基于改进半监督朴素贝叶斯的LPI雷达信号识别

发布时间：2018-04-23 14:28

本文选题：信号识别 + 朴素贝叶斯　；参考：《系统工程与电子技术》2017年11期

【摘要】：针对先验信息残缺的非合作电子对抗背景下的低截获概率雷达信号识别问题,提出一种基于改进的半监督朴素贝叶斯的识别算法。该算法首先提取出4种低截获概率(low probability of intercept,LPI)雷达信号的双谱对角切片作为识别特征;针对传统的半监督朴素贝叶斯(semi-supervised Na?ve Bayes,SNB)在更新训练样本集过程中会产生迭代错误的不足,利用改进的SNB(Revised SNB,RSNB)算法构建分类器,完成对测试样本的识别。该方法通过在无标记样本集生成的置信度列表中选取置信度较高的样本添加到有标记样本集中,再利用预测后的分类结果对分类器参数(即特征期望向量s醡i和方差向量σi)进行改进,有效解决了传统算法分类精度低且分类性能不稳定等缺点。理论分析和仿真结果表明,在LPI雷达信号识别问题,相比于SNB算法和传统的主成分分析加支持向量机法(principal component analysis-support vector machine,PCA-SVM),该算法具有更高的分类识别率和更好的分类性能。
[Abstract]:A new algorithm based on improved semi-supervised naive Bayes is proposed for low probability radar signal recognition under the background of non-cooperative electronic countermeasures with incomplete prior information. The algorithm firstly extracts four kinds of low probability of interception (LPI) radar signals with bispectral diagonal slices as recognition features, aiming at the shortcomings of iterative errors in updating the training samples by traditional semi-supervised naive Bayes Bayes (semi-supervised Bayes Bayes). The improved SNB(Revised SNB algorithm is used to construct the classifier to identify the test samples. In this method, samples with high confidence level are selected from the list of confidence levels generated from unmarked sample sets and added to the marked sample set. Then using the predicted classification results to improve the classifier parameters (I. E. feature expectation vector s press I and variance vector 蟽 I) effectively solve the shortcomings of the traditional algorithm such as low classification accuracy and unstable classification performance. The theoretical analysis and simulation results show that the algorithm has higher classification rate and better classification performance than SNB algorithm and principal component analysis-support vector machine method of principal component analysis plus support vector machine.
【作者单位】：空军工程大学航空航天工程学院;
【基金】：航空科学基金(20145596025,20152096019)资助课题
【分类号】：TN957.51

【相似文献】