Score level fusion for multimodal biometric identification

Abstract

It is a fact that biometric systems for identification have many advantages. More than one biometric characteristic are used in a multimodal biometric system to improve the accuracy of identification. Out of the several physical biometrics, palmprints and palm veins have many biometric features and they vary from person to person. They can be acquired in a touchless acquisition setup. A biometric system has image processing stages such as pre-processing, extracting ROI, feature extraction and feature matching. This paper presents an algorithm for segmenting Region of Interest from central palm region with respect to correct alignment of palm in vertical direction with fingers pointing upwards. Texture and appearance based features are extracted to form a feature vector. Weighted Euclidean distance is used for feature matching. The algorithm for fusion of palmprint and palm vein images at score level is investigated. Scores are determined using similarity scores and normalized using Min-Max normalization. Fusion is implemented using weighted sum method. The effect of fusion on identification is studied and evaluated with biometric parameters GAR and EER. The results obtained are analyzed.