LCST: The Impact of Compressive Sensing Technique on Massive MIMO Channel Estimation for 5G Wireless Sensor Networks

One of the main constraints on the efficiency of large-scale MIMO networks involves operator inefficiency. This is due to the fact that these systems' functionality relies on the inaccurate presentation of channel state information (CSI). This paper introduces a new channel estimation method that minimizes pilot overhead problems by applying the concepts of reducing mapping. Combining sparsity adaptive comparing search with compression sample fitting forms the basis of the suggested method. When it comes to spatial correlations, the signal sources in MIMO networks are pretty scattered. The channel estimation issue regarding MIMO systems can be solved using compressive sampling techniques, made possible by this distribution pattern. To enhance the channel estimation obtained by the Learning based Compressive Sensing Technique (LCST), a method based on the least square approach; this paper presents a new architecture for channel estimation that makes use of deep learning. To test how well the new LCST model works, it is cross-validated with the tried-and-true RSCST, or Radio Signals-based Compressive Sensing Technique. We accomplish this by simulating 5G and beyond networks with a multiple-input multiple-output (MIMO) system and a multiple-path channel profile, where the mobility is represented by the Doppler effects. The system model can accommodate any number of transceiver antennas, and the machine learning module is designed to work with any type of neural network. Compared to other well-known, conventional channel estimation methods, the suggested deep learning-based framework clearly outperforms them in terms of numerical results. Furthermore, out of all the artificial neural network designs that were considered, bidirectional long shortterm memory provides the lowest bit error ratio and the best channel estimation quality.