A Study Of Ecg-based Biometrics Technology

With increasing concern about information security in modern society, how to recognize human identification reliably has become an imp ortant research topic. Conventional methods such as certificates, passwords, etc. are not going to meet people’s need. Along with the development in technology, Biometric Identification Technology(BIT) is attracting more people and various related projects are coming out. By now, the biological features such as fingerprint, palm print, face and voice are widely used. However, they all have visible disadvantages that can not be ignored. Therefore, ECG(Electrocardiograph) based biometrics technology method has been proposed and gained lots of attention. ECG-based biometrics technology has more advantages than other technologies. It has high safety factor, and it is more suitable for medical fields than any other methods. ECG-based biometrics technology will be a powerful complement for current biometric identification systems.This thesis mainly studies the ECG-based biometrics technology. The main process can be divided into four parts: the preprocess of ECG, feature extraction, feature selection and classification. ECG signals in this study were filtered properly by the improved morphological filter, in order to remove noises including baseline drift, frequency interference, and so on. This thesis also analyzed the ECG’s biological features. Then, the processed ECG were analyzed with reference points detected in time domain and comprehensive features were extracted accordingly. These features have possibilities to be used for personnal identification. In this thesis, a comprehensive feature’s detection was carried out in the time domain. After that, This study used the variance analysis method to analyze the significant difference in the ECG groups,to find the stability of the features. Then, analyzed the significant difference between the groups to find the individual varieties of the features. This thesis also used wavelet packet decomposition method to extract the ECG’s energy distribution information, and fused with the features of time domain, to improve the identification accuracy. At last, this thesis used the support vector machine to deal with the laboratory datas, and the datas from the ECG-ID database. The experimental results show that the ECG features obtained by this thesis can achieve good identification results under different status.