Research on Model Evaluation Technology Based on Modulated Signal Identification

Nowadays, the electromagnetic space is complex and variable, and the accurate identification of modulated signals is becoming increasingly difficult. And model quality can directly affect signal recognition. Therefore, this paper constructs a model performance evaluation system based on the recognition effect of modulated signals, and builds a hierarchical model of evaluation indexes from the classification performance, complexity performance, noise robustness and adversarial robustness of the model, to make a comprehensive and credible evaluation of the model quality from multiple dimensions. Through the experiment, we found that the complexity and classification performance of the model can affect the robustness of the model to a certain extent. The results of the evaluation show that the evaluation system can make a comprehensive and reasonable assessment of the quality of the model under the modulation-based signal recognition task.