Automatic Radio Signal Modulation Classification Based on Dimensional Information Expansion and Deep Residual Network

Neural network models can effectively classify radio modulation types, outperforming traditional methods. However, in real-world scenarios, the insufficient dimensions of received signals restrict the performance of the classifier. In this paper, a novel modulation recognition approach is proposed. Deep residual network is applied for electromagnetic signal classification. In addition, a method for expanding the dimensional information of the signals is proposed to further improve the classification results. First, the dimensions of the original data are reduced to 256 to simulate the absence of dimensional information. Next, the classification capabilities of residual networks with and without the dimension expansion algorithm are compared on a benchmark dataset. Experimental results show that the proposed dimension expansion method improves the classification accuracy by 8.39% in cases with insufficient dimensions.