Fusion of Palm Dorsal Vein and Hand Geometry for Personal Identification Based on Linear Discriminant Analysis

A multi-modal biometric system based on palm dorsal vein and hand shape is proposed in this paper. Palm dorsal vein images are captured with an infra-red sensitive camera under near infra-red illumination. Gaussian matched filter is performed to extract vein patterns from Regions Of Interest (ROI) cropped from original images. As for hand shape features, width, height and triangle features are extracted from boundary images of palm dorsal. For the matching score level information fusion, the Linear Discriminant Analysis (LDA) algorithm is employed to project 2-dimensional matching score vectors to 1-dimensional space. Experimental results demonstrate that the fusion process can raise the system precision substantially, although it's relatively low using either single trait.