Multimodal Biometric Systems: Combining Electroencephalogram and Facial Recognition for Robust Individual Identification and Verification

Abstract

The primary factor driving the growing interest in novel biometric features is the vulnerability of traditional methods like fingerprint and facial recognition to forgery. This study focuses on a multimodal biometric identification system that integrates data from electroencephalograms (EEG) and facial features. To derive valuable insights from EEG data, we apply signal processing techniques such as filtering, segmentation, and feature extraction, alongside Daubechies-4 (DB4) wavelet analysis with five decomposition levels. The enhanced facial video features include entropy calculations and tracking of facial measurements. Six classifiers Gaussian Naïve Bayes, K-Nearest Neighbour, Random Forest, AdaBoost, Support Vector Machine, and Multilayer Perceptron were trained utilizing the combined EEG and facial data. Findings reveal that AdaBoost and Random Forest emerged as the most effective classifiers for this application, achieving accuracies of 99.87±0.13% and requiring EEG recording times of 2s and 1.5s, respectively, showcasing excellent precision.