| 摘要: |
| 为提高辐射源个体识别准确度，解决工程化应用问题，同时避免在信号样本有限的情况下单一识别算法的局限性，提出了一种结合卷积神经网络和随机森林的辐射源个体识别方法。该方法分别利用卷积神经网络和随机森林训练生成两组个体识别模型，然后采用识别概率统计法生成不同辐射源个体的综合权值向量，最后根据权重向量形成针对不同辐射源个体的综合识别模型。仿真结果表明，相较于单一算法，所提方法能够提升整体识别准确率，同时，对不同辐射源个体均有较好的适用性。 |
| 关键词: 辐射源个体识别 卷积神经网络 随机森林 综合权值 |
| DOI： |
|
| 基金项目: |
|
| Specific emitter recognition by combining convolutional neural network and random forest |
| HOU Siyao,LI Wei,LI Yongguang,LING Jie,HUANG Qianchuan |
| (Science and Technology on Electronic Information Control Laboratory,Chengdu 610036,China) |
| Abstract: |
| In order to improve specific emitter identification accuracy and solve the engineering application problems,a new specific emitter recognition algorithm based on the combination of convolutional neural network(CNN) and random forest is proposed.This algorithm can avoid limitations of single algorithm when the signal sample is limited.It provides two groups of specific emitter identification model by using CNN and random forest,then applies probability statistics to obtain the comprehensive weight of different specific emitter,finally produces comprehensive identification model by using comprehensive weight.Simulation results show that the proposed method can increase specific emitter identification accuracy and is suitable for different specific emitters. |
| Key words: specific emitter recognition convolutional neural network random forest comprehensive weight |
