Model-driven Deep Learning Based Channel Estimation for HPO-MIMO

Halved-Phase Only MIMO (HPO-MIMO) can alleviate the problems of high complexity and high power consumption of conventional MIMO. However, due to amplitude loss, HPO-MIMO cannot adopt the conventional receiver scheme. We propose an HPO-MIMO Channel Estimation (CE) scheme based on model-driven deep learning. Specifically, we use the Generalized Approximate Message Passing (GAMP) algorithm to generate preliminary channel estimation results. Then the Deep Neural Network (DNN) is used to refine the estimation values to obtain more accurate channel estimation results. Simulation results indicate that the presented DNN-based scheme retains the preponderance of the traditional GAMP algorithm but performs better than the latter. When the Mean Square Error (MSE) is −20dB, there is a Signal-Noise Ratio (SNR) gain of about 13dB between them. In addition, the DNN channel estimation method based on GAMP is proved suitable for different antenna sizes, antenna distances, and user densities.