Towards online applications of EEG biometrics using visual evoked potentials
Introduction
Recently, the electroencephalogram (EEG) method has attracted increasing attention among various bio-signals in the study of biometrics (Campisi and La Rocca, 2014, Thomas and Vinod, 2017). EEG activities show discriminant capabilities among subjects due to the morphological, anatomical, and functional traits of the brain. Compared with other commonly used biometrics such as fingerprints, faces, and irises, EEG offers advantages including robustness against attacks, continuous identification capability, intrinsic liveness detection, and universality (Campisi & La Rocca, 2014). Moreover, EEG has good potential in real-world biometric applications due to its characteristics of noninvasiveness, portability, and low cost.
Although EEG has shown great potential in biometrics, it remains a major challenge to design and implement a practical EEG-based biometric system for person identification. Although high correct recognition rates (CRRs) have been reported in the literature, a long duration of EEG data is still required due to the low signal-to-noise ratio (SNR) of EEG. The spontaneous EEG has always required data lengths up to minutes (Fraschini et al., 2015, Maiorana et al., 2016a, Maiorana et al., 2016b, Rocca et al., 2014, La Rocca et al., 2013, Crobe et al., 2016, Moctezuma et al., 2018, Nakamura et al., 2018). In contrast, visual evoked potentials (VEPs) can achieve higher speed and accuracy (Palaniappan, 2004, Min et al., 2017). For example, 1-s long VEP data elicited by object pictures were used to achieve a CRR of 99.06% (Palaniappan, 2004). Another study based on steady-state visual evoked potentials obtained a CRR of 98.60% using 5-s long EEG data (Min et al., 2017).
VEPs are the electrical brain potentials evoked by visual stimulation, which can be measured from the occipital area on the scalp. According to the modulation approaches of the stimulation signals, the VEPs used in person identification can be categorized into three major types (Min et al., 2017, Armstrong et al., 2015, Zhao et al., 2019, Zúquete et al., 2010, Ravi and Palaniappan, 2006, Snodgrass and Vanderwart, 1980, Palaniappan and Mandic, 2007a, Yeom et al., 2013): flash-VEPs (f-VEPs), steady-state VEPs (ss-VEPs), and code-modulated VEPs (c-VEPs). By using f-VEPs evoked by flashing pictures, a CRR of 89% was achieved among 15 subjects (Armstrong et al., 2015). The ss-VEPs obtained a CRR of 98.60% across 20 subjects using flickering grid-shaped line array stimulations (Min et al., 2017). The c-VEPs elicited by m-sequence stimulations achieved a CRR of 99.43% among 25 subjects in cross-session identification (Zhao et al., 2019). Although existing studies reported good performance with the three types of VEPs in person identification, a systematic comparison of these signals is still lacking. The selection of an optimal visual stimulus might be crucial for eliciting significant and robust individual difference in VEPs, which can facilitate person identification. This study therefore aimed to quantitatively compare the performance of the three types of VEPs in person identification.
Although the theoretical and methodological aspects of EEG-based biometrics have been investigated extensively, there are very few studies that demonstrate online EEG-based biometric systems (Harshit et al., 2016). To the best of the authors’ knowledge, an online demonstration of an EEG-based person identification system is still lacking, especially in a more realistic cross-session condition. The feasibility and practicality of an online EEG-based person identification system remain unknown. Permanence is a major problem in EEG-based person identification, which requires further longitudinal validations. The reproducibility of stable performance across multiple sessions is generally questionable due to the session-to-session variability of EEG signals. Very few studies have performed longitudinal evaluations, and all evidently showed an aging effect across sessions (Armstrong et al., 2015, Maiorana and Campisi, 2018). Furthermore, a practical biometric system should satisfy the requirements of online implementation of data recording and analysis. The issues of reliability and real-time performance on hardware and software require additional efforts. Recent advances in brain-computer interfaces (BCIs) (Gao et al., 2014) can provide useful guidelines for the design and implementation of online platforms for EEG acquisition and analysis. BCIs detect different classes of EEG signals for the same subject (i.e., the difference depends on brain states), whereas EEG biometrics detects individual difference in the same class of EEG signal (i.e., the difference depends on brains). Although the basic concepts and underlying mechanisms for BCIs and EEG biometrics are different, an EEG-based biometric system can be implemented in a very similar way to EEG-based BCI systems. The system framework consists of three major components: EEG recording, feature extraction and classification, and feedback presentation. Therefore, the study also aimed to implement an online VEP-based biometric system for person identification in a cross-session condition.
Here, we propose a general system framework for VEP-based person identification and demonstrate an online system with high CRRs. Compared with existing EEG-based biometric systems, the proposed system integrates the most advanced techniques in VEP-based BCIs with an EEG-based person identification system. Visual stimulation presentation and VEP detection play a key role in the implementation of the proposed system. The novelties and contributions of this study include: (1) a general system framework for person identification based on VEPs, (2) a quantitative comparison of the performance of three types of VEPs in person identification, and (3) a thorough demonstration of an online VEP-based biometric system for cross-session person identification.
Section snippets
System framework
Fig. 1 shows the VEP-based person identification framework, including two isolated steps: enrollment and identification. The enrollment step collected VEP data from a group of subjects to build a user database. For each subject, multiple VEP trials were required to train an individual model consisting of a VEP template and a spatial filter. Individual models labeled by subjects’ IDs were pooled in the database for later identification. In the identification step, a few trials were collected to
Offline identification
To evaluate the identification performance, we calculated CRRs for the three types of VEPs separately. To further explore inter-session variability, intra-session identification was compared with cross-session identification. The identification session data were used to evaluate intra-session performance using the leave-one-out cross-validation strategy.
