Next-Generation Biometric Authentication: Overcoming the Twin Identification Challenge with Advanced Facial Recognition and Multi-Modal Analysis Techniques

Conventional facial recognition struggles with individuals sharing near-identical facial features, particularly monozygotic twins. This research introduces a robust, real-time methodology to overcome this by integrating geometric, textural, and dynamic facial characteristics. The framework employs M...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Journal of Al-Qadisiyah for Computer Science and Mathematics Jg. 17; H. 3
Hauptverfasser: Hasan Nsaif, Azhar, Abduladheem Hasan, Rawsam
Format: Journal Article
Sprache:Englisch
Veröffentlicht: 30.09.2025
ISSN:2074-0204, 2521-3504
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:Conventional facial recognition struggles with individuals sharing near-identical facial features, particularly monozygotic twins. This research introduces a robust, real-time methodology to overcome this by integrating geometric, textural, and dynamic facial characteristics. The framework employs Multi-Task Cascaded Convolutional Networks (MTCNN) for face detection and alignment, followed by FaceNet for 128-dimensional facial embedding generation. MediaPipe's 468-point facial landmark extraction quantifies subtle structural variations via transformation matrix analysis and blend-shape evaluation, capturing static geometric discrepancies and dynamic micro-expressions. Validated on 7,200-image dataset (70% training, 30% testing), the system achieved 97.73% accuracy, operating efficiently on consumer-grade GPUs. This approach significantly enhances biometric technology, offering improved identity verification for genetically similar individuals in critical security applications like border control and secure access management, thereby addressing a key limitation in current facial recognition systems.
ISSN:2074-0204
2521-3504
DOI:10.29304/jqcsm.2025.17.32383