Facial identity across the lifespan

Area of Science:

• Cognitive Science
• Computer Vision
• Psychology

Background:

• Human facial appearance undergoes significant changes throughout the lifespan.
• Recognizing familiar individuals across different ages is a common human ability.
• Understanding the mechanisms of long-term facial recognition is crucial for various applications.

Purpose of the Study:

• To investigate the statistical properties of facial aging.
• To evaluate the effectiveness of different computational models for age-independent facial recognition.
• To explore how within-person facial variability aids recognition across the lifespan.

Main Methods:

• Analysis of statistical properties of faces sampled across the adult lifespan (20s to 70s).
• Simulation of computational models based on image-similarity and familiarization procedures.
• Assessment of model performance in recognizing individuals across different ages.

Main Results:

• Individuals retain idiosyncratic facial properties supporting moderate age-independent recognition.
• Image-similarity-based models showed limited success in cross-age facial recognition.
• Models incorporating familiarization and within-person variability demonstrated robust generalization and recognition.

Conclusions:

• Unique facial properties are key to recognizing individuals over time.
• Familiarization procedures enhance computational facial recognition models' ability to generalize across ages.
• Analysis of facial variability offers insights into recognition processes and visual representation.