Application of Deep Neural Network for NOMA Systems in 5G Communications

Currently, multi-channel access techniques may not be able to effectively handle massive data situations. For 5G communication systems, non-orthogonal multiple access (NOMA) can improve channel efficiency to support more users. However, in the transmitter, the complete channel state information (CSI) is required, while in the receiver, the computational complexity is high, which is the main disadvantage of NOMA system. The use of deep neural network (DNN) as a deep learning (DL) technique is an excellent solution to deal with various problems. In this work, DNN predicts parameter values to perform channel estimation in NOMA systems. From the results, the performance in terms of symbol error rate (SER) can be improved higher than standard estimation approach.