The Research And Realization Of Seismic Waveform Feature Extraction Algorithm

The characteristics of the seismic waveform include the waveform data characteristics of natural earthquake and artificial explosion. Extracting the seismic waveform is mainly to extract useful characteristics according to signal processing method and seismic waveform data's own characteristics for identifying the waveform of natural earthquake and artificial explosion from time domain, frequency domain and so on. The key problem of this is the selection and extracting of the signal characteristics.The Hilbert-Huang Transform method put forward by Chinese American academician of National Academy of Engineering in 1998. It is a new signal processing and analysis method which differs from the global analysis characteristics of former signal analysis methods and has a unique superiority in the localized characteristic area of the signal research. It is a localized analysis method which is a time frequency localization analysis method and is more suitable than Fourier Transform and Wavelet Transform. This method can reflect the signal energy distribution regularity on every scale of the time and frequency more really. If the HHT method is used in processing the seismic resources, the existing problem when using other methods can be effectively solved and we can get breakthroughs on different level in different aspects of seismic resources, such as signal decomposition, time frequency analysis, and instant parameter solution and so on.According to this, this paper using realized the extracting of the seismic waveform with HHT method. First, this paper analyzed the characteristics of natural earthquake and artificial explosion in time domain and frequency domain and did some necessary pre-process. Then carry out the empirical mode decomposition to the waveform signal of natural earthquake and artificial explosion and extract the three characteristics of corresponding maximum amplitude:the period (TAmax), cepstrum average (Cave) and the maximum of autocorrelation function (Mxc). At last, identify the waveform of natural earthquake and artificial explosion with support vector machine strictly according to the sample classify method. In the experiment, based on the C method, recognition rate is more optimistic, on the empirical mode decomposition of the intrinsic mode function to extract the first three of these three features the recognition rate of 99.8%.Based on the U method, the sample divided by the strict method, after repeated and a large number of experiments, on the empirical mode decomposition of the intrinsic mode function to extract the first three of these three features the average recognition rate over 97%.The results indicate that the waveform characteristic parameters extracted based on HHT method has a good identification rate. The period (TAmax), cepstrum average (Cave) and the maximum of autocorrelation function (Mxc) of corresponding maximum amplitude extracted according to this method can be used as a new characteristic parameter in the identifying the waveform signal of natural earthquake and artificial explosion. It has a good identification effect.