Abstract: During dorsal vein recognition processing,for the problem that there may be some distractive information in acquired image,a recognition algorithm based on effective dimensional feature was proposed.Firstly,adaptive median filtering algorithm is applied in acquired image for denoising.Secondly,image is divided into blocks,and every sub image feature is extracted based on Gaussian mixture model and gradient information.Then,according to feature similarity between sub images,the method which judges whether the sub image includes distractive information or not is proposed.Finally,vein image feature vector can be acquired through fusing the characteristics of all sub images with real vein information,and sparse representation algorithm is proposed to match the feature vector of unknown sample with diffe-rent effective dimensions.Experiments show that high recognition rate can be achieved by the proposed algorithm,and even though part of the region in vein image is blocked,good recognition effect can be also acquired.

Key words: Vein recognition,Gaussian mixture model,Sparse representation,Effective dimension