Information Geometry Approach Based on Deep Learning for Massive MIMO-OFDM Channel Estimation

Massive multiple input multiple output communication has outstanding advantages in spectrum and energy efficiency. Orthogonal frequency division multiplexing modulation is the fundamental modulation technology for broadband wireless communication. In recent years, the combination of the two has been widely used in 5G systems. Channel estimation plays an important role in massive MIMO-OFDM systems. However, with the significant increase of channel matrix dimension in massive MIMO-OFDM systems, it is challenging to perform accurate channel estimation with low computation complexity. The existing information geometry approach for channel estimation in massive MIMO-OFDM systems can achieve higher channel estimation accuracy with iterations. However, this algorithm based on heuristic strategy has unstable convergence performance. An information geometry approach based on deep learning for channel estimation is proposed in this paper to overcome the aforementioned shortcoming. This method can solve the optimization problem of damping factors in existing information geometry approach for channel estimation. It has better convergence performance with less number of algorithm iterations, and the complexity of channel estimation is reduced significantly.