Digital Self-Interference Cancellation Using Convolutional Neural Networks in In-Band Full-Duplex Systems

Abstract

    In-band full-duplex (IBFD) communication systems represent a major breakthrough in wireless communication, allowing transmission and reception of signals on the same channel simultaneously. However, a critical challenge in IBFD systems is mitigating Self-Interference (SI), which results from the leakage of the transmitted signal into the receiver, potentially overwhelming the received signal and degrading overall system performance. This paper investigates the application of Convolutional Neural Networks (CNNs) for digital Self-Interference Cancellation (SIC) in IBFD systems. We propose a CNN architecture consisting of three convolutional layers, designed to learn and suppress the nonlinear characteristics of self-interference effectively. Simulation results demonstrate that the proposed CNN-based SIC method achieves 52.492 dB interference suppression at an SNR of 30 dB, significantly improving the bit error rate performance compared to conventional methods.