长周期前驱波的检测及其局域均值能量提取与分类

Detection and Energy Feature Classification of Long-Period Precursor Wave with Local Mean Decomposition

地震预报迄今仍是一个世界性的科学难题,震前地壳内长周期形变前驱波检测的新方法与波形分析研究,可能成为未来地震预测的一个突破点.基于一种新颖的液体悬浮式长周期、超低频前驱波检测技术,实时连续记录了陕西省境内三个地震台站的超低频加速度信号;分析前驱波异常信号与地震发生的关系,发现异常出现与地震发生的关联度为51.59%;进一步利用局域均值分解算法对截取的前驱波异常信号进行分解,计算前五个分量的能量作为特征向量,输入支持向量机进行分类.结果表明,利用创新的前驱波检测技术和局域均值能量特征实现信号识别,异常前驱波均值能量特征与地震发生之间的关联度可达62.24%.

Accurate earthquake prediction still remains an unsolved problem.Detection and novel analysis methods of earth- quake precursor wave with long-period distortion may be a breakthrough for earthquake prediction in the future. Based on a new liq- uid-snspension detection technique, acceleration signals with super low frequency from three monitoring sites in Shaanxi Province were recorded consecutively;analysis of the relationship between the abnormal precursor waves and the occurrence of earthquake events shows that the correlation degree between the abnormal waves and the earthquake occurrences is 51.59% ;and then, local mean decomposition algorithm was further applied to analyze the precursor wave,the energy values of the first five components as the feature vectors were input into the support vector machine for classification. The results show that the correlation degree can reach as much as 62.24 %, because of the innovative precursor wave detection technique and the effective feature extraction of the local mean energy.