Nonlinear Self-Interference Cancellation Based on CNN-BiLSTM-Attention

For the self-interference problem in full-duplex wireless communication systems, a nonlinear self-interference cancellation method based on the CNN-BiLSTM-Attention network is proposed. The method first captures the spatial features of the input sequence through the CNN layer, then extracts the temporally-related features using a BiLSTM layer with multiple LSTM units that are bidirectionally connected. The Attention layer is used to extract the global dependencies within the time series data, allowing the network to dynamically adjust the weights of the individual time steps during learning. Finally, a fully connected layer is employed to construct the nonlinear self-interference signal, which is then removed at the receiver to achieve the purpose of self-interference suppression. Simulation results show that the algorithm exhibits excellent performance in nonlinear self-interference cancellation.