Why Temporal Persistence of Biometric Features, as Assessed by the Intraclass Correlation Coefficient, Is So Valuable...

Area of Science:

• Biometrics
• Pattern Recognition
• Data Science

Background:

• Temporally persistent traits are considered more valuable in biometrics.
• The specific impact of trait persistence on biometric analysis remains unclear.
• The intraclass correlation coefficient (ICC) has been proposed as a measure of temporal persistence.

Purpose of the Study:

• To investigate how temporal persistence of features influences biometric analysis.
• To identify specific aspects of biometric studies affected by feature persistence.
• To explain the importance of trait persistence in biometric performance.

Main Methods:

• Utilized synthetic features to model the effect of temporal persistence.
• Introduced the intraclass correlation coefficient (ICC) as an index for temporal persistence.
• Analyzed real-world datasets (eye-movements, gait) to validate synthetic findings.
• Applied a decorrelation step to real datasets to account for feature intercorrelations.

Main Results:

• Temporally persistent features demonstrably affect similarity score distributions in biometric identification.
• Synthetic data analysis clearly linked feature persistence to improved biometric performance.
• Real-world data analysis largely corroborated synthetic findings, with adjustments for feature intercorrelation.

Conclusions:

• Temporal persistence is a critical factor influencing biometric performance by altering similarity score distributions.
• The study provides a mechanistic understanding of why persistent traits are crucial for biometrics.
• Methodologies using synthetic data and decorrelation techniques can effectively study feature persistence in biometrics.