Abstract
The automatic modulation identification of a detected signal represents an essential task for an intelligent receiver and plays an important role in demodulating the intercepted signals for several communication systems. In this study, the authors propose an efficient algorithm of superposed modulations identification dedicated for two-way relaying multiple-input multiple-output systems with physical-layer network coding (PLNC). The aim of this work is to identify a pair of sources modulations from the superposed constellation, when PLNC is applied. For this purpose, the authors use the higher order statistics-based features in conjunction with genetic algorithm and information theory as a features selection method and the random forests as a classifier. Simulations are provided to assess the accuracy of the proposed algorithm through the average probability of correct identification for different modulation scheme pairs. It is shown that the algorithm achieves high-modulation identification in acceptable signal-to-noise ratio level at different relay position.