What the Heart Can(not) Tell: Potential and Pitfalls of Biometric Recognition Methods Based on Photoplethysmography

Area of Science:

• Biometrics
• Wearable Technology
• Physiological Signal Processing

Background:

• Wearable devices collect human physiological signals for applications like biometric authentication.
• Previous studies show potential for physiological signals in unique individual identification, but real-world validity is limited.
• Existing research often uses controlled settings, small datasets, and short-term evaluations, leading to optimistic performance estimates.

Purpose of the Study:

• To assess the reliability of photoplethysmography (PPG)-based biometric methods.
• To evaluate existing and novel PPG biometric approaches under various conditions.
• To propose guidelines for robust evaluation of PPG-based biometrics.

Main Methods:

• Replication of two published PPG biometric methods.
• Introduction and evaluation of a feature-based PPG biometric baseline method.
• Evaluation across multiple conditions, including laboratory and real-world scenarios.

Main Results:

• PPG biometric methods perform well on laboratory datasets.
• Effectiveness significantly declines in real-world environments due to signal variability, larger user populations, and temporal separation.
• Current systems face challenges with unseen users and longitudinal data.

Conclusions:

• PPG-based biometrics require more rigorous, real-world evaluation.
• Proposed guidelines emphasize longitudinal datasets, temporal splits, unseen-user assessments, and transparent reporting.
• Recommendations aim to enhance the reliability and reproducibility of physiological biometrics research.