Research On Intrusion Signal Recognition Method Of Optical Fiber Perimeter System

The fiber perimeter security system senses the vibration signal through the phase difference of the optical signals collected in the fiber sensor,and then identifies various types of intrusion signals according to the characteristic information of different vibration signals.The difficulty lies in that the system is not only susceptible to interference from surrounding environment,and the recognition effect is also susceptible to the vibration signal feature extraction and description methods.Therefore,on the basis of filtering out more disturbance signals,it is a worthwhile research to ensure the quality of signal decomposition and to select appropriate feature description methods.In this dissertation,three different non-linear and non-stationary signal analysis methods are studied,which are used in the fiber perimeter security system to compare the recognition rates of various intrusion signals.The results show that the fast adaptive empirical mode decomposition method guarantees signal decomposition performance,it is also better in terms of speed,combined with time domain features,improved recognition rate,and reduced false positive rate,so it is most suitable for fiber perimeter security systems.This dissertation studies the characterization,extraction and recognition of various types of signals.The main work contents are as follows:1.The type,characteristics,laying method and applicability of the sensor in the fiber perimeter security system are studied.The appropriate fiber optic sensor is selected according to the actual demand.Combined with the optical path mathematical model,the collected signal is theoretically included the information of vibration and location.2.The framing of the collected signal,the interception of the vibration signal and the filtering of the partial disturbance signal are studied.The short-time energy and the over-threshold rate characteristics of the time domain are used to discriminate the intercepted vibration signal,the statistical characteristics of the signal threshold rate are used after preprocessing.And the ratio difference coefficient feature excludes some of the disturbance signals.3.The principle of CEEMD,LMD and AEMD methods is studied,and it is applied to the decomposition and analysis of the intrusion signal of the fiber perimeter security system.Then the correlation coefficient method is used to select the optimal component,the corresponding kurtosis is calculated and used as the frequency domain feature.Finally,the feature vector is constructed.4.The three common classification algorithms are studied and applied to the intrusion signal recognition for comparative analysis.Finally,the optimal multi-core support vector machine is selected as the classifier for intrusion signal recognition.