K-fold validation; speech recognition

Hello everyone! i need some help with K-fold validation. i have three data set("EL", "MK", "Z0") each of which is a 1X1 matrix of length 20 data point. out of the 20 data, the first 10 data is silence audio and the last 10 data is speech audio. my required task is to perform K-fold validation with K=10. i'm further advised to split the data set into training and test sets keeping the training set balanced, with half "Speech" and half "Silence" audio samples. how can i approach this? any help will be deeply appreciated. thanks