A Novel SVM-based Automatic Modulation Classifier

Automatic modulation classification (AMC) recreates an essential part in both cooperative and noncooperative communication applications. AMC is employed to automatically determine the modulation types of the transmitted signals. This is done by noting the recived information instances, which would be manipulated by the noise as well as fading channels. It is a middle process between the signal detection and the data retrieving. In this paper, a novel feature-based ACM is proposed and its performance is investigated. The ACM is based on the support vector machine (SVM) method. Different analog and digital modulation types are examined versus the ACM. The classifier accuracy is more than 98% for high SNR. Even more, the accuracy is around 90% for low SNR. The confusion matrix show no chaos between the different modulation types discussed.