For Setting 1, seven feature sets as described in Table I are used to find the most appropriate set for the detection system. All classes (C1-C16) are taken into account as defined in Table II. Table III shows the overall accuracy of the correctly classified instances of the Setting 1 among five classifiers. The highest average accuracy is achieved when all the features (Feature Set 1) is used. The accuracy rankings among the five classifiers are 1-nearest neighbor (highest), 3-nearest neighbor multilayer perceptron, J48 decision tree, and BayesNet (lowest), respectively. With the use of only 7 in Feature Set 2, an equivalent accuracy can be achieved across all the classifiers.