In this work, we demonstrate that an $n \times n$ massive multiple-input multiple-output (MIMO) channel can be polarized using common matrix decomposition techniques: singular value decomposition (SVD) and QR decomposition. With full channel state information (CSI), we show that channel capacity is always attained by freezing certain number of worst subchannels, provided a total power constraint and sufficiently large $n$. We further prove that the capacity obtained through channel polarization is always greater than that achieved through channel equalization. Finally, we propose a low-complexity precoding scheme based on channel polarization, which outperforms the lattice-reduction-aided precoding scheme, in terms of capacity, decoding error rate, encoding complexity, and CSIT cost.
翻译:暂无翻译