Novel low-complexity MIMO detection based on constellation shift binary classification

Abstract

This thesis develops a novel technique for low-complexity MIMO detection known as Constellation Shift Binary Classification (CSBC). The proposed method utilizes the constellation structure to deduce a constellation shift that is used to reduce the problem of detecting the symbol into binary classification. The proposed method is proven to outperform other common techniques such as ZF and LMMSE. The proposed CSBC is a novel MIMO detection technique since it utilizes the constellation structure together with a binary classifier that is based on neural networks.