Supervised classification that performs a principal-component analysis on each class, and then an unknown is assigned to the class with which it has the lowest residual variance.