一种基于集成学习的DBN模型分类方法

A classification method of DBN model based on ensemble learning

雷达信号分类是雷达信号电子侦察的关键技术之一,针对利用深度学习模型进行雷达信号分类时其性能不稳定的缺点,提出了一种基于集成学习的深度信念网络模型进行分类的方法.通过深度信念网络模型不同层间的特征抽取,通过不同的分类器得到不同的分类结果,再将分类结果进行集成,得到最终的输出.待分类的雷达信号由12部雷达产生,包括常规、参差、频率捷变和抖动四种雷达.仿真结果表明,该模型的分类错误率较低,鲁棒性较好.

Radar signal classification is one of the key technologies of radar signal electronic reconnaissance. Aiming at the shortcomings of radar signal classification using deep learning model, a method of classification based on ensemble learning deep belief network model was proposed. Through the feature extraction between different layers of the deep belief network model, different classification results were obtained by different classifiers, and the classification results were integrated to obtain the final output. The radar signals to be classified were generated by 12 radars, including conventional, staggered, frequency agile and jitter radars. The simulation results showed that the model has lower classification error rate and better robustness.