Automatic Modulation Classification Using ResNeXt-GRU With Deep Feature Fusion
Lin Li, Yaxin Zhu, Zhigang Zhu
Abstract
With the integrated design and application of radar, communication and electronic reconnaissance, automatic modulation classification (AMC) is becoming increasingly significant in various applications, e.g. spectrum monitoring and cognitive radio, and has entered a stage of remarkable development. Recently, the AMC technology is suffering from complex and diverse signal types, low signal-to-noise ratio, poor algorithm robustness, and so forth. To cope with these flaws, this paper presents a modulation classification scheme based on deep feature fusion. In this work, the ResNeXt network is utilized to extract the distinctive semantic feature of the signal, and the gated recurrent unit (GRU) is employed to extract the time-series representation characteristic, respectively. Considering the complementarity between different features, a feature fusion model using discriminant correlation analysis (DCA) is proposed to fuse the output responses of the ResNeXt and GRU. The simulation results reveal that the proposed method achieves the superior performance, which is conducive to promoting the application of feature fusion in the AMC.