Digital modulation identification is a challenging and important operation in the security communication field. Identifying properly a digital modulation and its parameters is a key operation in many applications, such as cognitive radio, communication intelligence and dynamic spectrum allocation. However, traditional methods face a stalemate in which the accuracy of the identification process is quite low. Therefore, new descriptors are needed as an alternative solution or to complement the existing ones. Using a new method from the field of nonlinear dynamic systems, namely the phase diagram representation, we want to extract some features to help recognizing the type of used modulations. This method can be used even in noisy backgrounds, with the help of an additional processing algorithm, to give important information about the signal and the used modulation.