Digital Modulation Classification based on Higher-order Moments and Characteristic Function

In this paper, a method based on moments and characteristic function is proposed for the classification of digital modulation signals. The method is robust against the random phase of the signals. The features used in this method are the modulus of moments and the number of peaks within modified characteristic function. The former is particularly effective for classifying signals into subclasses at low SNR, while the latter is used for classifying signals within subclasses. To remove the effect of magnitude of received signals on modulation classification, the received signals are normalized by power. A signal power estimation algorithm is proposed accordingly.