A low-complexity algorithm based on variational Bayesian inference for MIMO channel estimation

With an increase in the number of transmitters in multiple-input multiple-output (MIMO) communication systems, there is a cubic rise in the computational complexity of the traditional sparse Bayesian learning (SBL) channel estimation algorithm. While various algorithms are effective for single-input single-output (SISO) systems, they are not suitable for the MIMO scenario. This paper introduces a MIMO channel estimation algorithm based on variational Bayesian inference (VBI) by assuming the independence of the variational distribution among different channels. The high-dimensional channel vectors estimated in the conventional MIMO-SBL algorithm are decomposed into multiple parallel lowdimensional channel vectors with different sparsity using VBI. Consequently, the complexity exhibits a linear relationship with the number of transmitters, as demonstrated through numerical analysis. Simulations confirm the improved estimation accuracy of the MIMO-VBI algorithm. Experimental results reveal that MIMO systems can achieve lower bit error rates using the MIMO-VBI algorithm, with reduced runtime for channel lengths exceeding 100 symbols.