基于小波和支持向量机的光纤微振动传感器模式识别

Pattern recognition of fiber-optic micro vibration sensor based on wavelet and SVM

为实现对双M-Z型光纤传感器的振动信号进行识别,提出一种基于小波能熵和支持向量机(SVM)的光纤传感信号模式识别方法。该方法对小波分解得到的各频段系数求解其能量信息熵,归一化后得到特征向量。其作为SVM的输入,通过选用合适的核函数和多类的分类方法,对SVM多类分类器进行建模。在多种振动信号的条件下,用测试样本对SVM分类器模型进行测试,测试结果表明:该方法对双M-Z型光纤微振动传感器的振动信号的分类达到了较高的识别率。

In order to recognize vibration signal from dual M-Z fiber-optic sensor,a mode recognition method based on wavelet energy entropy and support vector machine（SVM） is introduced.This method is applied to solve energy information entropy with each band coefficient obtained by wavelet decomposition,then normalized into feature vector,which is used as input for SVM multiclass classifier modeling,with proper kernel function and classification method.SVM model is tested by test sample in condition of multivibration signals,and test result indicates that high recognition accuracy can be achieved by this method.