MindID: Person Identification from Brain Waves through Attention-based Recurrent Neural Network

Person identi€cation technology recognizes individuals by exploiting their unique, measurable physiological and behavioral characteristics. However, the state-of-the-art person identi€cation systems have been shown to be vulnerable, e.g., antisurveillance prosthetic masks can thwart face recognition, contact lenses can trick iris recognition, vocoder can compromise voice identi€cation and €ngerprint €lms can deceive €ngerprint sensors. EEG (Electroencephalography)-based identi€cation, which utilizes the user’s brainwave signals for identi€cation and o‚ers a more resilient solution, draw a lot of aŠention recently. However, the accuracy still requires improvement and very liŠle work is focusing on the robustness and adaptability of the identi€cation system. We propose MindID, an EEG-based biometric identi€cation approach, achieves higher accuracy and beŠer characteristics. At €rst, the EEG data paŠerns are analyzed and the results show that the Delta paŠern contains the most distinctive information for user identi€cation. Œen the decomposed Delta paŠern is fed into an aŠention-based Encoder-Decoder RNNs (Recurrent Neural Networks) structure which assigns varies aŠention weights to di‚erent EEG channels based on the channel’s importance. Œe discriminative representations learned from the aŠention-based RNN are used to recognize the user’ identi€cation through a boosting classi€er. Œe proposed approach is evaluated over 3 datasets (two local and one public). One local dataset (EID-M) is used for performance assessment and the result illustrate that our model achieves the accuracy of 0.982 which outperforms the baselines and the state-of-the-art. Another local dataset (EID-S) and a public dataset (EEG-S) are utilized to demonstrate the robustness and adaptability, respectively. Œe results indicate that the proposed approach has the potential to be largely deployment in practice environment.