The MT-Transform has the advantages of having a uniform amplitude bounds, speed and simple hardware realization. Thus when conventional classifiers are used with MT-feature vector, the classification process represents a computational load. In addition to the fact that these conventual classifiers suffer from many other drawbacks. In this paper, we have used in MT-feature vector with the decision tree classifier (DTC) . Thus the recognition is invariant and the process has much speed and simplicity. This is a combined approach, i.e, the process combines feature extraction, feature selection, and classification. In addition, experimental results show that a higher classification rate has been achieved with less than 20% from the feature vector