Learning of Brain Connectivity Features for EEG-based Person Identification

The brain activity observed on multiple EEG electrodes is influenced by volume conductance and functional connectivity of a person performing a task. When the task is a biometric test, EEG signals represent the unique 'brain print' which is genetically defined by the functional connectivity that is represented by interactions between the electrodes, whilst the conductance component causes trivial correlations in EEG signals. Orthogonalisation using autoregressive modelling minimises the conductance component, and the connectivity features can be then extracted from the residuals. However, the results cannot be reliable for high-dimensional EEG data recorded via a multi-electrode system. The proposed method shows that the dimensionality can be significantly reduced if baselines that are required for estimating the residuals can be modelled by using EEG electrodes that make important contribution to the functional connectivity. The results show that the required models can be learnt by Machine Learning techniques which are capable of providing the maximal performance in the case of multidimensional EEG data. The study which has been conducted on a EEG benchmark including 109 participants shows a significant improvement of the identification accuracy.