Automatic Content Classification of Digital Modulation Signals without Binary Sequence Recovery

In this paper, we present a new classification method which aims to distinguish the contents of digital modulation signals without knowing about binary sequences. The differences of these binary sequences can be detected from the features extracted from their modulation signals directly. With the help of appropriately designed classifier, we can get the classification results in high accurate rates, and avoid the complex processing steps related with recovery of these binary sequences at the same time. We also provide an example and the method to deal with it in details. In numerical simulations, the mean probabilities of correct classification for signals we point out are more than 94%. And even in noisy conditions, the method is also effective.