摘要
针对地震地下流体观测数据易受观测系统、自然环境、场地环境、人为调试和不明原因等因素干扰,本文提出一种基于神经网络的地下流体观测数据干扰抑制的期望最大化算法。该方法利用噪声和信号幅度之间的统计特性差异来构建检测标准,并将包含信号和噪声的两个统计分布的混合物组成的模型拟合到数据中,然后通过期望最大化用于估计并得出数据中每个样本是否有噪声的概率(即可能性)。只有那些可能性大于特定阈值的样本才被认为是噪声。结合湖北地震台地下流体观测数据实例进行仿真实验,并与传统神经网络和小波分析算法对比。结果表明该算法能起到更好抑制噪声的目的,具有明显的优越性。
Aiming at the fact that the seismic observation data of underground fluid is easily disturbed by factors such as observation system,natural environment,site environment,man-made debugging and unknown reasons,this paper proposes an expectation maximization algorithm based on neural network for interference suppression of underground fluid observation data.This method uses the statistical characteristic difference between noise and signal amplitude to construct detection standards,and fits a model consisting of a mixture of two statistical distributions of signal and noise to the data,and then uses expectation maximization to estimate and obtain the probability(that is,the probability)of whether each sample in the data has noise.Only those samples whose probability is greater than a certain threshold are considered noise.The simulation experiment is carried out based on examples of underground fluid observation data from Hubei Seismic Station,and is compared with traditional neural network and wavelet analysis algorithms.The results show that the algorithm can achieve a better purpose of suppressing noise and has obvious advantages.
作者
周洋
王俊
ZHOU Yang;WANG Jun(Institute of Seismology,China Seismological Bureau(Key Laboratory of Seismological Geodesy),Wuhan Hubei 430071,China;Earthquake Administration of Hubei Province,Wuhan Hubei 430071,China;School of Computer Science,South-Central University for Nationalities,Wuhan Hubei 430071,China)
出处
《电子器件》
CAS
北大核心
2021年第5期1215-1220,共6页
Chinese Journal of Electron Devices
基金
中央高校基本科研业务费专项项目(CZT20015)
湖北省自然科学基金项目(2019CFB815)
中国地震局地震研究所和地壳应力研究所基本科研业务费专项项目(6292-6)。
关键词
地下流体
抑制噪声
神经网络
小波分析
期望最大化
underground fluid
noise suppression
neural network
wavelet analysis
expectation maximization
