Low-Complexity Linear and Non-Linear Digital Self-Interference Cancellation

To fully utilize the in-band full-duplex (IBFD) communication in practice, digital self-interference (SI) cancellation is indispensable. In the literature, several methods have been proposed to suppress the SI signal in digital domain, however, mostly linear and non-linear SI techniques are studied separately. In this work, we review existing digital SI cancellation methods and present a low-complexity solution for both linear and non-linear digital SI cancellation by exploiting frequency-domain processing as well as a neural network-based approach. We developed an orthogonal frequency division multiplexing (OFDM)-based IBFD transceiver in GNU Radio and use it in combination with USRP software-defined radios (SDRs) to experimentally demonstrate the performance in an unified framework. Our results show that the proposed SI cancellation is computationally 4 times more efficient and achieves a similar SI cancellation performance as the state of the art.