A low complexity channel decomposition and feedback strategy for MIMO precoder design

In classical MIMO systems, singular value decomposition (SVD) is adopted as a common way to decompose the MIMO channel into parallel subchannels. However, in addition to its high complexity, it is also sensitive to the ill-conditioning of the channel matrix. In this paper, we propose a channel decomposition strategy called LDL^H decomposition for low complexity MIMO precoder design. We show that the computation complexity of LDL^H achieves one degree of magnitude lower than that of SVD, as well as some SVD-based decomposition scheme such as geometric mean decomposition (GMD) and uniform channel decomposition (UCD). In addition, we also show that LDL^H-based precoder requires less quantization effort and feedback bandwidth.