A palmprint classification method based on finite ridgelet transformation and SVM

Abstract

A multi-scale palm print classification method based on FRIT (finite ridgelet transform) and SVM (support vector machine) was proposed. First, palm image with preprocessing was decomposed by using FRIT. As a result, ridgelet coefficients in different scales and various angels were obtained. The important linear feature of palmprint was included in the low frequency coefficients of FRIT decomposition coefficients. After the decomposition coefficients were transformed into feature vectors, SVM was chosen as a classifier. These feature vectors were regarded as feature parameters of palm print and sent into SVM to training. Two kernel functions are used as a discriminant function. Finally, SVM trained was used for classification of palmprint. The experiments were performed in PolyU Palmprint Database. The results indicate that proposed method has better performance than wavelet with SVM method, and classification accuracy used RBF (radial basis function) as kernel is higher than the use of polynomial kernel function. © 2011 Springer-Verlag.