SMU Data Science Review
Abstract
This paper presents the development of a secure voice authentication system that delivers an inclusive solution for all users, including those with disabilities. Leveraging a Text-Dependent Active Verification process, the system combines a spoken passphrase with voice biometric coefficients and audio vector embeddings for reliable user verification. A vector database is used to efficiently store data and perform similarity retrieval. Initially, the system achieves a 71% spoof detection accuracy, ensuring that only genuine samples proceed to the embedding stage, where it attains a 55.21% accuracy in vector embedding and similarity retrieval. Furthermore, this approach paves the way for user-specific voice-controlled environments. Overall, this study underscores the transformative potential of voice biometric authentication by merging cutting-edge signal processing with sophisticated machine learning techniques, setting a benchmark for future research in balancing robust security measures, user convenience, and ethical inclusivity.
Recommended Citation
Brooks, Erica; JACOB, LIJO; Lewis, Lani; Mittal, Gaurav; Negi, Shivam; and Javed, Faizan
(2025)
"Leveraging GenAI for Biometric Voice Print Authentication,"
SMU Data Science Review: Vol. 9:
No.
1, Article 3.
Available at:
https://scholar.smu.edu/datasciencereview/vol9/iss1/3
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License
Included in
Development Studies Commons, Other Biomedical Engineering and Bioengineering Commons, Science and Technology Studies Commons
