Unsupervised characterization of digital modulations

With the rising number of modulation types used in multi-user communication systems, we need to find efficient methods to discriminate them. Indeed, modulation recognition has become important in wireless communications for both civilian and military applications. Traditionally, to classify modulation types, most studies assume abundant a priori knowledge about modulated signals such as binary data rate, baud rate or carrier frequency. However, new transmission/receiver systems need to classify automatically digital modulations without specific information. In the course of making decision on modulation type, these parameters may have to be estimated in order to perform efficient demodulation. We propose a new way of estimating the baud rate of digitally modulated signals and then some perspectives to estimate the binary data rate. This approach is based on abrupt changes detection in the time-frequency plane.