CNN-SVM approach for EEG-Based Person Identification using Emotional dataset

The potential of brain electrical activity elicited by a set of emotional stimulation is investigated in this paper to build a biometric identification system based on electroencephalogram (EEG). Deep learning method such as Convolutional neural network (CNN) is used to extract unique neural features from EEG signals knowing that the brain signature is a repeatable discriminative characteristic, then features are classified with polynomial kernel function Support Vector Machine (SVM). Participants are presented, in the database, with 20 emotional videos stimuli and 32 electrodes were primarily, employed for this research. Our biometric system can achieve 99.99% identification accuracy depending on the employed configuration furthermore, results conclude that 5-channels EEG data (PO3, PO4, O1, Oz, O2) filtered in Beta frequency range [14 – 30] Hz carry the most discriminative information appropriate to biometric brain traits.