AR模型预测误差比在地震事件识别中的应用

Predication Error Ratio of AR Model and Its Application to Seismic Event Discrimination

根据天然地震与核爆地震信号的 P波差异 ,提出了基于 AR模型预测误差比的地震信号分类和识别方法。该算法首先应用 AR模型对地震信号 P波初至时段及其后一时段进行建模 ,分别求出其预测误差 ,从而计算得出 P波的预测误差比 ,然后利用预测误差比作为特征判据对地震信号进行分类与识别。对 2 0次地震时间的分类和识别实验结果表明 ,本文方法能够对天然地震及核爆地震两类事件进行有效的分类与识别。

For CTBT seismic verification, it is the ultimate objective to efficiently discriminate the underground nuclear explosion events. A novel identifying character based on the predication error ratio of AR model is presented. The differences in the spectrum power of P phase between natural earthquake and underground nuclear explosion are discussed firstly. Then P phase onset time and the next time intervals are modeled by AR models, and predication errors for each model are computed. Finally, the predication error ratio can be obtained. And it is used as a character for event discrimination. Experimental results with 20 actual seismic signals, which arose by underground nuclear explosions and natural earthquakes, show that the character can effectively classify the events.