ECG based person authentication using empirical mode decomposition and discriminant analysis

Abstract

Person identification or authentication through biometric features has been widely applied for basic access and high-level security. But conventional biometrics such as fingerprints and irises tend to be easily faked or duplicated. Therefore a new biometric modality is needed to overcome that problem. In this paper, we simulate a new model of biometric systems using physical signals of the body. The proposed biometric system is based on ECG signals as a characteristic of each subject. A total of 110 raw ECG signals with a duration of 5 seconds from 11 participants were demonstrated in the proposed system. Empirical mode decomposition (EMD) and statistical analysis are used for feature extraction. Discriminant analysis with cross-validation was applied to test the performance of the proposed method. In this research, the highest accuracy of 93.6% was obtained using subspace discriminant in the scenario of all feature attributes as predictors.