A Biometric Identification System with Kernel SVM and Feature-level Fusion

Abstract

The paper presents a biometric system with optimization for identification. The design combines 2 biometrics (fingerprint and palmprint) with feature-level functional fusion, avoiding the concatenation. Data classification is done with a kernel SVM (Support Vector Machine) model and a multi-class extension. The experimental achievements show that the performance improvements are provided by the feature-level fusion together with an optimized design of the biometric data classifier. The model can be applied in use-cases in which the identity of the individuals should be guessed only based on the biometric credential.