第一，对传统预分选流程进行改进，提出特征重要性估计算法，自适应选择高可

第一，对传统预分选流程进行改进，提出特征重要性估计算法，自适应选择高可分特征进行分选，完成分选流程，并进行了仿真验证；第二，将脉冲描述字与深度学习网络结合进行特征提取，并利用已知雷达库结合半监督理论，实现新型分选流程；第三，对分选得到的辐射源结果进行威胁等级评估；最后对特征重要性估计分选算法及威胁等级评估算法进行基于DSP6678的系统实现，满足工程设计要求。

第一，对传统预分选流程进行改进，提出特征重要性估计算法，自适应选择高可分特征进行分选，完成分选流程，并进行了仿真验证；第二，将脉冲描述字与深度学习网络结合进行特征提取，并利用已知雷达库结合半监督理论，实现新型分选流程 ；第三，对分选得到的辐射源结果进行威胁等级评估；最后对特征重要性估计分选算法及威胁等级评估算法进行基于DSP6678的系统实现，满足工程设计要求。

First, the traditional pre separation process is improved, the algorithm of feature importance estimation is proposed, the high separable feature is selected adaptively, the separation process is completed, and the simulation verification is carried out; second, the pulse description word is combined with the deep learning network to extract the feature, and the known Radar Library is combined with the semi supervision theory to realize the new separation process; third, the separation results Finally, the system implementation based on dsp6678 is carried out for the sorting algorithm of the feature importance estimation and the threat level evaluation algorithm to meet the engineering design requirements.<br>