Transcending Conventional Biometry Frontiers: Diffusive Dynamics PPG Biometry.

Area of Science:

• Biometrics
• Signal Processing
• Machine Learning

Background:

• Photoplethysmographic (PPG) signals offer unique physiological data for biometrics.
• Existing PPG biometric methods often rely on morphological features, which can be unstable.
• A need exists for robust and anti-spoofing biometric authentication systems.

Purpose of the Study:

• To develop the first biometric authentication system based on PPG signal dynamic characteristics.
• To leverage a Siamese convolutional neural network (CNN) for analyzing PPG signal dynamics.
• To evaluate the system's robustness, anti-spoofing capabilities, and performance against existing methods.

Main Methods:

• Extraction of biometric characteristics from PPG signal diffusive dynamics using the 0-1 test and (p,q)-planes.
• Implementation of a Siamese CNN for analyzing these dynamic features.
• Training and validation using a database of 40 real-world PPG signals.

Main Results:

• The proposed method demonstrates superior performance, achieving the best equal error rate (ERR).
• The system exhibits excellent processing times compared to eight other PPG-based biometric methods.
• Dynamic characteristics of PPG signals proved more stable over time than morphological features.

Conclusions:

• PPG signal dynamic-based biometrics using Siamese CNNs provide a robust and secure authentication solution.
• The method's reliance on vascular bed biostructure ensures individual uniqueness and stability.
• This approach offers a promising advancement in anti-spoofing biometric technology.