【摘要】 本文从长短时间窗(LTA-STA)得到启发模拟实时波段类型识别.事件为首都圈及其附近的186个天然地震和174个人工爆破事件,用于抽取特征的波形信号为各观测台站波形3分量中的垂直分量波形,在各个事件的所有观测台站的垂直分量波形中,通过滑动窗口按同一准则去除被噪声淹没的部分台站波形,只选择留下未被噪声淹没的台站波形.对连续波形,使用长窗口沿波形时间轴进行滑动,每滑动一个步长就进行一次滤波处理,以滤除噪声,当滤波后的长窗口波形满足阈值条件,此时停止长窗口滑动.然后在滤波前的长时窗口中选取短时间窗口波形,提取特征,使用支持向量机进行分类训练和识别,最后以事件为单位进行识别,事件划分按以训练集为300个事件,测试集为60个事件进行划分,进行了训练和识别.然后又将训练集按照训练集240个事件,测试集120个事件进行划分.得到较好的识别结果.本文结果说明了波形类型实时识别的可行性,也可为后续实时波形检测和识别提供借鉴.更多还原

【Abstract】 Being inspired by the LTA-STA, a real-time seismic source-type recognition is simulated in this paper. The events are 186 natural earthquakes and 174 man-made blasting events near the metropolitan Beijing and surrounding areas. The waveform signal being used to extract features is the vertical component waveform in the 3 components of the observation station’s waveforms. In the vertical component waveforms of all the observing stations, the waveforms of some stations which being submerged by noise are removed by the same criterion through the sliding window, and only those waveforms that are not submerged by noise are selected. For continuous waveforms, a long-time window is used to slide along the waveform time-axis. Two contiguous window wave-sections are separated by an adjustable sliding step. The filtering for each window wave-section is performed, one by one by every sliding step. When the filtered long-time window waveform satisfies some threshold condition, the long-time window sliding is stopped. Then select the short-time window wave-section in the original long-time window wave-section, filter the motion waveform, extract features, use the support vector machine for classification training and recognition, and finally recognize the event as sample, and divide the event-sample set into two sub-sets: 300 events as the train-set, left 60 events as test-set. Training and testing results are got by the two sub-sets. Furthermore, totally 360 events are partitioned into two sub-sets: 240 events as the train-set, and left 120 events as test-set. The results show that the feasibility of real-time seismic source-type identification, and also provide a good exemplar for subsequent real-time seismic source-type detection and identification.更多还原