基于多分类高斯SVM的光纤信号的模式识别方法  

Pattern recognition method of distributed optical fiber sensing signal based on multi-classification Gaussian SVM

作　　者：吴明埝 沈一春[2] 陈青青 王道根 李松林 谢书鸿[2] 尹建华 徐拥军 WU Mingnian;SHEN Yichun;CHEN Qingqing;WANG Daogen;LI Songlin;XIE Shuhong;YIN Jianhua;XU Yongjun(Zhongtian Power Optical Cable Co.Ltd.,Nantong 226463,China;Jiangsu Zhongtian Technology Co.Ltd.,Nantong 226463,China)

机构地区：[1]中天电力光缆有限公司,南通226463 [2]江苏中天科技股份有限公司,南通226463

出　　处：《激光技术》2025年第1期128-134,共7页Laser Technology

基　　金：江苏省科技项目产业前瞻与关键核心技术基金资助项目(BE2022130)。

摘　　要：为了有效提升光纤信号识别精度,采用了一种基于多分类的高斯支持向量机(SVM)的信号事件类型判别方法,先通过汉宁窗卷积的方法以及95%能量的原则来识别事件发生始末段信息,再从时域、频域以及尺度域等角度出发,对归一化后的多种特征参数的均值与离散性进行分析,并选取合适的主要特征参数,最后采用基于多分类高斯SVM算法对3组不同事件类型进行了分类识别,通过理论分析和实验验证,取得了不同类型光纤事件信号的数据。结果表明,对30组实验数据的事件类型进行模式识别,正确率在96%以上。该方法流程满足了光纤传感的事件信号高精度识别要求,对光纤传感器应用具有较重要的参考价值。In order to effectively improve the accuracy of fiber optic signal recognition,a signal event type discrimination method based on multi classification Gaussian support vector machine(SVM)was adopted.Firstly,the Hanning window convolution method and the principle of 95%energy were used to identify the information of the beginning and end stages of event occurrence.Then,from the perspectives of time domain,frequency domain,and scale domain,the mean and discreteness of various normalized feature parameters were analyzed,and appropriate main feature parameters were selected.Finally,the multi classification Gaussian SVM algorithm was used to classify and recognize three different event types.Theoretical analysis and experimental verification were conducted,and data on different types of fiber optic event signals were obtained.The results showed that pattern recognition of event types in 30 sets of experimental data achieved an accuracy rate of over 96%.This method process meets the high-precision identification requirements of event signals in fiber optic sensing and provides important reference value for the application of fiber optic sensors.

关 键 词：传感器技术 多分类高斯支持向量机 模式识别 事件信号 

分 类 号：TP212.14[自动化与计算机技术—检测技术与自动化装置]