svm will work on only two classes or groups. You need to send only two groups data to train it. Check the below example:
load fisheriris
% the data has three classes, pick only two classes out of it.
inds = ~strcmp(species,'setosa'); % don't consider this class
X = meas(inds,3:4);
y = species(inds);
%
SVMStruct = svmtrain(X,y)
Where as if I use all the classes, it will throw the same error as you got .
SVMStruct = svmtrain(meas,species) ; % this will give error as three classes are present
Also check that svmtrain is outdated, you need to use fitcsvm.
