HIGHER ORDER SINGULAR SPECTRUM ANALYSIS OF EEG
- 物理技术－已发表论文 
奇异谱分析是脑电信号分析的一种新方法,脑电信号的奇异谱可以反映脑电的特征,它有助于研究大脑的动力学行为。奇异谱分析方法是基于二阶统计的方法,反映的是信号时间上和空间上的一种线性相关关系。而脑电信号属于非线性信号,其内在的非线性关系很难通过奇异谱得到真实的反映,从而会丢失某些有用的信息。提出一种新的基于高阶统计的脑电奇异谱分析方法,并将其运用于正常脑电和癫痫患者的脑电分析中。大量的实测信号样本仿真实验结果表明,正常脑电和癫痫脑电的奇异谱有明显的不同。此外,基于高阶统计的奇异谱和基于二阶统计的奇异谱相比更能反映出信号的细节。Singular spectrum analysis (SSA) is a novel way to analyse EEG, which can reflect the features of EEG and is helpful to study the dynamic behavior of human brain. SSA method is based on two order statistic, what the method reflects is the linear correlation on spacetime of signals. EEG is a kind of nonlinear signal, however, the intrinsic nonlinear correlations are difficult to reflect by SSA, so, some useful information will be lost. A novel SSA method is proposed, which is used for analysing EEG of normal subjects and epileptic subjects by a large number samples. The results of computer experiment show that the singular spectra of normal subjects are significantly different from epileptic subjects. More-over, the singular spectra based on higher order statistic is better for reflecting the details about EEG than singular spectra based on two order statistic.