Non-Linear Self-Interference Cancelation for Full-Duplex Transceivers Based on Hammerstein-Wiener Model

Abstract

The main challenge in a full-duplex transceiver design is created by the self-interference caused by the coupling of the transmitted signal to the transceiver’s own receiver. The effect of the non-linear operation of both the power amplifier at the transmitter and the low noise amplifier at the receiver are considered in the self-interference cancelation. The performance of three self-interference cancelers are studied: linear cancelation, auto-regressive moving-average (ARMA) based cancelation and a neural network (NN) based canceler. The NN based cancelation outperforms both the linear and ARMA based canceler but requires considerably more operations than the other two.