A Machine Learning Approach for Person Authentication from EEG Signals

Nowadays, electroencephalogram (EEG) is an emerging modality in the field of biometrics and helps to acquire characteristic brain signals from the scalp surface corresponding to various activity states. This paper presents a computerized method for automatically identifying individuals based on their recorded EEG signals A public dataset specifically designed for testing EEG biometric approaches was used. This data was collected over three sequential sessions involving 12 different stimuli and 21 subjects. During authentication, the recorded test EEG pattern was matched with the respective template stored in the database. The experiments show that the proposed random forest-based machine learning model can achieve approximately 83.2% authentication accuracy. This demonstrates that EEG could be reliable for biometric identification and authentication across various contexts.