A Multi-Instant Machine Learning Algorithm Model for Infant Fingerprinting Verification

Abstract

Understanding the challenges associated with infant biometrics, like immature ridge formation, very small fingerprints, and their fast physiological development, this research implements a multi-instance machine learning technique to enhance accuracy. Multiple instances of fingerprints for each infant were recorded, which enhanced the robustness of the system by dealing with inter-class variability. The datasets used 500 pixels-per-inch fingerprints of infants less than 12 months old that were recorded during two different instances. After preprocessing techniques, feature extraction was performed using a convolutional neural network (CNN), and matching scores were computed. The resulting scores were analyzed using a combination of three learning algorithms, CNN, CNN + SVM, and CNN + RF. Metrics showed that CNN + RF model outperformed all others with regard to its performance, having achieved verification accuracy of 81% compared to 75% and 82%, respectively (the previous multi-instance learning technique had achieved highest accuracy of only 69.05%), training accuracy of 98% compared to 58% and 50%, respectively, validation accuracy of 85% compared to 84% and 83%, respectively, and minimum training loss of 0.2 while having considerably high validation loss of 2.5. Although higher computational cost, CNN + RF proved most active for real-world applications in healthcare and identity management. Upcoming research should explore larger datasets to further enhance model precision. 

Received: 21 August 2025 | Revised: 26 January 2026 | Accepted: 22 June 2026

Conflicts of Interest

The authors declare that they have no conflicts of interest to this work.

Data Availability Statement

Data available on request from the corresponding author upon reasonable request.

Author Contribution Statement

Olatayo Moses Olaniyan: Conceptualization, Methodology, Software, Validation, Investigation, Resources, Data curation, Writing – original draft, Visualization, Supervision, Project administration. Kazeem Aderemi Bello: Conceptualization, Methodology, Investigation, Resources, Data Curation, Writing – original draft. Rendani Wilson Maladzhi: Conceptualization, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing. Johnson Olugbenga Joseph: Conceptualization, Investigation, Resources, Data curation, Writing – original draft. Adebimpe Esan: Methodology, Software, Validation, Formal analysis, Writing – review & editing, Visualization, Supervision, Project administration. Quadri Ademola Mumuni: Conceptualization, Investigation, Resources, Data Curation, Writing – original draft. Bolaji Abigael Omodunbi: Formal analysis, Writing – review & editing. Chinedu Michael Chiejine: Software, Project administration.