Low Complexity Self-Interference Cancellation Based on Two-Stage RLS Filtering for Full-Duplex Radios

Full-duplex wireless communication can double the spectral efficiency by transmitting and receiving information simultaneously, but it suffers from strong self-interference (SI). Baseband SI cancellation (SIC) can estimate and subtract the SI from the receiving signal, while perfect SIC is unachievable for the existence of the intended signal. To enhance SIC capability, we propose a two-stage recursive least squares (RLS) filter architecture with low computation complexity and strong environmental robustness. In the first stage, the filter acts as a traditional RLS filter to cancel SI. In the second stage, we introduce the reconstructed intended signal as an additional input of the RLS filter, and the filter jointly estimates the SI channel and intended channel to attenuate the SI signal approaching the noise floor. Besides, to improve the robustness of the proposed method, we introduce an adjusted diagonal loading into the two-stage RLS to reduce the convergence time. The computation complexity and storage complexity are analyzed as well. Simulation results show that the proposed method has up to 10 dB SIC improvement over the traditional RLS, and achieves the same performance as the two-stage least squares (LS) method after two iterations. On the other hand, our proposed two-stage RLS filter reduces the computation complexity by about 81.6% compared to the two-stage LS method when the filter order is 16.