EUVIMEDEUVIMED entdecken
Crossref

Development of an Optimized Lightweight CNN Model Using Chicken Swarm Optimization for Finger-Vein Biometric Recognition

O. G. Adedara, A.O. Oke, O.A. Oyinloye, O. Makinde, F.W. Ipeayeda, A. S. Falohun

Advances in Multidisciplinary & Scientific Research Journal Publication · 30.06.2026 · 14 (2) · 9-29

DOI: 10.22624/aims/maths/v14n2p2 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Advances in Multidisciplinary & Scientific Research Journal Publication, online erschienen am 02.06.2026. Verfasst von O. G. Adedara und 5 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Finger-vein recognition has gained prominence in biometric security due to its internal vascular patterns, which offer high uniqueness and resistance to forgery. Nevertheless, optimizing performance for the intricate, high-dimensional features of finger vein patterns remains challenging. This study introduces a finger vein verification and identification system that combines lightweight convolutional neural networks (CNNs) with Chicken Swarm Optimization (CSO) for hyperparameter tuning, referred to as the CSO-CNN technique. Evaluated on a dataset of 400 images (150 genuine, 250 impostor), images underwent preprocessing for enhancement and region of interest extraction. The CSO-CNN optimized hyperparameters to derive discriminative features, classified using SoftMax classifier. At an optimal threshold of 0.8, CSO-CNN yielded a False Acceptance Rate (FAR) of 7.20%, False Rejection Rate (FRR) of 14.00%, accuracy of 90.25%, and recognition time of 125.52 seconds. Comparatively, the standard CNN achieved FAR of 8.80%, FRR of 17.33%, accuracy of 88.00%, and recognition time of 175.53 seconds. The Equal Error Rate (EER) further validates CSO-CNN's superiority at 12.00% versus 15.45% for CNN. Confusion matrix analysis indicated CSO-CNN correctly identified 129 genuine and rejected 232 impostors, with fewer misclassifications. Paired t-tests confirmed significant enhancements, with p-values of 0.005 and 0.000. CSO's integration facilitated efficient hyperparameter tuning, boosting feature extraction and reducing recognition time. The finger vein's internal, stable vascular patterns contributed to robust accuracy. These outcomes establish CSO-CNN as an effective solution for finger-vein biometrics, enhancing security and efficiency for real-world applications. Keywords: Finger vein recognition, vascular biometrics, lightweight convolutional neural networks, hyperparameter optimization, Chicken Swarm Optimization, deep learning Adedara O. G., Oke, A.O., Oyinloye O.A., Makinde, O., Ipeayeda, F.W., & Falohun A. O. (2026): Development of an Optimized Lightweight CNN Model Using Chicken Swarm Optimization for Finger-Vein Biometric Recognition. Journal of Advances in Mathematical & Computational Science. Vol. 14, No. 2. Pp 9-28. Available online at www.isteams.net/mathematics-computationaljournal. dx.doi.org/10.22624/AIMS/MATHS/V14N2P2

Zitieren

APA

Adedara, O. G., Oke, A., Oyinloye, O., Makinde, O., Ipeayeda, F., & Falohun, A. S. (2026). Development of an Optimized Lightweight CNN Model Using Chicken Swarm Optimization for Finger-Vein Biometric Recognition. Advances in Multidisciplinary & Scientific Research Journal Publication, 14(2), 9-29. https://doi.org/10.22624/aims/maths/v14n2p2

Vancouver

Adedara OG, Oke A, Oyinloye O, Makinde O, Ipeayeda F, Falohun AS. Development of an Optimized Lightweight CNN Model Using Chicken Swarm Optimization for Finger-Vein Biometric Recognition. Advances in Multidisciplinary & Scientific Research Journal Publication. 2026;14(2):9-29. doi:10.22624/aims/maths/v14n2p2

BibTeX
@article{Adedara2026,
  author = {Adedara, O. G. and Oke, A.O. and Oyinloye, O.A. and Makinde, O. and Ipeayeda, F.W. and Falohun, A. S.},
  title = {Development of an Optimized Lightweight CNN Model Using Chicken Swarm Optimization for Finger-Vein Biometric Recognition},
  journal = {Advances in Multidisciplinary & Scientific Research Journal Publication},
  year = {2026},
  volume = {14},
  number = {2},
  pages = {9--29},
  doi = {10.22624/aims/maths/v14n2p2}
}
Kostenlos starten →