Dear pranjal,
First of all I thik that you will need more of theese samples (3 is not enough for the classifier to be exact). All of them should be in one matrix, where you have got different samples in the rows and the colums of the matrix are your attributes (for example frequency). Then you want to use it like this:
knn = ClassificationKNN.fit(X, d, 'NumNeighbors', k);
[LABEL,POSTERIOR,COST]=predict(knn, Xt);
where X is your samples matrix, d is a column vector of decisions of the classifier (e.g. 0 - healthy, 1 - apnea and so on) and k is a number of neighbors. Predict function will calculate the predictions, they are stored in LABEL. Xt is the matrix of samples that you want to obtain predicts for.
Good luck!
