Data-Enhanced Bayesian MIMO-OFDM Channel Estimation Strategy with Universal Noise Model
Jia‐Cheng Jiang, Hui‐Ming Wang · 2020
Model-based methods are dominant in current systems for their optimal designs under given models, but may suffer from inaccurate modeling assumptions. Recently, data-based deep learning methods have achieved remarkable performances by training a large amount of data but encounter some challenges such as, lack of available training data and explainability. In this paper, we propose a novel hybrid idea to integrate the strengths of both data and model-driven methods, named model based method enhanced by data, which is training affordable, theoretically interpretable and model flexible. To show the idea more concretely, we consider a multiple-input multiple-output (MIMO) orthogonal frequency division multiplexing (OFDM) channel state information (CSI) acquisition approach. Specifically, we utilize a universal mixture of Gaussian (MoG) model to deal with the nongaussianity of the noise and interference in complex communication environments, which can adaptively adjust involved parameters to fit the true distribution by observed data. We propose a variational Bayesian framework to derive the specific form of minimum mean square error (MMSE) estimator. Simulations are performed to verify the efficiency of our proposed method and the accuracy of our analysis.