ECG Based Biometric Authentication System using Deep Learning Methods

This paper is based on electrocardiogram (ECG) signals, which are easy to acquire, and research shows that it is reliable for biometric authentication. To improve the speed of authentication, achieving more stable results with different datasets, a deep convolutional neural network based on multiresolution analysis to extract features that will allow performing a closed-set authentication is proposed in this paper. This network utilises a far simpler block architecture inspired by VGG architecture, and the less dense architecture improves speed. This proposed method improved classification accuracy more than the previously proposed schemes. The method’s performance is calculated using the ECG-ID database and the MIT-BIH Arrhythmia Database. It produces an accuracy of approximately 99% on both datasets. Finally, a comparison was made between the proposed classifier's performance(accuracy, speed) and well-known CNN architectures.