Detection And Feature Extraction Of Earthquake Precurse Wave For Classification

The physical mechanism of the occurrence and incubation is not clear in anearthquake, the earthquake prediction still remains an unsolved problem worldwide.Since Hiroo Kanamori put forward the concept of "precursor wave" in1974, manydomestic and overseas earthquake workers start to record and analyse these waves.Although different examples was showed, they reached an agreement on thecharacteristic and value of the precursor, it’s with a long period, bandwidth and acomplex frequency spectrum constituent, the research of the precursor is limited in thetraditional methond and the time domain feature analysis, improving the accurace ofthe earthquake prediction using the characteristic digged with modern signal process israrely seen, new methods and waveform analysis of the detection of the earthquakeprecursor wave with long-period distortion in the earth crust may be a breakthrough forthe earthquake prediction in the future.Under the support of earthquake monitoring center of Shannxi province. weemplaced the precursor recording equipment which was based on a novel liquid floatedlong cycle, ultra-low frequency detecting techniques in Ziwu, Louguan and Qianling inShannxi and recored the singal in real-time for a year, the magnanimity data wasanalysed, from the relation of the abnormal precursor and the earthquake we found theaccurance of the prediction was51.59%; Then the signal process was used to find thecharacteristic hidden in the waves, the local mean decomposition, waveletdecomposition and wigner-vile distribution time-frequency analysis method was usedto extract the effective feature of the precursor and the SVM was used to classify.The research result shows that, the classification accuracy based on the method oflocal mean decomposition is62.24%, the classification accuracy based on the methodof wavelet decomposition is72.03%, the classification accuracy based on the methodof wigner-vile distribution is66.43%, the result indicates that to realize the signalidentification with innovative precursor detecting techniques and modern signalprocessing, the precursor classification accuracy can reach as high as72.03%, andonce again proves the research value of the precursor theoretical and the applicationvalue of the time-frequency analysis method in non-stationary signal especially in theearthquake signal analysis, and further research will be launched in succession.