EUVIMEDEUVIMED entdecken
Crossref

MSFR; Multi-Stressor Robustness Profiling of DTW and Kinematic Features for Cross-Modal Online Signature Biometrics

Rajaa Mrayeh Mohammed

Journal of Al-Farabi for Engineering Sciences · 2026

DOI: 10.59746/9zq6t278 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Journal of Al-Farabi for Engineering Sciences, 2026. Verfasst von Rajaa Mrayeh Mohammed.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Cross-modal online signature verification is affected by acquisition-modality shift, corrective harmonization, and synthetic augmentation, yet feature robustness is usually examined under only one stressor at a time. This study develops the Multi-Stressor Feature Robustness (MSFR) framework to compare how eleven DTW-based and kinematic descriptors respond across these three stress conditions while explicitly preserving their different experimental meanings. MSFR re-organizes archived standardized effect-size summaries into a feature-level response vector comprising raw stylus–finger susceptibility, genuine and forgery post-harmonization residuals, and writer-level original-versus-augmented finger sensitivity. The framework uses balanced residuals, an exploratory residual-to-raw ratio, Holm-adjusted evidence where preserved, Pareto dominance, threshold-sensitivity analysis, and perturbation analysis of the Pareto front. Raw stylus–finger standardized differences range from 0.21998 to 1.2894. Post-harmonization residual effects are very small (mean absolute residual 0.00142 for genuine signatures and 0.00155 for forgeries). Augmentation produces a different pattern: Curvature is the most sensitive descriptor (archived |d| = 0.936), whereas six descriptors remain below the 0.20 descriptive threshold. Local Variance and Speed Mean form the first point-estimate Pareto frontier; under independent ±1% and ±5% perturbations both remain non-dominated in 100% of 100,000 simulations, while at ±10% Local Variance remains non-dominated in 100% and Speed Mean in 98.72%. Robustness is stressor-specific rather than universal. MSFR should be interpreted as an evidence-bounded pre-model diagnostic layer, not as proof of statistical equivalence, classifier improvement, or deployment readiness, because the archived material does not preserve the raw paired vectors and writer-disjoint classifier outputs needed for those claims.

Zitieren

APA

Mohammed, R. M. (2026). MSFR; Multi-Stressor Robustness Profiling of DTW and Kinematic Features for Cross-Modal Online Signature Biometrics. Journal of Al-Farabi for Engineering Sciences. https://doi.org/10.59746/9zq6t278

Vancouver

Mohammed RM. MSFR; Multi-Stressor Robustness Profiling of DTW and Kinematic Features for Cross-Modal Online Signature Biometrics. Journal of Al-Farabi for Engineering Sciences. 2026. doi:10.59746/9zq6t278

BibTeX
@article{Mohammed2026,
  author = {Mohammed, Rajaa Mrayeh},
  title = {MSFR; Multi-Stressor Robustness Profiling of DTW and Kinematic Features for Cross-Modal Online Signature Biometrics},
  journal = {Journal of Al-Farabi for Engineering Sciences},
  year = {2026},
  doi = {10.59746/9zq6t278}
}
Kostenlos starten →