부산대학교 도서관

Gaussian approximation of interference (GAI)-based BP (GAI-BP) serves as the state-of-the-art (SOA) BP detection algorithms for large-scale MIMO systems. Nevertheless, the GAI-BP suffers from high computational expense due to the exponentiation operations for computing the a prior probability. To address this issue, in this work, we propose a threshold-updating strategy-based GAI-BP (TU-GAI-BP) detection algorithm. The TU-GAI-BP first employs the threshold to select the target a prior probability which tends to be converged after several iterations, then freezes the probability in the following iterations. Simulation results prove that compared with the SOA GAI-BP, the proposed TU-GAI-BP achieves about 71% computational reduction and still remains comparable error performance in 32 × 16, 16-QAM MIMO scenario. Additionally, we design the hardware architecture for the proposed TU-GAI-BP detector.