Multimodal Personal Verification Using Likelihood Ratio for the Match Score Fusion

Area of Science:

• Computer Science
• Biometrics
• Pattern Recognition

Background:

• Multimodal biometric systems offer enhanced security and accuracy compared to unimodal systems.
• Effective fusion of scores from multiple biometric traits is crucial for robust personal verification.
• Existing methods like Support Vector Machines (SVM) and sum rule have limitations in handling complex biometric data.

Purpose of the Study:

• To develop and evaluate a novel personal verification system integrating multiple biometric modalities.
• To improve the accuracy and reliability of person identification using a likelihood ratio test for score fusion.
• To compare the proposed system's performance against established approaches like SVM and sum rule.

Main Methods:

• Feature extraction using Zernike Moments (ZM) from face, fingerprint, hand shape, and palm print data.
• Score-level fusion of multiple biometric matchers employing the likelihood ratio test.
• Estimation of genuine and impostor score densities using a Gaussian Mixture Model (GMM).

Main Results:

• The proposed system demonstrated superior identification performance compared to SVM and sum rule methods.
• Experimental results confirmed higher accuracy in personal verification using the novel fusion technique.
• The system effectively handles score fusion from diverse biometric traits.

Conclusions:

• The developed multimodal biometric verification system achieves excellent identification accuracy.
• The likelihood ratio test-based fusion approach provides a robust method for personal verification.
• The system's high performance makes it suitable for a wide range of person identification applications.