Comparison of EEG-Based Biometrics System Using Naive Bayes, Neural Network, and Support Vector Machine

Various systems for recognizing or authenticating a person's identity have been developed using biometric-based technology. However, the current authentication system still has several weaknesses, such as being easily imitated, low accuracy, and being highly susceptible to manipulation. Electroencephalogram (EEG) is one of the biometric features obtained from electrical activity in the human brain. EEG signals tend to be different (unique) from one another. For this reason, in our research, we are trying to build a biometric (person recognition) system based on EEG signals with machine learning algorithms such as Naive Bayes, Neural Network, and Support Vector Machine (SVM). EEG data were taken from 43 participants at baseline condition (without stimulation) involving 4 channels such as FP1, FP2, F7, and F8. In addition, the device used for EEG recording is an OpenBCI Ultra Cortex Mark IV. The pre-processing stages such as Finite Impulse Response (FIR), Automatic Artifact Removal EOG (AAR-EOG), Artifact Subspace Reconstruction (ASR), and Independent Component Analysis (ICA) were done using Matlab EEGLab ToolBox. Power Spectral Density (PSD) is calculated as a feature for classification. In this study, the classification process is run on each channel by using all EEG sub-bands in order to find the best EEG channel that describes the best human biometrics system. It is found that the Neural Network produces the highest accuracy value of 97.7% on the F7 channel compared to SVM and Naïve Bayes.