Efficiently Updating ECG-Based Biometric Authentication Based on Incremental Learning.

Area of Science:

• Biometrics
• Signal Processing
• Machine Learning

Background:

• Biometric authentication using electrocardiograms (ECGs) is gaining interest.
• ECG signal variability due to emotional or physical states poses authentication challenges.

Purpose of the Study:

• To propose an adaptive ECG-based authentication method using incremental learning.
• To address ECG signal variability for robust subject identification.

Main Methods:

• An incremental support vector machine (SVM) was employed for adaptive authentication.
• ECG data from 11 subjects over six days were collected; days 1-5 for training, day 6 for testing.

Main Results:

• Incremental learning reduced the false acceptance rate from 6.49% to 4.39%.
• True acceptance rate increased from 61.32% to 87.61% per ECG wave after incremental learning.
• Testing one day after the latest training showed a false acceptance rate within 3.5-5.33% and improved true acceptance rates (70.05-87.61%).

Conclusions:

• The proposed adaptive ECG authentication method effectively handles signal variability.
• Incremental learning significantly enhances the accuracy and reliability of ECG-based biometrics.