A MACHINE LEARNING FRAMEWORK FOR BIOMETRIC AUTHENTICATION USING ELECTROCARDIOGRAM (ECG)

Abstract

This paper presents a system for how with suitably embrace and modify AI (ML) techniques used to construct electrocardiogram (ECG)- based biometric authentication systems. The proposed system can assist agents and engineers in ECG-based biometric verification components to define the boundaries of required datasets and get preparing information with great quality. To determine the limits of datasets, a use case analysis is conducted. In light of different application scenarios for ECG-based verification, three distinct use cases (or validation classes) are developed. By providing more qualified preparing information given to corresponding AI models, the accuracy of ML-based ECG biometric authentication systems are expanded in result. The ECG time cutting method with the R-top mooring is utilized in this system to secure ML preparing information with great quality. In the proposed system, four new measurement metrics are acquainted with assess the quality of the ML training and testing data. Additionally, a Matlab toolkit, containing all proposed tools, metrics, and test data with exhibitions utilizing different ML techniques, is developed and made publicly available for further analysis. For developing ML-based ECG biometric authentication, the proposed system can guide experts to establish the appropriate ML solutions and the ML training datasets along with three identified user case scenarios. For analysts taking on ML techniques to design new systems in other research domains, the proposed framework