Finger vein recognition with manifold learning

Finger vein is a promising biometric pattern for personal identification in terms of its security and convenience. However, so residual information, such as shade produced by various thicknesses of the finger muscles, bones, and tissue networks surrounding the vein, are also captured in the infrared images of finger vein. Meanwhile, the pose variation of the finger may also cause failure to recognition. In this paper, for the first time, we address this problem by unifying manifold learning and point manifold distance concept. The experiments based on the TED-FV database demonstrate that the proposed algorithmic framework is robust and effective.