A Mahalanobis distance-based method for palmprint biometric identification

Autor/autori: Sorin SOVIANY, Mariana JURIAN

Rezumat: Articolul prezinta o solutie de identificare biometrica a persoanelor, in care se utilizeaza caracteristici extrase din regiuni de interes ale palmei. Metoda propusa utilizeaza distanta Mahalanobis pentru evaluarea similaritatii paternurilor biometrice, iar stabilirea identitatii persoanelor se realizeaza prin maximizarea scorurilor de similaritate evaluate. Utilizarea distantei Mahalanobis in evaluarea similaritatii paternurilor biometrice prezinta avantaje semnificative comparativ cu alte tipuri de distante, in special prin proprietatile sale de invarianta la scalare si de exploatare a corelatiilor dintre caracteristici. Metoda propusa imbunatateste precizia sistemelor biometrice de identificare.

Cuvinte cheie: distanta Mahalanobis, identificare, regiuni de interes

Abstract: The paper presents a people biometric identification solution which are using features from the palmprint regions of interest (ROI). The proposed method uses the Mahalanobis distance in order to evaluate the biometric patterns similarity; the people identification is resulting through the matching scores maximization. The Mahalanobis distance using for the biometric patterns matching is suitable because of its main properties providing significant advantages for biometric data, especially the scaling invariance and feature correlation exploiting. The proposed method improves the identification biometric systems accuracy.

Keywords: Mahalanobis distance, identification, region of interests

DOWNLOAD PDF