Automatic modulation classification (AMC) has been used as a crucial component to improve the overall performance in cognitive radio system, thereby allowing any user to promote secure and reliable communication. In this letter, we propose a semi-blind AMC approach based on a simplified distributed space-time block coding (D-STBC) scheme, in which the STBC code is artificially generated by availing three nodes in the cooperative network. The proposed system aims at distinguishing between the modulation type and order among different M -ary shift keying linear modulations, by exploiting both the temporal and spatial signal dimensions. After the application of the zero-forcing STBC-decoding, the performance of the proposal is investigated using three different machine learning classifiers. For this, a comparative study has been carried out by resorting to selected features which were based on the combination of the fourth and the sixth-order of high-order cumulants and high-order moments of the D-STBC-decoded signal. In addition, to the usually adopted receiver operating characteristics classification metric, we propose to have recourse to the accuracy, the recall, the precision, the specificity, and the F-factor, to ensure the impartiality of the comparison. Support-vector machine classifier is shown to exhibit the highest performance in terms of the whole measures.