Toward Reliable Asynchronous Brain–Computer Interfaces via Periodic–Aperiodic Feature Integration

Asynchronous brain–computer interfaces (BCIs) enable users to initiate commands at will, offering a more natural and flexible interaction paradigm compared with conventional synchronous systems. However, accurately distinguishing between intentional control (IC) and non-control (NC) states remains a critical challenge, particularly in noninvasive steady-state visual evoked response (SSVER)-based BCIs. Existing approaches either rely on stimulus-locked periodic features or incorporate auxiliary physiological signals, but their performance is often limited by response variability, nonstationarity, and reduced practicality. In this study, we propose a unified framework that jointly leverages periodic and aperiodic features of neural signals (UFPAF) to enhance state recognition in asynchronous BCIs. Specifically, the proposed method integrates SSVER-based periodic features with aperiodic characteristics derived from the 1/ f -like structure of the power spectrum, enabling a more comprehensive representation of brain activity. We evaluate the proposed framework through extensive offline experiments on both electroencephalography (EEG) and optically pumped magnetometer-based magnetoencephalography (OPM-MEG) data, comparing it with four representative methods, and through online validation in real-time brain-controlled spelling and virtual vehicle control tasks. The results demonstrate that incorporating aperiodic information significantly improves detection accuracy and robustness, particularly under conditions of weak responses and signal variability. These findings underscore the importance of jointly modeling periodic and aperiodic neural dynamics and suggest that the proposed approach provides a promising solution for achieving reliable and naturalistic asynchronous BCI systems.