Windowed Information Geometry Approach for Low-Complexity Channel Estimation in Massive MIMO-OFDM Systems

Accurate channel estimation is essential in massive multiple-input multiple-output orthogonal frequency division multiplexing (MIMO-OFDM) systems. As the number of base station antennas further increases, high dimensional channel estimation becomes challenging. To tackle this problem, we first introduce window functions into received signal model to obtain the windowed received signal model. The channel estimation is formulated as obtaining the $\mit {a} \;\mit {posteriori}$ distribution’s marginals of the beam domain channel within the information geometry (IG) framework. The calculation of the marginals is turned into an iterative projection process. We derive the windowed information geometry (WIG) approach by calculating an approximate solution of the projection in the presence of window functions. To reduce the computational complexity, we further propose a low-complexity calculation method for WIG. Specifically, by choosing suitable window functions, we can transform the original matrix multiplications into sparse matrix multiplications, which dramatically reduces the complexity. Furthermore, by adjusting the values of thresholds used in the low-complexity calculation method, we can make a superior trade-off between complexity and performance of WIG. Finally, complexity and performance of WIG are evaluated through simulations. Simulation results show that WIG can achieve similar performance to the simplified IG (SIG) while the complexity is much lower.