MLResNet: An Efficient Method for Automatic Modulation Classification Based on Residual Neural Network
Ming Liang Xue, Ming Huang, Jing Jing Yang, Ji Da Wu · 2021
In the face of a complex electromagnetic environment, the modulation mode of communication signals is becoming increasingly complicated. Existing modulation mode recognition methods of communication signals cannot accurately and quickly identify the modulation mode of communication signals. In this letter, we propose an efficient architecture for automatic modulation classification (AMC) based on residual neural network (ResNet). We combine the improved residual neural network with long short-term memory network (LSTM) to obtain a new network structure (MLResNet), which solves the problems of gradient disappearance and too many parameters. In the experiments, MLResNet reaches the overall 24-modulation classification rate of 96.60% at 18 dB SNR on the well-known DeepSig dataset.