Person authentication using electroencephalogram (EEG) brainwaves signals

• Main Authors: Liew, Siaw Hong, Choo, Yun Huoy, Low, Yin Fen, Zeratul Izzah, Mohd Yusoh
• Other Authors: Wai, Yie Leong
• Format: Book Chapter
• Language: English
• Published: The Institution of Engineering and Technology 2019
• Subjects: QA75 Electronic computers. Computer science
• Online Access: http://ir.unimas.my/id/eprint/40730/1/Book%20Chapter%20%28IET%20EEG%20Signal%20Processing%29.pdf; http://ir.unimas.my/id/eprint/40730/; https://digital-library.theiet.org/content/books/10.1049/pbhe016e_ch4
• Institution: Universiti Malaysia Sarawak

This chapter starts with the introduction to various types of authentication modalities, before discussing on the implementation of electroencephalogram (EEG) signals for person authentication task in more details. In general, the EEG signals are unique but highly uncertain, noisy, and difficult to analyze. Event-related potentials, such as visual-evoked potentials, are commonly used in the person authentication literature work. The occipital area of the brain anatomy shows good response to the visual stimulus. Hence, a set of eight selected EEG channels located at the occipital area were used for model training. Besides, feature extraction methods, i.e., the WPD, Hjorth parameter, coherence, cross-correlation, mutual information, and mean of amplitude have been proven to be good in extracting relevant information from the EEG signals. Nevertheless, different features demonstrate varied performance on distinct subjects. Thus, the Correlation-based Feature Selection method was used to select the significant features subset to enhance the authentication performance. Finally, the Fuzzy-Rough Nearest Neighbor classifier was proposed for authentication model building. The experimental results showed that the proposed solution is able to discriminate imposter from target subjects in the person authentication task.