Hybrid-Layers Neural Network Architectures for Modeling the Self-Interference in Full-Duplex Systems

Full-duplex (FD) systems have been introduced to provide high data rates for beyond fifth-generation wireless networks through simultaneous transmission of information over the same frequency resources. However, operation FD is practically limited by self-interference (SI), and efficient SI cancelers are sought make realizable. Typically, polynomial-based employed mitigate SI; nevertheless, they suffer from complexity. This article proposes two novel hybrid-layers neural network (NN) architectures cancel with low The first architecture referred as hybrid-convolutional recurrent NN (HCRNN), whereas second termed dense (HCRDNN).In HCRNN, a convolutional layer extract input features using reduced scale. Moreover, then applied assist in learning temporal behavior signal localized feature map layer. In HCRDNN, an additional exploited add another degree freedom adapting settings order achieve best compromise between cancellation performance computational complexity analysis proposed provided, optimum their training selected. simulation results demonstrate that HCRNN HCRDNN-based attain polynomial state-of-the-art NN-based astounding reduction. Furthermore, show design flexibility hardware implementation, depending on system demands.