电晕放电辐射信号的特征提取和模式识别方法研究  

Feature Extraction and Pattern Recognition of Corona Discharging Signals

作　　者：胡小锋[1] 刘卫东[1] 周帅 

机构地区：[1]军械工程学院静电与电磁防护研究所,石家庄050003

出　　处：《装备环境工程》2017年第4期57-61,共5页Equipment Environmental Engineering

摘　　要：目的研究电晕放电辐射信号的特征提取和模式识别方法。方法在分析信号特征提取方法的基础上,对实测的电晕放电辐射信号特征提取,利用概率神经网络开展电晕放电辐射信号目标识别,检验特征提取的有效性。结果以奇异值作为输入特征量的PNN在整体上效果更优,稳定性好,对两类不同放电辐射信号的正确识别率均可达到80%以上,并且当输入特征量个数达到10个时,对实测样本的正确识别率均达到了最高值。电晕放电的正确识别率为96.7%,火花放电的正确识别率为93.3%。结论该方法能基本满足实际放电信号的识别应用。Objective To research methods for feature extraction and pattern recognition of corona discharge radiation signals. Methods Based on the analysis of signal feature extraction method, the signal feature of corona discharge radiation measured was extracted. The probabilistic neural network was adopted to identify corona discharge radiation signal target to test the effectiveness of the proposed feature extraction. Results The PNN with singular value as the input characteristics was overall better in effect and good in stability. Its correct rate of recognition of two kinds of different discharge radiation signals could be higher than 80%. When ten characteristics were input, the correct recognition rate reached the peak of the measured samples. The correct recognition rate of corona discharge was 96.7%. The correct recognition rate of spark discharge was 93.3%. Conclusion This method can basically meet the recognition and application of actual discharge signals

关 键 词：电晕放电 特征提取 模式识别 概率神经网络 

分 类 号：TJ06[兵器科学与技术—兵器发射理论与技术] TN911.7[电子电信—通信与信息系统]