Inter-Cell Interference Suppression for Channel Estimation and Data Detection in MIMO-OFDM Systems via Score-Based Variational Inference

Abstract

In multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) systems, conventional receivers based on linear channel estimation and subband-averaged interference covariance reconstruction suffer serious performance degradation under strong interference. To address this issue, we propose a probabilistic framework for joint channel estimation and interference suppression using score-based prior learning. Specifically, we formulate a variational inference problem that exploits the bilinear structure of the interference by explicitly separating the interference channels from the interference symbols. To solve the resulting highly nonconvex bilinear inference problem, we develop a score-based variational inference algorithm that combines score-based turbo message passing for target-channel estimation with a reverse mean propagation framework for interference-channel inference, while the interference symbols are updated via linear minimum mean square error (LMMSE) estimation. We further design an interference rejection combining (IRC) detector with subcarrier-level covariance reconstruction based on the estimated interference channels, thereby avoiding the performance loss caused by conventional subband averaging. Numerical results demonstrate that the proposed algorithm is able to effectively suppress the interference in channel estimation, as well as to accurately reconstruct the interference subspace for interference rejection in data detection, and thus significantly outperforms existing baselines.