Study On Algorithms Of Palmprint Recognition

With the fast development of computer and internet, people,entertainment and economic activities have close relationship with computer and internet, and the whole society run more and more depending on network. The security mechanisms of confidential information protecting need to be strengthen continuously. Traditional security and authentication cannot guarantee the safety of authentication. However, biometrics based on personal biologic feature can solve the authentication problems efficiently and safely, which is concerned widely.The palmprint recognition is a new biometrics technology, which use palm texture as biologic characters to recognize identity. As palmprint is typical of ease acquisition, and its major feature is apparent, stability and divisibility, palmprint automatic identification is a very potential identification technology.Extensive foundations of theory and application have been made in the field of palmprint recognition over ten years'research. Especially for the preprocessing and segmentation stage, the algorithms are relatively ample and mature. The two- dimensional locality preserving projections (2DLPP) is used for extracting feature. The Minimax Probability Machine(MPM) is used for data classification. Both of the two methods are the main object in this paper for studying.The main work in this paper is as follows:①The two-dimensional locality preserving projections (2DLPP)-based on manifold learning which is used for extracting feature would be reconstructed as incremental learning algorithm called GSVD-I2DLPP . From the theory of two dimensional locality preserving projections, the objective function of 2DLPP equates with the generalized eigenvalue equation. There are several methods for computing the generalized eigenvalue equation. For instance, the method based on the orthogonal-triangular (QR) decomposition and the method based on the generalized singular value decomposition (GSVD). In this paper, it uses the method based on the generalized singular value decomposition (GSVD) to solve the problem of generalized eigenvalue equation in the 2DLPP. At the same time, the incremental 2DLPP based on the generalized singular value decomposition (GSVD) is proposed. The experimental results show that the proposed GSVD-I2DLPP gives the same classification performance as the 2DLPP algorithm but costs less time for computing than the 2DLPP algorithm.②After studying the Minimax Probability Machine (MPM) which is used for binary classification problem, the disadvantage of the MPM is that the MPM take the importance of sample into account but neglect the relative importance of each feature with respect to the classification task. To overcome this limitation, we construct a Feature Weighted MPM(FWMPM) based on the Boosting . First, we estimate the relative importance of each feature by computing the Boosted distance .It is also the weight of each feature. Then, we make use of the weights for computing the inner product and Euclidean distance in kernel functions. In this way the MPM can allay being affected by trivial relevant feature. The experimental results and analysis show that the FWMPM has the better performance of class than the standard MPM.