Feature-level fusion of global and local features for finger-vein recognition

Multi-features system, as an effective method to improve the performance of biometric-based identification, has been one of the hot research fields on personal identification. In this paper, a novel method of finger-vein recognition based on the feature level fusion of global and local features is proposed. First, local texture information is characterized as the local feature using a Gabor filter framework, and the global feature is extracted by moment invariant method. Then, global-local feature vectors (GLFVs) from finger-veins are generated using canonical correlation analysis (CCA) and a novel weighted fusion strategy. Based on GLFVs, the nearest neighborhood classifier is employed for classification finally. Experimental results show that the proposed method has good performance in personal identification.