作者：邹瑛珂1, 贾云飞1, 刘素芸2
作者单位：1. 南京理工大学机械工程学院，江苏 南京 210094;
2. 山西北方兴安化学工业有限公司，山西 太原 030008
关键词：协方差叠加经验模态分解;经验模态分解;特征提取;随机森林;人车地震动信号
摘要：
为解决在野外环境中对低信噪比人车地震动信号进行正确识别的问题，该文提出一种基于希尔伯特包络线提取和改进经验模态分解的信号分解方法——协方差叠加经验模态分解的人车地震动信号识别算法。首先对目标的地震动信号进行希尔伯特变换，对其进行平滑，获取信号的包络线，然后对包络线进行EMD分解后运算协方差选出一个与原信号最相关的IMF和一个最无关的包含高频噪音的IMF作差平均并加回原信号中实现对原信号的信噪比提升。再次进行EMD分解从而得到具有高信噪比的新IMF分量。通过对所得到的IMF分量进行时频域的信号分析就可以实现对于目标地震动信号的特征提取，最后使用随机森林分类算法对信号进行分类，从而实现对人车目标的识别和分类。最后识别准确率大于90%，高于其他传统方法在该环境下的识别率。
Improved EMD decomposition based recognition algorithm for pedestrian and vehicle ground motion signals
ZOU Yingke1, JIA Yunfei1, LIU Suyun2
1. College of Mechanical Engineering, Nanjing University of Technology, Nanjing 210094, China;
2. Shanxi North Xing’an Chemical Industry Co., Ltd., Taiyuan 030008, China
Abstract: In order to solve the problem of correct recognition of human vehicle ground motion signals in the field environment, this paper proposes a human vehicle ground motion signal recognition algorithm based on Hilbert envelope extraction and improved empirical mode decomposition. Firstly, Hilbert transform is used to smooth the ground motion signal of the target, and the envelope of the signal is obtained. Then EMD is used to decompose the envelope, and covariance is calculated to select an IMF which is most related to the original signal and an IMF which contains high-frequency noise which is most irrelevant to the original signal. The difference is averaged and added back to the original signal to improve the signal-to-noise ratio of the original signal. EMD decomposition is performed again to obtain a new IMF component with high SNR. Through the time-frequency signal analysis of the IMF component, the feature extraction of the target ground motion signal can be realized. Finally, the random forest classification algorithm is used to classify the signal, so as to realize the recognition and classification of the human and vehicle targets. Finally, the recognition accuracy is more than 90%, which is higher than that of other traditional methods in this environment.
Keywords: covariance superposition empirical mode decomposition;empirical mode decomposition;feature extraction;random forest;human vehicle ground motion signal
2022, 48(4):85-94 收稿日期: 2020-12-28;收到修改稿日期: 2021-02-08
基金项目:
作者简介: 邹瑛珂（1996-），男，重庆市人，硕士研究生，专业方向为信号处理与模式识别
参考文献
[1] 王建平, 焦国太, 秦栋泽, 等. 基于地震动信号的目标识别[J]. 科学技术与工程, 2013, 13(1): 191-195
[2] 鹿洪荣, 隋文涛. 基于LPSO最大峭度熵反褶积的齿轮箱故障诊断[J]. 组合机床与自动化加工技术, 2020(10): 117-121
[3] 陶良小, 靳平, 王同东, 等. 基于线性阵列包络线偏移叠加的车辆检测方法[J]. 兵工学报, 2015, 36(5): 869-873
[4] 吴甄非, 余志军, 刘海涛. 基于HHT的人员脚步信号识别算法[J]. 计算机工程, 2011, 37(19): 168-170
[5] MURALIDHARAN A, SUGUMARAN V, SOMAN K P, et al. Fault diagnosis of helical gear box using variational mode decomposition and random forest algorithm[J]. Structural Durability & Health Monitoring, 2014, 10(1): 55-80
[6] 刘自然, 胡毅伟, 石璞, 等. 基于改进经验小波变换的滚动轴承故障特征提取方法研究[J]. 中国测试, 2019, 45(10): 10-15
[7] XU S Q, CHEN Z W, CHAI Y, et al. Sampling of bandlimited signals in the offset linear canonical transform domain based on reproducing kernel Hilbert space[J]. International Journal of Wavelets Multiresolution and Information Processing, 2020, 18(2): 14
[8] 陈志恒, 罗文斌, 常俊杰, 等. 基于EMD的神经网络空耦超声储油罐液位检测[J]. 中国测试, 2021, 47(1): 9-14
[9] 孙新建, 李成业. 岩石爆破震动信号分析的EEMD滤波方法研究[J]. 水利水电技术, 2013, 44(1): 101-104
[10] JAYANTHAN A M B , SELVI K K. An automated ECG signal diagnosing methodology using random forest classification with quality aware techniques[J]. International Journal of Trend in Scientific Research and Development, 2020, 4(3): 1174-1179.
[11] 赵亚男. 基于CART决策树和RBF神经网络的山东省空气污染状况预测评估[J]. 统计学与应用, 2019, 8(5): 725-733.
[12] 周继裕, 陈思露, 符少文. 基于均值滑动滤波算法和STM32电子秤的设计[J]. 现代电子技术, 2017, 40(10): 10-12
