Single heartbeat ECG authentication: a 1D-CNN framework for robust and efficient human identification.

Area of Science:

• Biometrics
• Machine Learning
• Signal Processing

Background:

• Individual authentication systems require robust and efficient methods.
• Electrocardiogram (ECG) signals contain unique physiological patterns suitable for biometric identification.
• Existing authentication methods may face challenges with accuracy, scalability, or real-time processing.

Purpose of the Study:

• To propose and evaluate a novel 1D-CNN framework for individual authentication using single ECG heartbeats.
• To assess the sufficiency of a single heartbeat segment for creating a robust biometric system.
• To investigate the effectiveness of SMOTE for handling data imbalance in ECG-based authentication.

Main Methods:

• A 1D-CNN framework was developed for individual authentication.
• Single heartbeat samples were generated from ECG signals using R-to-R segmentation, length thresholding, and interpolation.
• The Synthetic Minority Oversampling Technique (SMOTE) was applied to address sample distribution imbalances.
• The framework was validated on four public ECG databases (NSRDB, MIT-ARR, ECG-ID, MIMIC-III).

Main Results:

• The 1D-CNN framework achieved perfect scores (100% accuracy, precision, sensitivity, F1-score) on individual NSRDB and MIT-ARR databases.
• Performance remained high (>99.6%) on mixed datasets with larger populations and diverse conditions.
• The model demonstrated excellent scalability across small and large subject groups.

Conclusions:

• A single heartbeat segment is sufficient for a robust 1D-CNN based individual authentication system.
• The proposed framework offers high accuracy and scalability, suitable for security applications.
• Future work should explore multimodal biometrics and real-time implementation for broader applicability.