Performance of the FearID earprint identification system

Area of Science:

• Forensic Science
• Biometrics
• Pattern Recognition

Background:

• Earprints are valuable forensic evidence found at crime scenes.
• Automated analysis of earprints can enhance identification accuracy and efficiency.
• Previous methods lacked robust automated classification capabilities.

Purpose of the Study:

• To assess the evidential strength of earprints in forensic investigations.
• To develop and evaluate an automated system for earprint classification.
• To determine the accuracy of earprint matching algorithms.

Main Methods:

• Collected a dataset of 1229 donors' earprints across three countries.
• Utilized operator-defined contours and specialist-defined anatomical landmarks for image segmentation.
• Developed and trained an automated classification system for 'matching' or 'non-matching' print pairs.

Main Results:

• Achieved a 4% equal error rate for laboratory-quality earprint comparisons.
• For reference databases with two prints per ear, 90% of searches yielded the best match within the top 0.1% of the hitlist.
• Print/mark comparisons resulted in a less favorable equal error rate of 9%.

Conclusions:

• The automated earprint identification system demonstrates significant potential for forensic applications.
• The system shows high accuracy in matching lab-quality prints, supporting its use in forensic casework.
• Further research is needed to optimize performance for real-world, degraded earprint evidence.