Digital Self-Interference Cancellation Using Kernel Adaptive Filtering in Hilbert Spaces

This paper introduces a novel adaptive kernelbased algorithm for digital self-interference cancellation (SIC) in in-band full-duplex (FD) wireless communication systems. FD transceivers, which enable simultaneous transmission and reception on the same frequency band, present an attractive solution to the increasing demand for scarce wireless spectrum. However, operating in the full-duplex mode results in a strong self-interference (SI) signal, which in turn prevents the potential of doubling the spectrum efficiency to be realized. Therefore, effective suppression of SI is crucial for the feasibility of FD wireless communications. This study investigates a new approach for rapidly identifying and tracking the SI channel, hence well-suited for dynamic wireless communication environments. We propose an application-tailored adaptive projected subgradient method (APSM) in the Cartesian product of multiple reproducing kernel Hilbert spaces, enabling us to utilize various kernels. We verify the efficacy of the proposed APSM-based approach by employing a custom hardware testbed. In a nonlinear scenario with overthe-air measurement data, the proposed method achieves 50 dB of digital SIC outperforming a linear technique by 15 dB.