Need to calculate error in recognition rate using SVM classifier and rbf kernel.

I have train data with 28 features and 4000 samples. I have test data with 28 features and 800 samples. I have 62 classes as output. I loaded test and train data. I want to classify this using SVM (non linear) with 62 classes output. Please send me a program using SVM classifier and rbf kernel so that I can calculate error in recognition rate. Also send me a program using 10 fold method.