EUVIMEDEUVIMED entdecken
Crossref

FGCM-SViT: Fidelity-Guided Cross-Modal Siamese Vision Transformer for Writer-Independent Online Signature Verification

Hayder Majid Sachit Al-Rikabi

Journal of Al-Farabi for Engineering Sciences · 2025

DOI: 10.59746/bh3c7n57 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Al-Farabi for Engineering Sciences, 2025. Verfasst von Hayder Majid Sachit Al-Rikabi.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Writer-independent online signature verification (OSV) must handle large within-writer variation, skilled forgery, limited enrollment data, and differences between stylus and finger input. Existing Siamese CNNs learn metric embeddings, while Transformer models improve global dependency modeling. However, most methods treat synthetic samples as equally reliable and rarely connect augmentation fidelity with cross-modality learning. This study proposes FGCM-SViT, a Fidelity-Guided Cross-Modal Siamese Vision Transformer. It combines stylus-guided harmonization, statistically weighted augmentation, a three-channel dynamic signature representation, and a shared-weight Vision Transformer. The design is motivated by prior analysis of a DeepSignDB-derived cohort of 76 writers and 1,216 genuine finger signatures. After harmonization, stylus-versus-processed-finger effects were almost zero for genuine signatures (mean |d| = 0.00142) and forgery conditions (mean |d| = 0.00155). In contrast, curvature showed a large writer-level augmentation shift (d = -0.936). FGCM-SViT therefore assigns a feature-sensitive fidelity score to each augmented sample and uses it during pairwise metric learning. The joint objective combines contrastive discrimination, cross-modal alignment, and fidelity regularization. A writer-disjoint five-fold protocol, four verification protocols, ablation tests, bootstrap confidence intervals, and multiplicity-aware comparisons are prespecified. No unexecuted EER, FAR, FRR, or ROC-AUC values are reported.

Zitieren

APA

Al-Rikabi, H. M. S. (2025). FGCM-SViT: Fidelity-Guided Cross-Modal Siamese Vision Transformer for Writer-Independent Online Signature Verification. Journal of Al-Farabi for Engineering Sciences. https://doi.org/10.59746/bh3c7n57

Vancouver

Al-Rikabi HMS. FGCM-SViT: Fidelity-Guided Cross-Modal Siamese Vision Transformer for Writer-Independent Online Signature Verification. Journal of Al-Farabi for Engineering Sciences. 2025. doi:10.59746/bh3c7n57

BibTeX
@article{AlRikabi2025,
  author = {Al-Rikabi, Hayder Majid Sachit},
  title = {FGCM-SViT: Fidelity-Guided Cross-Modal Siamese Vision Transformer for Writer-Independent Online Signature Verification},
  journal = {Journal of Al-Farabi for Engineering Sciences},
  year = {2025},
  doi = {10.59746/bh3c7n57}
}
Kostenlos starten →