How to scale test data using svm.cpp class file (LibSvm)

See more:

C++

OpenCV

Hi, Iam doing an project based on opencv and SVM (support vector machines). In there in order to identify user shown gesture i need to do the classification using SVM. I have added svm.cpp and svm.h files to my project and below is my code for the prediction part.

svm_node x[featureVectorSize];

// inside a loop
for(int k;0 < featureVectorSize; k++)
{
    x[k].index = k+1;
    x[k].value = FeatureVectorVal[k]; // i have a vector to store feature values for test data    //(without scaling)
}

    svm_model* newmodel = svm_load_model("Mymodels\\A.scale.model");
    double *ProbVal = new double[1];
    double result = svm_predict_probability(newmodel,x,ProbVal);
    cout<<"result "<< result<<"\n";
    cout<<"ProbVal "<< *ProbVal<<"\n";
    svm_free_model_content(newmodel); 
    svm_free_and_destroy_model(&newmodel);

//some code to show output

Q1. when i do the classification with model A it always gives matching result value as -1. That means not matching with the test data. After that i have hardcoded the scaled feature values for test data (write above FeatureVectorVal to a text file and scale using scale.exe which is provided by the Libsvm package). Then i did the prediction with the same model.Now it gave me result as +1. So that means i have to input scaled test data to svm_node structure. But in svm.cpp i couldn't find any function for scaling my test data. How can i scale my test data using svm.cpp functions?

Q2. When i do prediction with more than one model (using a array of model file paths) program crashes after prediction with first model and prompt me to ;

if ((model->param.svm_type == C_SVC || model->param.svm_type == NU_SVC) &&
	    model->probA!=NULL && model->probB!=NULL)

line in svm.cpp file.

Need help guys..Thanks in advance :-)