ECG based biometric identification method using QRS images and convolutional neural network

Electrocardiogram (ECG) signals, which are commonly used in medical applications, have been started to use as a biometric modality for biometric applications thanks to its liveness indicator that makes it stronger against spoofing attacks. Due to improving computational power of computer systems, several convolutional neural network (CNN) based methods have been recently proposed for ECG biometric identification in order to increase identification performance and classification accuracy. In this work, we proposed an ECG based biometric identification method using QRS (QRS wave) images and two-dimensional CNN. In the(dagger) proposed method, ECG signals were segmented by applying noise removing and QRS detection algorithms. After these segments were aligned according to their R-points, they were transformed to two-dimensional ECG signals called QRS images of size 256x256. Finally, biometric identification task was achieved by developing a CNN based ECG biometric identification method which uses the QRS images as an input. The identification performance of the proposed method was compared to other CNN based ECG biometric identification methods proposed in the literature. The experimental results show that the proposed method provides an accuracy of 98.08% and an identification rate of 99.275% for a public ECG database of 46 persons.