Multimodal Feature Fusion Recognition of Modulated Signals Based on Image and Waveform Domain

摘要

Communication signal modulation type recognition has a wide range of applications in electronic reconnaissance equipment such as electronic support, electronic intelligence and radar threat warning systems. The common modulation feature recognition algorithms usually only focus on one feature, ignoring the complementarity between different features. Considering the importance of feature fusion, this paper proposes a feature fusion method based on deep learning model. Extracting the image domain features and I/Q waveform domain features of the signal through suitable deep learning models, then combine the extracted features and use Kernel Principal Component Analysis (KPCA) to reduce the dimensionality of the joint features, finally obtain the classification recognition result in the classifier. Simulation experiments show that the signal recognition method based on feature fusion can have a higher recognition rate at low SNR than when only single features are considered, which can reach 93.15% at -2 dB.