Real-Time Multimodal Classification of Internal and External Attention

The current attentional state can be divided into several categories, for example, the direction of attention. Often, this state is subconscious or its constant report impossible. Thus, an automated surveillance of the attentional state could be beneficial. In this paper, we performed a classification of multimodal data (EEG and eye tracking) to model internally- and externally-directed attention. 10 participants performed 6 different tasks of which 3 were associated with internal and 3 with external attention. In the first step, we showed that a combination of the two modalities led to an improvement of classification accuracy (average 72.67%) compared to single modality classifications. In a second step, the analysis was performed in real-time. The system was tested on one participant with an average accuracy of 60.87%. These results allow for an optimistic outlook on a reliable real-time multimodal classification system of internal and external attention.