EEG Identification Based on Brain Functional Network and Autoregressive Model

Due to the advantages of high concealment, non-stealing and liveness detection, electroencephalography (EEG) is a promising biometric technology for security requirements and personal identification. Many EEG identification method focuses on specific task. The generalization of these methods to other tasks or datasets is limited. In this work, we propose a new EEG identification method, which is based on the combination of brain functional network and autoregressive model, under different tasks. Specifically, we employ Phase Locking Value (PLV) between paired EEG channels to construct the functional connectivity matrix. Using a data-driven threshold, we get the corresponding binarized brain network. Then the degree of each node is obtained by graph theory. Additionally, the autoregressive (AR) coefficients are employed using Burg algorithm. The feature vector is formed combined these AR coefficients and node degrees. The support vector machine (SVM) is used to accomplish the EEG personal identification task. Experimental results on three public datasets with different tasks show that the proposed method has superior performance in terms of classification accuracy and stability.