基于信号时频关联分析的光纤入侵振源识别算法

An Intrusion Recognition Algorithm Based on Time-frequency Domain Analysis of Signal in Optical Fiber Intrusion System

本文针对光纤预警系统提出了一种基于信号时频关联分析的光纤入侵振源识别算法.在特征提取阶段,本文采用信号的周期性和有序程度2个指标来描述信号的内在规律性,引入短时自相关算法和小波变换算法分别从时域和频域的角度提取信号的基音周期和小波功率熵特征,形成对信号的时频关联分析;在分类器设计阶段,根据本文所提取特征的特性,采用K临近算法作为分类器.经大量实测数据测试和仿真验证,证实了该算法模型具有较好的分类效果和较低的算法复杂度.

An algorithm of intrusion recognition,based on association analysis of time-frequency domain for optical fiber vibration signal,is proposed in this paper.In the process of extracting signal features,period and order degree are adopted to describe the signal inherent regularity.In time domain,the short-time autocorrelation function is introduced to obtain the pitch period.In frequency domain,the power entropy of wavelet coefficient is acquired with the method of wavelet transform.The combination of those two methods achieves association analysis of time-frequency domain for signal.In classifier design phase,the k-nearest neighbor algorithm is used as the classifier on basis of the feature extracted in the first phrase.The proposed algorithm model is proved to have good performance in classification after having been tested with numerous actual data and simulations.Also,the complexity of this algorithm is relatively low which makes the algorithm highly applicable.