基于数据挖掘的光纤周界振动信号识别  

Recognition of vibration signal around optical fiber based on data mining

作　　者：冯丽佳 苑柳青 熊丽丽 FENG Lijia;YUAN Liuqing;XIONG Lili(Hebei Polytechnic Institute,Shijiazhuang 050091,chian)

机构地区：[1]河北工程技术学院,石家庄050091

出　　处：《激光杂志》2023年第7期137-142,共6页Laser Journal

基　　金：河北省教育厅科学技术研究项目(No.1803114)。

摘　　要：针对光纤周界振动信号中的噪声干扰以及通过单一特征无法全面描述振动信号的问题,研究基于数据挖掘的光纤周界振动信号识别方法。采用SVD(奇异值分解)方法对所采集的振动信号进行去噪处理,基于秩阶次值重构信号,由此消除噪声;针对无噪音振动信号,将峭度。排列熵与瞬时频率标准差作为特征,并进行特征融合,获取振动信号的特征向量;将特征向量输入概率神经网络中,利用模拟退火算法确定网络中的平滑因子参数值优化网络结构,通过学习与训练的过程输出振动信号识别结果。实验结果显示秩阶次值k为5时信号去噪性能最好,信号特征之间差异较为显著,易于区分,识别准确率平均为94.08%。Aiming at the noise interference in the optical fiber perimeter vibration signal and the problem that the vibration signal can not be fully described by a single feature,the optical fiber perimeter vibration signal recognition method based on data mining is studied.SVD(singular value decomposition)method is used to denoise the collected vibration signal,and the signal is reconstructed based on the rank value,so as to eliminate the noise;For the noiseless vibration signal,the kurtosis is.Permutation entropy and instantaneous frequency standard deviation are used as features,and feature fusion is carried out to obtain the feature vector of vibration signal;The feature vector is input into the probabilistic neural network,the smoothing factor parameter value in the network is determined by simulated annealing algorithm,the network structure is optimized,and the vibration signal recognition result is output through the process of learning and training.The experimental results show that the signal denoising performance is the best when the rank value k is 5,and the difference between the signal features is obvious,which is easy to distinguish.The average recognition accuracy rate is 94.08%.

关 键 词：数据挖掘 光纤周界 振动信号识别 信号去噪 特征提取 概率神经网络 

分 类 号：TN212[电子电信—物理电子学]