HHT变换联合人工神经网络的核信号时频特征分析与识别  

Time-frequency feature analysis and recognition of nuclear signals based on HHT and ANN

作　　者：金晶[1] 魏彪[1] 冯鹏[1] 唐跃林[1] 周密[1] 任勇[1] 米德伶[1] 

机构地区：[1]光电技术及系统教育部重点实验室,重庆大学光电工程学院,重庆400044

出　　处：《核技术》2010年第6期451-456,共6页Nuclear Techniques

基　　金：国家军工预研专项基金项目(JW2025067);重庆市自然科学基金项目(CSTC2009BB2188)资助

摘　　要：用HHT变换联合BP人工神经网络分析处理252Cf自发裂变中子源驱动式核信息系统输出的随机核信号,获得信号时频特征,并分类识别核信号样本。理论分析和实验结果表明,基于HHT变换边际谱的特征提取方法,HHT反映信号时频域联合分布能较好地提取实际非平稳随机核信号的时频特征;用BP神经网络分类器对核信号样本分类识别,取得较高正确率,验证了此方法的有效性和合理性。In this paper, for a nuclear information system driven by a 252Cf spontaneous fission neutron source, the random nuclear signals were analyzed and processed using both Hilbert-Huang Transform （HHT） and artificial neural networks （ANN）. We got its time-frequency features and recognized its nuclear signal samples based on their classification. The results show that the time-frequency feature extraction method based on Hilbert marginal spectrum is applicable to separate the time-frequency feature of non-stationary nuclear random signal because it can reflect the time-frequency distributions. After classified recognition of nuclear signal samples using BP neural network, we also got ideal result that further verified the effectiveness and reasonability of the method.

关 键 词：随机核信号 希尔伯特-黄变换 人工神经网络 特征提取 分类识别 

分 类 号：TL375.17[核科学技术—核技术及应用] TL816.3