Research and application of combined kernel SVM in dynamic voiceprint password authentication system

Abstract

The kernel function plays an important role in the classification of support vector machines (SVM). In order to solve the problem that a single SVM kernel function can not achieve optimal learning ability and generalization ability in recognition classification at the same time, here we present a new combined kernel function by analyzing and comparing the characteristics of various kernel functions. The new combined kernel function, which is a weighted fusion of radial basis function and polynomial kernel function, has the advantages of both global kernel function and local kernel function. Moreover, we compare the evaluation results of the combined kernel SVM and the SVM using traditional kernels in dynamic voiceprint password authentication system (DVPAS). The experimental results show that the newly constructed multicore SVM classifier obviously superior to linear kernel function, polynomial kernel function and radial basis function in DVPAS.