A Hierarchical Data Fusion and Classification Model for Biometric Identification Systems

Autor/autori: Sorin SOVIANY, Mariana JURIAN

Rezumat: Articolul prezinta un model ierarhic de clasificare a datelor biometrice, destinat a asigura imbunatatirea performantelor procesului de recunoastere a persoanelor folosind sisteme de identificare biometrica. Abordarea ierarhica se bazeaza pe combinarea mai multor clasificatori in cadrul unei ierarhii cu fuziune biometrica multi-nivel. Fuziunea biometrica multi-nivel include atat fuziunea pre-clasificare cu selectia optima a caracteristicilor, cat si fuziunea postclasificare cu aplicarea unei reguli de insumare ponderata a scorurilor de similaritate. Solutia asigura cresterea preciziei procesului de recunoastere biometrica inclusiv prin aplicarea unei strategii adecvate de selectie a caracteristicilor, deoarece nu toate componentele vectorilor de caracteristici asigura acelasi grad de imbunatatirea performantei.

Cuvinte cheie: fuziune biometrica multi-nivel, model ierarhic, eroare de generalizare

Abstract: The paper presents a hierarchical biometric data classification model which is designed to provide the performance enhancement for the persons recognition task in biometric identification systems. The hierarchical approach is relying on more classifiers combination within a multi-level biometric fusion hierarchy. The multi-level biometric fusion model includes both of pre-classification fusion with optimal feature selection and the postclassification fusion based on the similarity scores weighted sum. The proposed solution increases biometric recognition accuracy based on a suitable feature selection, as much as not all of the feature vectors components support the performance improvement degree.

Keywords: multi-level biometric fusion, hierarchical model, generalization error

DOWNLOAD PDF