光纤周界入侵信号特征提取与识别方法的研究

提出一种基于互补经验模态分解(CEEMD)奇异值熵结合多核支持向量机(SVM)的入侵信号特征提取与识别方法。首先,采用CEEMD方法对入侵信号进行分解得到若干个本征模态函数(IMF)；其次,再对IMF分量进行奇异值分解,计算其奇异值熵；然后,根据奇异值熵筛选出有用IMF分量,构建特征向量；最后,采用多核支持向量机识别入侵信号。采用实际采集的攀爬,敲击,汽车,风等场外入侵信号进行了实验验证,结果表明：CEEMD方法有效解决了EEMD的残留白噪声问题,多核SVM比单核SVM具有更好的识别率,攀爬入侵信号识别率达到95%。

Research on intrusion signal extraction and recognition of optical fiber sensor perimeter

A intrusion signal extraction and recognition method based on complementary ensemble empirical mode decomposition(CEEMD),singular value entropy and multiple kernel support vector machine(SVM) is proposed.Firstly,the intrusion signals were decomposed using the CEEMD and a series of intrinsic mode functions(IMF) were gotten.Subsequently,IMFs were decomposed by singular value decomposition(SVD) and singular value entropy was calculated.Then,according to the singular value entropy,the useful IMF component was selected,and the feature vector was constructed.Finally,the multiple kernel support vector machine was used to identify the intrusion signal.The experiments were carried out by using the actual intrusion signals,such as climbing,knocking,car,wind,and so on.The experimental results show that the CEEMD method can solve residual white noise of EEMD,and the multiple kernel SVM has better recognition rate than the single kernel SVM,and the climbing intrusion signal recognition rate is 95%.