Biometric Individual Authentication System using High Performance ECG Fiducial Features

The Electrocardiographic (ECG) recording is a reliable human heart vital status measurement. Automatic processing of these signals through several computational approaches such as machine learning tools has recently emerged in modern biometric systems. Evaluating ECG potential for biometrical applications has been the purpose of several research papers. In this paper, we developed a new model for individual authentication by detecting high-performance fiducial features of ECG signals. We used SVM and Naive Bayes classifiers to study the impact of high-order statistical features of QRS complexes and R-R intervals within ECG-ID and MIT-BIH Arrhythmia Databases. We integrated these features into a biometric model that we developed. The system reaches an accuracy of 96% up to 99% for the ECG-ID and MIT-BIH databases, respectively. The obtained results approve the reliability of the developed model for robust biometric recognition.