A Study On Resting Electroencephalogram(EEG) Based Biometrics In Ubiquitous Environment

With the booms of mobile communication, especially mobile smart phone, technologies to identify individuals for mobile security calls for some more strict requirements in user-friendly, real-time and ubiquitous aspects. In addition to traditional approaches (for example, password check), some advanced biometric methodologies have been applied in practice, such as fingerprint and iris based solutions; however, these solutions generally lack a true ubiquitous nature for mobile security. Electroencephalogram (EEG) reflects the electrophysiological activities of brain. Some previous studies have demonstrated the uniqueness of EEG in different persons which is adequate for biometrics. In this paper, aiming at the real-time implementation in ubiquitous environments, I propose an EEG based approach to collect and process single-channel resting EEG signals and then to identify individuals.The main tasks of this thesis are as follows:1)Design ubiquitous EEG collection module which should use less electrodes and less time in collecting data for EEG biometrics;2)Design efficient EEG de-noising and processing module which could automatically remove noise, and then extract useful EEG features in time and frequency domain.3) Compare the results of three kinds of classifiers:k-Nearest Neighbors (kNN), Fisher’s Discriminant Analysis and Back Propagation (BP) Neural Network (NN). And then design a signal-to-noise (SNR) solution to exclude the intruders.In this paper, I conduct experiments which comprise three types of tests:accuracy test, time dimension test and capacity dimension test. The results of these experiments demonstrate that the accuracy rate can be over90%when we collect EEG signals over50seconds. In the SNR based intruder-exclusion technology, with the increase of SNR threshold, the true acceptance rate (TAR) of clients will decrease and the true rejection rate (TRR) of intruders will increase. So, it proves the considerable performance of intruder-exclusion technology, as well as provides us a flexible selection of SNR threshold according to the practical requirements. In addition, with the prolongation and extension of EEG database, the accuracy will decline, which indicates my solution is just adequate for individual identification within small pools and the EEG database should be updated frequently. In all, my work validates the ubiquitous EEG based biometrics in multiple aspects.