THE STUDY AI-POWERED BIOMETRIC IDENITIFICATION SYSTEMS AUTHENTICATION MECHANISM USING ELECTROCARDIOGRAMS INFORMATION

Abstract

The foundation for developing relevant Machine Learning (ML) approaches to build Electrocardiogram (ECG) based biometric authentication systems is presented in this study. The suggested framework may assist researchers and developers working on ECG-based biometric authentication methods in defining the parameters of necessary datasets and obtaining high-quality training data. Use case analysis is used to establish dataset bounds. Three separate use cases, or authentication categories, are established based on different application situations using ECG based authentication. Increasing the amount of qualified training data provided to machine learning schemes that correlate with them would raise the accuracy of machine learning-based ECG biometric identification methods. This framework uses the ECG time slicing approach with the R-peak anchoring to get high-quality ML training data. Four additional measure indicators are included in the suggested framework to assess the caliber of ML training and testing data. Additionally, a Matlab toolbox is created and made accessible to the public for additional research. It includes all suggested mechanisms, measurements, and example data with demos utilizing different ML algorithms. The suggested framework may guide researchers in creating the appropriate ML settings, ML training datasets, and three user case scenarios in order to build ML-based ECG biometric authentication. In order to generate high-quality ML-based training and testing datasets and to leverage new measure metrics, the suggested framework remains valuable for researchers who are embracing ML approaches to create novel schemes in various study fields.