mfcc with classifiers (GMM and SVM)

for an audio , i am interestion to know that which mfcc extracted features should focus in order to check perfomance of classifiers like GMM and SVM...

i am now new to this field, please tell me from basics