A Study on Construction of ELF Band Electromagnetic Field Measuring System and Application to Tectonic Activity Detection

(1)建立相控阵网络以观测电磁辐射：我们增加了观测点的数量。我们实现了数字滤波器，以缩小观测频带。通过GPS实现了数据的准确时间戳。(2)高灵敏度磁场传感器和高增益滤波放大器的开发：我们减小了传感器的尺寸和重量，并将其置于耐候外壳中，以便于设置和稳定。(3)相敏检波器与GPS同步及数据采集装置的研制：研制了基于GPS的定时同步电路。本文实现了基于实时Linux的数据采集系统。(4)基于统计模型和声学特征的地壳电磁辐射探测系统的研制：提出了监测线性预报误差的自动探测方法。从福冈西部地震发生前后的辐射异常中，我们得到了地震前兆存在的信息。(5)异常电磁辐射源估计及概率模型的探讨：以倒谱系数为特征参数，通过GTM对观测数据进行分类，证实了观测数据与地图位置的一致性。(6)滤波方法和异常辐射探测方法的研究：提出了利用绝对误差的TPE、伊卡和NMF的背景噪声抑制方法。每种方法都能很好地去除全局背景噪声。(7)神经网络信号分析：利用小波变换后的观测数据作为参考，建立神经网络。我们确认了我们的神经网络可以检测到异常信号。(8)电磁场分析：通过监测磁场方位角的变化，可以自动检测异常电磁辐射。

(1) Construction of phased array network to observe EM radiation: We increased the number of observation sites. We implemented digital filter to narrow down observing frequency band. We achieved accurate timestamp of data by GPS.(2) Development of the high-sensitive magnetic field sensor and high-gain filtering amplifier: We reduced size and weight of the sensor, and set it in the weather-resistant case for easy setting and stability.(3) Development of phase sensitive detector with GPS synchronization and data logging device: We developed the timing synchronous circuit using GPS. We achieve data logging system based on real-time Linux.(4) Development of detection system for electromagnetic radiation from crust by statistical model and acoustic signature: We proposed automatically detection method monitoring on liner prediction errors. We got knowledge that indicates the existence of earthquake-precursor from detected anomalous radiations around occurrence of west-off Fukuoka earthquake.(5) Source Estimation of anomaly electromagnetic radiation and deliberation of probability model: We confirmed the observed data are classified with map position by GTM when using cepstrum coefficients as feature parameter.(6) Deliberation of filtering method and deliberation of anomalous radiation detection method: We proposed background noise reduction methods by TPE using absolute errors, ICA and NMF. Each method removes global background noise well.(7) Signal analysis by neural network: We made neural net using wavelet-transformed observed data as reference. We confirmed our neural net could detect anomalous signal.(8) Analysis of electromagnetic field: We found out anomalous electromagnetic radiation can be detected automatically by monitoring the variance of magnetic field azimuth.