查看文章

In non-orthogonal multiple access (NOMA) system, accurate detection and hence the successive interference cancellation (SIC) becomes crucial for reliable decoding. Moreover, channel state information (CSI) acquisition in NOMA is challenging due to potential interference. A DL-based NOMA receiver can easily decode messages for multiple users with accurate SCI and better CSI. The one-dimensional conventional neural network (1D-CNN) model is used to extract the input signal characteristics from OFDM, and the time series gated recurrent unit (GRU) network is then used to infer the signal at the receiver terminal. The InDNN is initially trained offline using simulation data that relies on channel statistics. Subsequently, it is utilized in the online deployment stage to directly recover the transmitted symbols. The Adam optimizer is employed to optimize the loss during the model learning process. Preliminary results …