双选信道下基于图模型的超奈奎斯特信号频域均衡方法

Modern mobile communication applications raise the requirement of high quality support for high mobility users. In this paper, we present a Bayesian graphical model based frequency domain equalization method for faster-than-Nyquist (FTN) signaling in doubly selective channels.The conventional frequency domain minimum mean squared error (FD-MMSE) equalizer suffers high complexity due to the interferences induced by adjacent frequency symbols. To tackle this problem,a low complexity iterative message passing method namely,belief propagation is employed on the Bayesian graphical model to detect the FTN symbols. Compared to the low complexity variational inference method, the proposed algorithm considers the conditional dependencies between symbols and therefore can improve the performance. Simulation results show that the proposed equalization method has similar performance of the MMSE equalizer and outperforms the variational inference method.