particularly in high signal-to-noise ratio (SNR). Our proposed algorithm can achieve better BER performance with low complexity and it is proven by simulations.