Deep Learning-Aided Channel Estimation for Large Scale Fading Decoding in Massive MIMO Systems

A Massive Multiple-Input-Multiple-Output (MIMO) system is a wireless communication technique in which base stations (BS) are equipped with a large number of antennas. Massive MIMO systems have significant inter-cell interference caused by the phenomenon of pilot contamination. A dual-layer decoding technique employing Deep Neural Networks (DNNs) for channel estimates can effectively reduce both coherent and non-coherent interference in multi-cell massive MIMO networks. Channel estimation is a crucial component of enormous cellular MIMO networks that directly affect the performance of a wireless communication system. This study presents two DNNs that are designed to accurately estimate channel state information and effectively decrease interference. Initially, we present a Multi-Layer Perceptron (MLP) that exclusively comprises completely connected layers. Additionally, we present a Convolutional Neural Network (CNN) that comprises one-dimensional convolution layers and fully connected layers. During the training phase, the neural networks acquire knowledge of intricate patterns and utilize the spatial and temporal correlation seen in massive MIMO channels. This allows for the development of a strong system that can successfully handle pilot contamination and reduce interference. The simulated results illustrate the effectiveness of the DNN methods compared to the Minimum Mean Square Error (MMSE) channel estimate in terms of Sum Spectral Efficiency (SE) in several circumstances, such as multi-user, multi-antenna, and variable channel conditions. The suggested system model attains a notable increase in sum spectral efficiency, hence enhancing data transmission rates and overall system performance.