Research on improved receiver of NOMA-OFDM signal based on deep learning

In this paper, the receiver of multi-user uplink non orthogonal multiple access (NOMA) combined with orthogonal frequency division multiplexing (OFDM) signal is studied. The designed receiver is based on deep neural network(DNN) in deep learning, which consists of a layer of DNN and soft decision. It solves the problems of channel estimation error, time delay and restriction of decoding between users in traditional detection methods. It can recover the symbols of all users at one time and perform channel estimation and signal detection jointly. The simulation results show that the receiver has strong robustness to the user's power allocation. It is not only suitable for linear channel, but also for nonlinear channel. When increasing the number of users, the receiver can also detect well.