Fig. 5 shows averaged waveforms of f-VEPs, ss-VEPs, and c-VEPs for two example subjects (s1 and s5) in two distinct sessions (d1 and d2). As
Discussion
The basic concept and underlying mechanism of EEG-based biometrics have been proposed previously (Campisi and La Rocca, 2014, Thomas and Vinod, 2017, Del Pozo-Banos et al., 2014). The results of our study demonstrate the discriminant capability of VEPs for person identification (see Fig. 6, Fig. 7). The individual differences in VEPs can be generally attributed to the morphological, anatomical, and functional specificity of the visual cortex. To our knowledge, this study is the first to
Conclusions
In this study, we systematically compared the performance of three types of VEP signals in EEG-based person identification. The results showed that the c-VEP paradigm achieved the highest CRRs of 100% using 3.15-s VEP data (a 5.25-s duration including 2.1-s intervals) in the intra-session condition and 99.48% using 10.5-s VEP data (a 17.5-s duration including 7-s intervals) in the cross-session condition with a group of 21 subjects. An online system demonstration using c-VEPs was successfully
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
This work was supported in part by the National Key Research and Development Plan of China (2017YFA0205903), the National Natural Science Foundation of China (61671424, 61335010, and 61634006), and the Strategic Priority Research Program of the Chinese Academy of Science (XDB32040200).
References (45)
- et al.
Brainprint: Assessing the uniqueness, collectability, and permanence of a novel method for ERP biometrics
Neurocomputing
(2015) - et al.
Electroencephalogram subject identification: A review
Expert Systems with Applications
(2014) Deep learning for EEG-based biometric recognition
Neurocomputing
(2020)- et al.
EEG-based person identification through Binary Flower Pollination Algorithm
Expert Systems with Applications
(2016) - et al.
Electrophysiological evidence for an early attentional mechanism in visual processing in humans
Vision Research
(1999) - et al.
EEG-based user identification system using 1D-convolutional long short-term memory neural networks
Expert Systems with Applications
(2019) - et al.
Task-related component analysis for functional neuroimaging and application to near-infrared spectroscopy data
NeuroImage
(2013) - et al.
Steady-state visually evoked potentials: Focus on essential paradigms and future perspectives
Progress in Neurobiology
(2010) - et al.
Person authentication from neural activity of face-specific visual self-representation
Pattern Recognition
(2013) - et al.
State-of-the-art methods and future perspectives for personal recognition based on electroencephalogram signals
IET Biometrics
(2015)
The psychophysics toolbox
Spatial Vision
Brain waves for automatic biometric-based user recognition
IEEE Transactions on Information Forensics and Security
High-speed spelling with a noninvasive brain–computer interface
Proceedings of the National academy of Sciences of the United States of America
Filter bank canonical correlation analysis for implementing a high-speed SSVEP-based brain–computer interface
Journal of Neural Engineering
A high-security EEG-based login system with RSVP stimuli and dry electrodes
IEEE Transactions on Information Forensics and Security
Minimum spanning tree and k-core decomposition as measure of subject-specific EEG traits
Biomedical Physics & Engineering Express
Using rapid visually evoked EEG activity for person identification
EEG biometrics for user recognition using visually evoked potentials
EEG biometrics using visual stimuli: A longitudinal study
IEEE Signal Processing Letters
An EEG-based biometric system using eigenvector centrality in resting state brain networks
IEEE Signal Processing Letters
Visual and auditory brain-computer interfaces
IEEE Transactions on Bio-Medical Engineering
Exploiting the P300 paradigm for cognitive biometrics
IJCB
Cited by (18)
Cybersecurity in neural interfaces: Survey and future trends
2023, Computers in Biology and MedicineAutomatic detection of code-modulated visual evoked potentials using novel covariance estimators and short-time EEG signals
2022, Computers in Biology and MedicineCitation Excerpt :The authors used a template-matching-based identification method and they found that the c-VEPs obtained the highest correct recognition rates (CRRs) of 100% and 99.48% in the intra-session and cross-session conditions, respectively. Their results indicated that the c-VEP paradigm can achieve the highest CRR in EEG-based person identification system [26]. Zheng et al. proposed a high-performance brain switch based on cVEPs to realize a fast reaction without false positives (FP).
A wavelet-based capsule neural network for ECG biometric identification
2022, Biomedical Signal Processing and ControlCitation Excerpt :However, extensive research has shown that they are more likely to be hacked and manipulated by impersonation attacks [1]. In recent years, physiological signals such as electroencephalography (EEG) [2,3] electromyography (EMG) [4,5], electrocardiogram (ECG) [6–8], and photoplethysmography (PPG) [9,10], have demonstrated strong potential in biometric recognition systems. Among these physiological signals, the ECG signal, which is a recording of electrical activities of the heart in a non-invasive way, has been demonstrated to have the needed characteristics such as uniqueness, universality, liveness detection, permanent acquisition for a robust and reliable recognition process.
An Overview of Brain Fingerprint Identification Based on Various Neuroimaging Technologies
2024, IEEE Transactions on Cognitive and Developmental SystemsPerson identification using autoencoder-CNN approach with multitask-based EEG biometric
2024, Multimedia Tools and ApplicationsA Review on Applications of Electroencephalogram: Includes Imagined Speech
2024, International Journal of Advanced Computer Science and Applications