【摘要】 地震时间序列的模型参数可以作为核爆地震信号识别的特征。现有基于ARMA(autoregressive moving average)模型参数的地震信号特征提取方法是一种线性方法,且只利用了信号的二阶统计信息,识别精度不高。为此,利用地震波的混沌特性,提出了一种核爆地震非线性特征提取方法:首先对地震波信号进行相空间重构,然后利用Volterra级数在重构的相空间内建立自适应预测模型,最后提取模型参数作为特征。在核爆地震分类实验中,非线性特征与线性特征相比,表现出更好的分类性能。研究结果表明:综合利用地震波信号的线性、非线性以及高阶统计信息对于核爆地震识别是非常重要的。更多还原

【Abstract】 ARMA parameters can be used as features for discrimination between nuclear explosions and earthquakes.Current methods for feature extraction are linear and only takes advantage of the second-order statistic,so the classification accuracy of is not high.To solve this problem,a method for extracting nonlinear features of nuclear explosions and earthquakes was proposed based on the chaotic feature of seismic waves.Firstly,the phase space of seismic waves was reconstructed.Secondly,the adaptive prediction model based on Volterra series was established in the phase space.Finally,the model parameters were taken as the features of seismic samples.In the classification experiment of of nuclear explosions and earthquakes,nonlinear features obtained better performance than linear features.Investigated results show that the combination of linear,nonlinear and higher-order statistical information is critical for the classification of nuclear explosions and earthquakes.更多还原