Intelligent Linear and Nonlinear Analysis for Biometric Fingerprint Recognition

Linear and nonlinear component analysis are proposed for fingerprint recognition in this study. Principal component analysis (PCA) has been used for dimension reduction so that the significant characteristics of various samples are indicated by the dominant eigenvectors of the corresponding covariance matrices. Then a set of principal components are selected for cluster separation. It is a multivariate analysis technique to achieve a fundamental idea of reducing correlated variables within a dataset while leaving most uncorrelated variables to represent the entire dataset. PCA gives rise to satisfactory results with reduced dimension, on the other hand, however, this linear approach can not fully indicate nonlinear phenomenon. In this case, nonlinear component analysis (NCA) is also developed as a nonlinear approach for iteratively reconstruction in the data space for the data in the feature space. Compared with PCA, NCA can extract more useful features for purposes of classification. By means of principal component analysis and nonlinear component analysis as well as control oriented identification, biometric characteristics can be captured.