Widely Memory-Canonical Piecewise Linear-Based Self-Interference Canceller for Full-Duplex Radios

In this paper, a widely memory-canonical piecewise linear (WMCPWL)-based model for the digital self-interference cancellation (DSIC) in In-band full-duplex communication systems is proposed. Unlike previous polynomial-based and neural networks (NN)-based models, the proposed one decomposes the total self-interference (SI) signals into widely linear and WMCPWL components, the latter of which is introduced to mimic the nonlinearity and memory effects between transmitting and receiving paths. Moreover, the image rejection ratio (IRR) which generally quantifies the IQ imbalance can be extracted by estimating SI channel coefficients corresponding to the widely linear part. Simulation results demonstrate that the proposed model decreases both computational complexities and model parameters while providing a similar cancellation performance compared with other state-of-the-art models.