基于LCD和改进PNN的光纤周界振动信号识别

Optical fiber perimeter vibration signal recognition based on LCD and improved P NN

摘要:

针对马赫-曾德尔光纤周界系统振动信号扰动信 息提取及识别中的问题,提出了一种 基于局部特征尺度分解(LCD)和改进概率神经网络(PNN)的识别方法。首先,采用LCD将振动 信号分解成一系列内禀尺度分量(ISC),再将分解得到的ISC分量每连续3阶一组进行独立成 分分析(ICA),提取扰动信息。其次,提取振动信号的峭度、排列熵、瞬时幅度标准差和瞬 时频率标准差构造具有准确描述能力的特征向量。最后,采用经模糊C均值聚类(FCM)优化后 的PNN对振动信号进行识别分类。利用六种振动信号实验数据进行验证。结果表明,该方法 能够高效准确的识别六种振动信号,平均识别率达到97.17%,识别时 间为0.78 s。该方法在 有效信息提取和振动信号识别方面明显优于传统的LCD算法和PNN算法,具有实际应用价值。

Abstract:

Aiming at the problem of vibration signal disturbance information extraction and recognition in Mach-Zehnder fiber perimeter system,a rec ognition method based on local feature scale decomposition and improved probabil istic neural network is proposed.First,the is used to decompose the vibration signal into a series of intrinsic scale components,and then the intrinsic scale components obtained by the decomposition are subjected to independent component analysis for every 3consecutive groups to extract disturbance information.Sec ondly,extract the kurtosis,permutation entropy,instantaneous amplitude standa rd deviation and instantaneous frequency standard deviation of the vibration sig nal to construct a feature vector with accurate description ability.Finally,th e fuzzy C-means clustering-optimized probabilistic neural network is used to id entify and classify the vibration signals.The experimental data of six kinds of vibration signals are used for verification.The results show that this method can efficiently and accurately identify six kinds of vibration signals,with an average recognition rate of 97.17% and a recognition time of 0.78s.This method is significantly superior to traditional local feature scale decomposition algor ithms and probabilistic neural network algorithms in terms of effective informat ion extraction and vibration signal recognition,and has practical application v alue.