Combining local and global features based on the eigenspace for vein recognition

Abstract

Dorsal hand vein recognition is an emerging biometric technique researched today. In this paper, we propose a novel approach, the local feature-based ensemble 2-directional 2-dimensional linear discriminant analysis (LFBE(2D) 2 LDA), for dorsal hand vein recognition. The characteristic of the approach is to combine local and global information for vein recognition. First, we use block-based (2D) 2 PCA (B(2D) 2 PCA) to extract local feature from the dorsal hand vein image. Then, the global features are extracted by (2D) 2 LDA from the local feature-based ensemble to represent the vein image for classification. This method not only combines local and global feature, but also takes full advantage of the discriminant information and descriptive information of the images. The experiment result on our large dorsal hand vein database shows that high accuracies (98.55%) have been obtained by our proposed method. Keywords-dorsal hand vein recognition; block-based 2-directional 2-dimensional principal component analysis (B(2D)2PCA); 2-directional 2-dimensional linear discriminant analysis ((2D) 2LDA)