Automatic modulation classification using KELM with joint features of CNN and LBP

Abstract

Signal automatic modulation classification refers to modulation methods which can automatically classify and identify different communication signals. As the middle part of the signal detection and demodulation, modulation classification plays an important part in the area of military and civilian fields, such as software radio and electronic countermeasures. In recent years, modulation classification technology has made great achievements, but as the engineering requirements improving and the wireless communication channel environment become increasingly complex, there are still many problems to be solved. In order to solve the problems which many modulation methods failed to consider the relationship between different characters, an algorithm based on the feature fusion of the convolutional neural network (CNN) is proposed. In this paper, a smooth pseudo Wigner–Ville distribution is used to make the one-dimensional signal into an image, a model of CNN is used to extract image features, and the image features extracted by CNN and the artificial features are put together to make the feature fusion together. The fusion feature further improves the modulation classification performance. The simulation results show that the algorithm proposed in this paper improves the performance under low signal-to-noise ratio, indicating that this algorithm of constructing fusion features is more effective.