基于现代谱估计的光纤周界安防系统模式识别

发布时间：2018-06-29 02:02

本文选题：光纤周界安防系统 + 模式识别　；参考：《光通信技术》2017年05期

【摘要】：针对分布式光纤周界安防系统,为了有效识别事件类型并兼顾系统实时性,提出一种两级振动模式识别方案。第一级在时域上提取短时能量和短时过阈值率两个特征,与动态阈值进行比较,筛选疑似扰动样本;第二级对疑似扰动样本进行现代功率谱估计,提取各频段功率分布特征,并联合时域特征构建特征向量输入支持向量机(SVM)进行模式识别。
[Abstract]:For distributed optical fiber perimeter security system, a two-stage vibration pattern recognition scheme is proposed in order to identify the event type effectively and give consideration to the real-time performance of the system. In the first stage, two characteristics of short-time energy and short-time over-threshold rate are extracted in time domain, and compared with the dynamic threshold, the suspected disturbance samples are screened, and the second stage is used to estimate the power spectrum of the suspected disturbance samples to extract the power distribution characteristics of each frequency band. Feature vector input support vector machine (SVM) is combined with time domain features for pattern recognition.
【作者单位】：上海海事大学信息工程学院;
【分类号】：TN253

【相似文献】