摘要
针对微震信号难以精确识别的问题,提出一种基于RST-NMF微震信号时频分析和分类方法。首先对微震信号进行S变换得到时频矩阵,然后在频率方向上进行重排,再借助非负矩阵分解技术得到时、频域的分解向量,从中提取宏观、微观统计量构造信号的特征空间,最后采用SVM进行分类。在三道沟井田的试验结果表明,RST时频分析方法对频域分散的能量团有很好的聚集性,时频矩阵经NMF分解最大程度上获取微震信号的局部特征和内在联系,提取分解向量的宏观和微观统计量保证了信号特征空间的完备性,有效地避免了分类时过拟合的发生,分类准确率达到了94%。
Aiming at the problem of micro-seismic signals being difficult to accurately identify,a time-frequency analysis and classification method based on RST-NMF model was proposed.Firstly,S-transform was performed for a micro-seismic signal to obtain a time-frequency matrix,and this matrix was rearranged in frequency direction.The non-negative matrix factor(NMF)technique was used to obtain resolved vectors in time domain and frequency one.Macro and micro statistics were extracted from these vectors to construct the signal’s feature space,and SVM was used to do classification.The test results of SAN Daogou minefield showed that the time-frequency analysis method of RST has a good clustering for frequency domain dispersed energy groups;local characteristics and internal relations of a micro-seismic signal are obtained to the greatest extent from the time-frequency matrix with the NMF technique;extracting macro and micro statistics of resolved vectors can ensure the completeness of a signal feature space and effectively avoid the occurrence of over-fitting during classification;the correct rate of classification can reach 94%.
作者
张法全
王海飞
王国富
叶金才
ZHANG Faquan;WANG Haifei;WANG Guofu;YE Jincai(School of Information and Communication,Guilin University of Electronic Technology,Guilin 541004,China;Guangxi Provincial Key Lab of Wireless Wideband Communication&Signal Processing,Guilin 541004,China)
出处
《振动与冲击》
EI
CSCD
北大核心
2019年第17期1-7,共7页
Journal of Vibration and Shock
基金
国家自然科学基金(61362020
61761009)
关键词
微震信号
RST
NMF
SVM
micro-seismic signal
RST(rearranged S transform)
NMF(non-negative matrix factor)
SVM(support vector machine)
