On the performance evaluation of Bayesian network classifiers in modulation identification for cooperative MIMO systems

Modulation recognition plays an important role in monitoring the intercepted signals. In this paper, we present an algorithm for modulation classification designed for cooperative MIMO system based on pattern recognition approach. This is done using higher order statistics (HOS) features and a Bayesian network classifiers. In order to evaluate the effectiveness of Bayesian network methods, a comparative study is performed between the naive Bayes using discretization (NBD), the tree augmented naive Bayes (TAN) and the decision tree (J48) classifiers. Through the receiver operating characteristics (ROC) curves, the probability of identification and the training time, we show that the NBD and the TAN classifiers achieve nearly similar performance compared to the J48 classifier. Hence, these classifiers can be used to distinguish between different M-ary shift keying linear modulation types and thus lead to better monitoring of the intercepted signals in broadband technologies with low complexity.