The Wavelet packet decomposition features applied in EEG based authentication system

Abstract

Biometric authentication is recently used for verification someone’s identity according to their physiological and behavioural characteristics. The most popular biometric techniques are fingerprints, facial and voices recognition. However, these techniques have disadvantage in which it can be easily to be imitated and mimicked by hackers to access a device or a system. Therefore, this study proposed electroencephalogram (EEG) as a biometric technique to encounter this problem. The wavelet packet decomposition is explored in this study for feature extraction method. The wavelet packet decomposition feature is represented in the average wavelet, root mean squared (RMS) wavelet and power wavelet were selected as features to extract a meaningful information from the original EEG signal based on the visual representation. These features were applied to classify between familiar and unfamiliar image responses (visual representation) and to recognize 13 subjects by using Support Vector Machine (SVM), k-Nearest Neighbor (KNN) and Random Forest (RF). The analysis of the classification between familiar and unfamiliar images responses obtained that gamma frequency (30 – 45 Hz) achieved the highest correct recognition rate (CRR) and KNN obtained the accuracy of 92.8% was obtained with KNN in the classification between familiar and unfamiliar image responses. Using the gamma frequency band, the classification between the EEG responses of the 13 subjects was evaluated using the percentage of false acceptance rate (FAR) and false rejection rate (FRR). From the overall view, the value of FAR is lower than FRR. These values were used in authentication system as threshold for security level. As the result of classification between the subjects, SVM performed better compared as KNN and RF in which the error rate for acceptance of unauthorized person and rejection of authorized person were the lowest.

A KCP-DCNN-Based Two-Step Verification Multimodal Biometric Authentication System featuring QR Code Fabrication

“…Many research articles incorporate biometrics that has gone via several stages from preprocessing and feature detection of biometric data to machine learning techniques utilizing extracted features and deep learning approaches without feature engineering (Chen et al, 2021) (Stergiadis et al, 2022). In computational approaches, neural networks are commonly utilized for recognition as they provide high precision in this area (Rosli et al, 2021). An increase in authentication performance is reported by machine learning and deep learning approaches in detecting distinct characteristics that differentiate individuals from others.…”