Multichannel EEG-based biometric using improved RBF neural networks

Electroencephalogram (EEG) brainwaves have recently emerged as a promising biometric that can be used for individual identification. In this study, we present a new visual stimuli-driven, non-volitional brain responses based methodological framework towards individual identification. The non-volitional mechanism provides an even more secure way in which the individuals are not aware of security credentials and thus can not manipulate their brain activities. Given the intercorrelated structure of brain functional areas, instead of making the identification decision relying on any single EEG channel, we propose a new identification approach based on the decision-level fusion of multichannel EEG signals, using the Radial Basis Function (RBF) neural network and its improved versions. Specifically, the identification decision is determined according to the identification patterns reflected from multiple EEG channels over the desired brain functional region. We evaluate the performance of our proposed methods based on four different visual stimuli and four independent EEG channels. Experimental results show that, the proposed fusion technique can significantly improve the identification accuracy, compared to the conventional single channel based solution. For RBF network, the accuracy of identifying 37 subjects could reach over 70%, which is better than the average accuracy of about 55% achieved through individual channels. For the improved RBF networks, the frequency-based decision making could reach the accuracy of 90%, while the probability-based method could reach over 91%. Our study lays a foundation for future investigation of more accurate and reliable brainwave-based biometrics.