SVD Based Sphere Detector for MIMO Channels

It is well known that Maximum Likelihood(ML) criteria results in the best BER performance for multiple-input multiple-output (MIMO) systems, but exhaustive search is unacceptable in practical systems mapped by large QAM constellations. Sphere detector is promising by setting constraints on the points to lie in a hypersphere. Generally, the constraints are set through QR decomposition. In this paper, a new scheme of sphere detector is proposed and described in detail. Instead of QR decompositions, singular value decomposition (SVD) is used in the new scheme to get the geometric parameters of the hypersphere, and then construct the search tree needed for detecting iteration. The proposed scheme can achieve the same BER performance with less computation compared to the traditional algorithms.