Zero-calibrated Brain-computer Interface Based on Fourier Phase Information

Zero-calibration brain-computer interfaces (BCIs) represent a highly promising and significant research area, addressing the calibration issues caused by EEG data drift. Domain generalization is employed to achieve zero-calibration BCIs and invariant feature extraction is a critical technology for domain generalization. Phase information, as an invariant feature extraction method, has been applied to achieve zero-calibration interfaces. However, phased-based methods have failed to consider the impact of frequency on phase information. This paper explores the influence of frequency on phase-invariant feature extraction and proposes a zero-calibration BCI based on Fourier phase information. We validated the effectiveness of the proposed algorithm on public dataset. Experimental results indicate that extracting phase-invariant information within an appropriate frequency range can enhance the system's generalization performance. Moreover, it is important to avoid performing the transfer within the α band when using spectral transfer to increase the similarity of EEG data.