【摘要】 研究了如何从天然地震和人工爆破事件的波形记录中提取出有效、适用的波形特征,以用于对爆破事件的识别.首先对波形记录进行了4层小波包变换;然后对变换得到的最后一层小波包系数提取3种波形特征:能量比特征、香农熵特征及对数能量熵特征;最后利用v-SVC支持向量分类机对这3种特征的分类能力进行了外推检验.通过选用不同地区、不同台站、不同震级的天然地震与人工爆破的波形记录,力求提取的特征量能尽可能地反映天然地震与人工爆破波形的本质区别,尽量弱化震中距、震级等因素对识别效果的影响.结果表明,上述3种特征中以香农熵特征的识别效果最好,能反映天然地震与人工爆破的本质区别,可作为识别天然地震与人工爆破的一个有效判据.更多还原

【Abstract】 Research on how to extract seismic wave features from earthquakes and explosions and how to discriminate explosions from earthquakes based on these features. Firstly,the transform of 4-layer wavelet packet is performed on the wave records. Secondly,the last layer coefficients of wavelet packet from the transform are employed to extract 3 types of wave features: energy ratio,Shannon entropy and logarithmic energy entropy. Thirdly,these features are supplied to a classifier of v-SVC support vector machines for verifying the capabilities of these features. In order to weaken undesirable effect of event epicenter-distance and magnitude on the recognition,we tried to extract more essential features of the wave records gathered from different regions,different observatories and various events almost covering whole magnitude ranges. The results show that,among the above three features,the feature of Shannon entropy is the best candidate for discriminating explosions from earthquakes. This may be an effective criterion in explosion recognition.更多还原