Abstract: In order to preserve the local structure of the palm-dorsa vein space, locality preserving projection (LPP) was applied to palm-dorsa vein recognition. In small-sized sample cases such as image recognition, the matrix of the eigenvalue equation is usually singular. To solve the problem, kernal principal component analysis (KPCA) method was presented to reduce the palm-dorsa vein space dimensions. Then LPP was used to extract the local features. The algorithm was tested in the existing palm-dorsa vein database. The results showed that the new method has much higher recognition rate and the feature extraction time is 2.6 s, so it satisfies the real-time system specifications.