Parallel factor analysis can decompose human EEG during a finger movement task

To achieve a non-invasive brain computer interface by using the human scalp EEG, we applied the parallel factor analysis (PARAFAC) decomposition to the EEG data measured during the self-initiated and externally cued finger movement tasks. The results showed that the PARAFAC model can decompose the human scalp EEG and identify the intention, which concerns with the finger movement. By applying the core consistency diagnostic to the PARAFAC model, we discussed the feasible method for identifying the appropriate number of the components. Our results suggest that the combination of the EEG and other neuroimaging modality should be applied for producing the appropriate number of the components of the PARAFAC model for the purpose of the non-invasive brain computer interface, while the core consistency diagnostic could decide the minimum number of the components.