一种地震动信号特征提取与分类方法  

A Feature Extraction and Classification Method of the Vibratory Signal on Ground

作　　者：常克武 郭慧杰 葛军 CHANG Ke-wu;GUO Hui-jie;GE Jun(China Satellite Navigation Office,Beijing 100195,China;Beijing Institute of Radio Metrology and Measurement,Science and Technology on Metrology and Calibration Laboratory,Beijing 100039,China)

机构地区：[1]中国卫星导航系统管理办公室,北京100195 [2]北京无线电计量测试研究所,计量与校准技术重点实验室,北京100039

出　　处：《宇航计测技术》2020年第3期61-64,共4页Journal of Astronautic Metrology and Measurement

摘　　要：为了实时检测、识别和预警对地下基础设施的挖掘破坏活动,本文提出一种地震动信号特征提取与分类方法。通过提取小波包变换域和集合经验模态变换域的多域能量联合分布特征向量,构建改进的径向基神经网络分类模型,利用机器学习的方法提取稳定的信号多域融合特征,并实现准确的信号特征分类预测。由多类别挖掘信号的仿真实验结果可以看出,本文的算法和模型能有效提升地震动信号分类的准确率,对地震动干扰信号具有较强的鲁棒性。A method is proposed to detect,identify and warn the excavation and destruction of underground infrastructure in real time,by feature extraction and classification of the vibratory signal on ground.An improved radial basis neural network classification model is constructed by extracting the multi-domain energy joint distribution feature vector of the wavelet packet transform domain and the ensemble empirical mode transform domain,which adopts machine learning methods to extract stable multi-domain fusion features of the target signal and achieve accurate feature classification prediction.It can be concluded from the simulation experiment results of multi-category ground vibratory signals that the algorithm and model proposed in this paper can effectively improve the classification accuracy of the ground vibratory signals and have strong robustness to ground vibratory interference signals.

关 键 词：特征分类 径向基神经网络 多域特征融合 地震动信号 

分 类 号：TN911.7[电子电信—通信与信息系统]