Electroencephalogram-based multimodal attention level classification using deep learning techniques

This study aims to develop a novel attention level prediction method using a multimodal brain-computer interface system that integrates electroencephalogram (EEG), electrocardiogram (ECG), and electrooculogram (EOG) signals to enhance prediction accuracy and robustness. We propose the Multi-Feature Enhanced Attention Network (MEAN), which leverages the complementary strengths of these signals: EEG provides insights into brain electrical activity, ECG captures heart rate variability to reflect emotional and cognitive states, and EOG records eye movements for contextual attention level information. The model is designed to address the limitations of single-modality signals, such as noise susceptibility and limited information range. Experimental results demonstrate that MEAN achieves an average accuracy of 0.9524, outperforming traditional models. The model exhibits superior adaptability, particularly in handling EEG and multimodal data, and shows enhanced predictive performance compared to existing approaches. In conclusion, the proposed MEAN model effectively integrates multimodal physiological signals to improve attention level prediction, offering a robust and accurate solution for applications requiring attention level monitoring. This research provides a foundation for advancing applications in education, work efficiency assessment, and cognitive enhancement technologies, highlighting the potential of multimodal approaches for understanding and predicting attention states.