Date

2022-2-10

Author

Yahyaei, Reza

Metadata

Show full item record

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Item Usage Stats

614
views

491
downloads

Cite This

Continuous biometrics using electroencephalography (EEG) has attracted noticeable amount of research for some time. Many state-of-the-art methods have been proposed and achieved satisfactory performances of more than 95 % in recognition rate. In particular, features based on spectral bands and functional connectivity have been preferred for their superior performance. Unlike majority of the previous research, this study chose to focus on the efficiency and ease of implementation. In this direction, we propose a new feature for EEG-based biometrics called the mean curve length (MCL), which is a simple measure of signal complexity based on the Katz fractal dimension. In this study, we evaluated its performance in person identification and authentication, and compared it with other features. For this, a large standard dataset comprising 109 subjects under the eyes-open (EO) and eyes-closed (EC) resting state conditions was utilized. In order to keep the results realistic, minimal preprocessing was performed on the signals, and no subjects or channels were excluded based on their artifacts. A Mahalanobis distance-based classifier was employed for both identification and authentication tasks. For high-dimensional features such as functional connectivity metrics, we used principle component analysis to implement a modified Mahalanobis classifier incorporating dimensionality reduction. The results of our analyses indicated that MCL provides a remarkably high individual distinction comparable to the commonly preferred features, while being vastly more efficient. Specifically, the recognition accuracies were 99.4 % (EO) and 98.8 % (EC) for identification, and for authentication, the equal error percentages of 6.33 % (EO) and 10.50 % (EC) were obtained. Our study demonstrates MCL as a fast and accurate biometric feature that is promising for real-life and real-time applications. It promotes the effectiveness of nonlinear signal measures in individual discrimination, and encourages to look beyond the conventional time and frequency domain measures of brainwaves.

Subject Keywords

Mean Curve Length, Fractal Dimension, EEG, Biometrics, Mahalanobis Classifier

URI

https://hdl.handle.net/11511/96714

Collections

Graduate School of Informatics, Thesis