High-Order Modulation Recognition of Communication Signals Based on Feature Fusion

To address the issue of poor recognition performance of high-order modulation signals under low signal-to-noise ratio conditions, a recognition model based on feature fusion and multi-kernel classifier is designed for an open dataset of high-order modulation signals. Firstly, the advantages of convolutional neural networks and gated recurrent units are fully utilized to extract deep features of high-order modulation signals. Additionally, high-order cumulants of the high-order modulation signals are extracted as supplementary features. Finally, a multi-kernel learning classifier is constructed using multi-kernel linear combination to fuse the aforementioned two types of features, effectively improving the recognition accuracy of high-order modulation under low signal-to-noise ratio conditions. Experimental results demonstrate that the feature fusion recognition model significantly improves the recognition performance of high-order modulation signals under low signal-to-noise ratio conditions, with an average recognition accuracy exceeding 82%.