In-Ear EEG Biometrics for Feasible and Readily Collectable Real-World Person Authentication

Abstract

The use of EEG as a biometrics modality has been investigated for about a decade, however its feasibility in real-world applications is not yet conclusively established, mainly due to the issues with collectability and reproducibility. To this end, we propose a readily deployable EEG biometrics system based on a 'one-fits-all' viscoelastic generic in-ear EEG sensor (collectability), which does not require skilled assistance or cumbersome preparation. Unlike most existing studies, we consider data recorded over multiple recording days and for multiple subjects (reproducibility) while, for rigour, the training and test segments are not taken from the same recording days. A robust approach is considered based on the resting state with eyes closed paradigm, the use of both parametric (autoregressive model) and non-parametric (spectral) features, and supported by simple and fast cosine distance, linear discriminant analysis and support vector machine classifiers. Both the verification and identification forensics scenarios are considered and the achieved results are on par with the studies based on impractical on-scalp recordings. Comprehensive analysis over a number of subjects, setups, and analysis features demonstrates the feasibility of the proposed ear-EEG biometrics, and its potential in resolving the critical collectability, robustness, and reproducibility issues associated with current EEG biometrics.• Recording scalp-EEG with multiple electrodes is timeconsuming to set-up and cumbersome to wear. Such a

On the Influence of Affect in EEG-Based Subject Identification

“…A 99% authentication accuracy has been reported in [30] using single-channel signals from consumer-grade EEG sensor technology, by choosing custom tasks and custom acceptance thresholds for each subject. More recently, and with the aim to ease acquisition and better fulfill the collectability requirement, the use of a wearable in-ear EEG sensor has been proposed [32]. Results using Power Spectral Density (PSD) features and Support Vector Machines (SVM) with different kernels showed an average accuracy of 95.7% at a user verification task, without mixing the training and validation data from the same recording days.…”

“…[47], or by randomly selecting training and validation samples regardless of the data acquisition days [48]. This was called the biased scenario in [32], and can be affected by session-specific exogenous conditions (e.g. capacitative coupling of electrodes and cables with other devices, induction loops created between the employed equipment and the body, power supply artifacts) [35] or other noise-dependent features related to contamination artefacts from subjects' movements (e.g.…”

“…capacitative coupling of electrodes and cables with other devices, induction loops created between the employed equipment and the body, power supply artifacts) [35] or other noise-dependent features related to contamination artefacts from subjects' movements (e.g. eye blinks, chewing) [32]. This approach benefits from the recording-daydependent EEG characteristic in the classification, it is hence biased in favour of high classification rate and it cannot evaluate the repeatability/reproducibility of the application [32].…”

“…eye blinks, chewing) [32]. This approach benefits from the recording-daydependent EEG characteristic in the classification, it is hence biased in favour of high classification rate and it cannot evaluate the repeatability/reproducibility of the application [32]. Still, recordings for the DEAP and MAHNOB datasets were acquired during a single session per subject and do not allow for a cross-session-validation study.…”

“…The matrices contain the value of the cosine distances encoded as a color according to the legend shown at the right-hand side. Please, note that the first two subjects in MAHNOB have almost double the samples than the rest 3.3.2 Protocol P2: Rigorous scenario The importance of using data recorded across multiple sessions per user to validate the robustness of an EEG-based biometrics application has been remarked in a number of previous works [32], [35]. It is widely accepted that a sound experimental setting for biometric purposes should include recordings for each user over multiple days.…”