Nonlinear self-interference cancellation based on complex-valued neural networks for co-platform

In the contest for electromagnetic spectrum dominance, multifunctional integrated platforms face the challenge of co-platform self-interference (SI) caused by simultaneous signal transmission and reception, which significantly degrades the performance of multifunctional systems. Traditional SI cancellation (SIC) methods employ polynomial models. However, due to their large number of parameters and high complexity, these methods are difficult to deploy widely in practical scenarios. To address this issue, a novel method based on a complex-valued convolutional neural network with Mish activation function (M-CVCNN) is proposed. M-CVCNN can simultaneously exploit information in both the amplitude and phase of signals, ensuring effective cancellation performance while significantly reducing the number of model parameters. Experimental results demonstrate that the M-CVCNN canceller successfully reduces the power of nonlinear SI signals by 7.16 dB with only 178 parameters.