On normalized belief of Gaussian BP in correlated large MIMO channels

This paper proposes a normalized belief with the aid of node selection (NS) method in Gaussian belief propagation (GaBP) detection especially for a large multiuser multiple-input multiple-output (MU-MIMO) configurations in bands below 6 GHz. GaBP is known as an efficient detector regarding low computational complexity as well as high detection capability only if the BS has hundreds of receive antennas and each antenna experiences uncorrelated fading. When the wavelength is long such as below 6 GHz, we face two inconvenient problems for the GaBP detection. The first one is a limitation that a base station (BS) has tens of antennas at most. Another one is impairments induced by spatial fading correlation in the BS side. As a result, GaBP is subject to ill convergence behavior of iterative detection due to lack of channel hardening effects obtained by a massive number of uncorrelated receive antennas. To improve the convergence property, we propose the novel normalized belief that is adaptively scaled according to the instantaneous channel state. Finally, we demonstrate the validity of the normalized belief with the aid of NS methods, in terms of suppression of bit error rate (BER) floor.