Signal Detection in NOMA Systems using DNN with Bidirectional LSTM

Abstract

Deep and machine learning-based algorithms are two new methodologies to solve time series prediction challenges. Traditional regression-based modeling has been found to provide less accurate findings than these techniques. A deep learning-assisted method for detecting signals in non-orthogonal multiple access systems with orthogonal frequency division (OFDM-NOMA) is described. The deep neural network (DNN) with a bi-directional long short-term memory (Bi-LSTM) is used to detect signals using different deep learning-based optimizers such as Sgdm, RMSprop, and Adam. The combination that detects most accurately is determined by comparing neural networks and optimizers. The simulations show that the deep learning technique can outperform the conventional successive interference cancellation (SIC) method and demonstrate that the Bi-LSTM-based deep learning algorithm may effectively detect signals in NOMA system scenarios in the Long-Short- Term Memory (LSTM) model. As a result, deep learning is a reliable and essential method for detecting NOMA signals.