Hierarchical approach of discriminative common vectors for bio metric security

In face recognition task the Linear Discriminant Analysis method attempts to find projection directions that minimize within class scatter and maximize between class scatter. But since the dimension of the sample space is typically larger than the number of samples in the training set the within class scatter matrix is singular. Therefore, LDA method cannot apply directly. This is called as the ?small sample size problem?. This problem is solved by discriminative common vectors approach. The discriminative common vectors approach is base on a variation of Fisher´s Linear Discriminant Analysis. In this, the common vectors are extracting by eliminating the differences of the image samples in each class of images. Then the discriminative common vectors, which will be use for classification, are obtaining from the common vectors. This method classifies only known persons present in the database. This limitation can be solving by our proposed method. In this paper, we propose a new face recognition method called the hierarchical classification method based on a variation of discriminative common vector method. Hierarchical classification allows the system to make best use of information available in a database allowing the system to even classify individuals not present in the database to obtain collateral information like age group, race, etc., this is made possible by obtaining the features of individuals in one level then obtaining the features of groups in the higher levels. When classification fails at lower level then classification is attempt at a higher level. The number of iterations required is equal to the number of hierarchy levels in the training set. The proposed method yields an optimal solution for maximizing the Discriminant Common Vector method. Our test results show that the hierarchical classifier system is superior to other methods in terms of recognition, accuracy and efficiency.