Large-scale evaluation of multimodal biometric authentication using state-of-the-art systems.

Area of Science:

• Biometrics and Authentication Systems
• Human-Computer Interaction
• Pattern Recognition

Background:

• Prior multimodal biometric studies often used non-COTS systems with limited accuracy.
• Previous research was typically conducted on smaller populations (hundreds of users).

Purpose of the Study:

• To evaluate multimodal biometric system performance using COTS fingerprint and face recognition.
• To assess accuracy improvements on a large-scale population (approx. 1,000 individuals).
• To introduce novel normalization and fusion techniques for enhanced accuracy.

Main Methods:

• Utilized state-of-the-art Commercial Off-the-Shelf (COTS) fingerprint and face biometric systems.
• Tested systems on a population of nearly 1,000 individuals.
• Implemented and evaluated established multimodal fusion methods alongside new normalization techniques.

Main Results:

• Demonstrated significant accuracy gains with multimodal systems compared to unimodal systems.
• Achieved higher accuracy even with highly accurate COTS systems.
• New normalization and fusion methods further boosted system performance.

Conclusions:

• Multimodal biometric authentication, particularly fingerprint and face, offers substantial accuracy improvements.
• COTS systems are viable for large-scale, accurate multimodal biometric deployments.
• Novel fusion and normalization strategies enhance multimodal biometric system effectiveness.