光纤振动信号特征提取及线性分类方法

Feature extraction and linear classification for Fiber vibration signals

在光纤预警系统(OFPS)中产生的入侵事件主要分为有害入侵和无害入侵。目前对于这两类扰动常规的特征提取方法通常是采用时域分析,但是对于不同有害入侵事件其时域特征区分不明显,因此时域处理不能更好体现它们之间的细节差别。通过对有害入侵信号的频谱进行统计研究发现,不同信号的频谱分布存在较为明显的差异性,因此本文将入侵信号变换到频域并借鉴声信号的处理方法,提出了一种基于能量占比特征的有害入侵事件识别算法。对采集到的振动信号进行预处理并计算功率谱密度(PSD),计算各信号不同频段的能量占比,并将其作为信号分类识别的特征。之后将能量占比特征作为样本送入分类器进行OFPS振动信号识别。在分类器的选择上,本文采用线性判别分析(LDA)分类器对信号进行识别,LDA能最大限度的保持原始数据信息,并有效区分振动信号。通过实验结果表明该算法在OFPS振动信号的识别研究中提高了有害入侵信号的识别率,从而验证了本算法的可行性,同时有效减少了识别时间。

The intrusion events in the optical fiber pre-warning system （OFPS） a re divided into two types,which are harmful intrusion event and harmless interference event.At present,the sig nal feature extraction methods of them are usually designed from the view of time domain.Howe ver,the differences of time-domain characteristics for different harmful intrusion events are not obvi ous,which cannot reflect the diversity of them in detail.In our statistical study,we find that the spectrum distributions of different intrusion signals have obvious differences,so that the intrusion signal is transformed int o the frequency domain for analysis. In this paper,an identification method of harmful intrusion event based on ener gy ratio is proposed according to the analysis method of acoustic signal.Firstly,the intrusion signals are pre- processed and the power spectral density （PSD） is calculated.Then the energy ratios of different frequency bands are calculated and the corresponding feature vector of each type of intrusion signal is further formed,which is inp ut into the classifier for training and recognition.The linear discriminant analysis （LDA） classifier is used to identi fy the signal,which can make use of the information of original data and distinguish different intrusion signals efficiently.Experimental results show that the algorithm improves the recognition accuracy of the intrusi on signal,which verifies the feasibility and validity of the algorithm.