Combining Spectral Activation with Brain Network Interactions for EEG-based Attention Recognition

Attention is a higher cognitive function of the human brain and is crucial in daily life. Recently, attention-based Brain-Computer Interface (BCI) has become a new hotspot in the field of BCI. However, current studies often employ single-task experimental paradigms that fail to capture the dynamic transitions between cognitive states, limiting their ability to effectively engage participants' attention. Moreover, most studies rely on both frequency-domain and time-domain methods for analyzing EEG signals. Frequency-domain approaches overlook the temporal dynamics of signals, while time-domain methods lack spectral resolution. Such one-dimensional analyses are insufficient to reveal the time-frequency interactions of EEG signals. In this study, we designed an alternating attention and relaxation experimental paradigm and proposed a method that combines spectral activation with brain network interactions for attention recognition. The experiment involved 69 participants, achieving the highest accuracy of 91.22%. Furthermore, we found significant differences between attention and relaxation states across frequency bands. In the attention state, the prefrontal and parietal regions showed enhanced connectivity. This study provides a new paradigm and theoretical support for understanding attention regulation mechanisms and advancing brain-computer interface applications.