ECG signals analysis for biometric recognition

Electrocardiogram (ECG) as a new biometric trait has the advantage of being a liveliness indicator and difficult to be spoofed or falsified. According to the utilized features, the existing ECG based biometric systems can be classified to fiducial and non-fiducial systems. The computation of fiducial features requires the accurate detection of 11 fiducial points which is a very challenging task. On the other hand, non-fiducial approaches relax the detection process but usually result in high dimension feature space. This paper presents a systematic study for ECG based individual identification. A fiducial based approach that utilizes a feature set selected by information gain IG criterion is first introduced. Furthermore, a non-fiducial wavelet based approach is proposed. To avoid the high dimensionality of the resultant wavelet coefficient structure, the structure has been investigated and reduced using also IG criterion. The proposed feature sets were examined and compared using radial basis functions (RBF) neural network classifier. The conducted experiments using Physionet databases revealed the superiority of our suggested non-fiducial approach.