Convolutional Neural Networks with SE for Modulation Format Identification

We propose a convolutional neural network model that incorporates an attention mechanism for modulation format classification. The simulation results indicate that the model's loss function reaches its minimum value after 250 epochs, showing no significant signs of overfitting. The accuracy of modulation format recognition achieved by the model is 98.5%, which is approximately 2% higher than that of the model without the attention mechanism. Furthermore, compared to the model lacking the attention mechanism, both the number of parameters and the time required to test a single image have not increased significantly. This suggests that our algorithm demonstrates superior performance, providing a new perspective for future modulation format recognition.