Face-voice based multimodal biometric authentication system via FaceNet and GMM

Area of Science:

• Computer Science
• Information Security
• Biometrics

Background:

• Information security is critical in IT due to industry advancements.
• Biometric authentication, using unique physiological and behavioral traits, is essential for secure user identification.
• Current systems need reliable personal recognition to restrict access to authorized users.

Purpose of the Study:

• To enhance the accuracy of multimodal biometric authentication by combining voice and face recognition.
• To reduce the equal error rate (EER) in biometric systems.
• To present a novel scheme for robust user identification.

Main Methods:

• Utilized the Gaussian Mixture Model (GMM) for voice recognition.
• Employed the FaceNet model for face recognition.
• Implemented score-level fusion to integrate voice and face modalities for final identity determination.

Main Results:

• The proposed multimodal biometric system achieved enhanced accuracy.
• The implemented score-level fusion effectively combined voice and face recognition results.
• The system demonstrated a significantly lower equal error rate compared to existing methods.

Conclusions:

• Multimodal biometric authentication combining voice and face recognition offers superior security.
• The proposed GMM and FaceNet-based scheme with score-level fusion is effective in reducing EER.
• This approach provides a more reliable and accurate method for user authentication.