Abstract:

BACKGROUND: High correlation between electroencephalogram (EEG) sources can cause rank deficit of the correlation matrix of EEG scalp recordings. Classical spatio-temporal EEG source localization cannot localize such sources.
OBJECTIVE: To develop a novel method to image correlated EEG sources.
METHODS: An algorithm, termed multivariate correlation coefficient matrix decompositions (MVMD), was proposed. Correlation index, obtained by decomposing cc matrix, was a measure of the degree of correlations between the specified channels and the system. The numerical experiment proved that this kind of matrix decomposition was correct and reasonable.
RESULTS AND CONCLUSION: MVMD transforms the relations between each pair of variables into those between each pair and the system generating all the variables. Such transform is useful in visualizing the relations between variables. It has promising prospect in cognitive neuroscience and other fields associated with multiple variable analysis.