【摘要】 地震早期预警系统能够在震后数秒内快速估计地震影响，并通告可能受影响的区域，可以有效避免人员伤亡和财产损失。然而，传统地震早期预警系统需要耗费大量资金建设高密度地震台网。近些年，随着智能手机的普及，基于智能手机的地震早期预警(SEEW)研究受到广泛关注。SEEW系统的误警率和漏检率一直是研究的难点，也是影响系统实际应用部署的关键。采用深度学习、统计推断等模型构建了一种SEEW决策方法：(1)通过在长短时记忆(LSTM)模型中引入注意力机制，建立手机地震信号触发模型；(2)提出了一种基于时空一致的手机触发比例算子，并建立了基于统计推断的地震判别模型。通过仿真平台测试了方法性能，结果表明，该方法与类似方法相比，在极低的误警率下，地震成功预警率有显著提升。更多还原

【Abstract】 Earthquake Early Warning(EEW)can quickly estimate the impact of an earthquake within seconds after the earthquake and notify the areas that may be affected by the earthquake, which can effectively avoid casualties and property damage. However, traditional earthquake early warning systems need the building of a high-density seismic network, which costs a lot of money. In recent years, with the popularity of smartphones, the research on smartphone-based earthquake early warning(SEEW)systems has received widespread attention. The false alarm rate and missed detection rate of the SEEW system are always the tough issues in the study, and it is also the key to the actual application and deployment of the system. We use deep learning and statistical inference models to build a SEEW decision-making method. Firstly, a mobile phone seismic signal trigger model is established by introducing the attention mechanism into the long short term memory(LSTM)model; Secondly, a proportional operator of the triggered mobile phones based on temporal and spatial consistency is proposed, and a statistical inference-based earthquake discrimination model in the server is established. The performance of this method is tested through simulation platform. The results show that, compared with similar methods, the proposed method significantly improves the earthquake early warning rate under an extremely low false alarm rate.更多还原