Simple and efficient multibiometric technique for subject recognition using multilead electrocardiogram signal

In this paper, a simple and effective multibiometric technique for subject recognition using multiple lead electrocardiogram (ECG) signals is presented. The proposed technique significantly improves the recognition performance of a biometric system by using multiple sources available in the same modality group. A total of 30 subjects with 12 lead ECG measurements obtained from PTB Diagnostic ECG database (PTBDB) with sampling rate of 1000 Hz were used to verify the approach. Normalization plays an important role in the identification stage as it uniquely matches between ECG signals from bipolar limb leads and also the supplementary augmented unipolar limb leads. Based on the experimentation results, self-similarities are prominent and distinct from one person to another by obtaining high correlation values and relatively good classification accuracies ranging from 93% to 100% for all the leads. This result also suggests the robustness, reliability and stability of the proposed method for multibiometric system.