Based on the statistical learning theory

kernels are often presented as measures of similarity measure.Kernels play a very important role in support vector machine (SVM) algorithms

which can mapper the un-classification data to high dimensional space

to get the optimal classification result.Unfortunately

SVM lacks ability to deal with the system or data with ambiguous characters.So this paper proposes to SVM with fuzzy kernel

which is obtained by similarity measure method.Compared to Gaussian RBF kernel function and fuzzy sigmoid kernel function

SVM with fuzzy similarity kernel gets more accuracy results with less computation requirements

and alleviates current limitation of the fuzzy kernel function.