脑电信号多重粗粒化复杂度分析方法研究

为了准确地描述脑电信号的精细结构和复杂成分，提出了基于多重粗粒化的脑电信号复杂度方法。根据脑电信号幅值微弱，幅值跨越大，一定幅值范围的脑电具有特定认知和生理意义的脑电特性，采用多重赋值，可尽量保留对信号复杂结构的描述；基于统计学特点(以百分比为阈值)对已经归一化的信号划分幅值域，进行粗粒化，基于传统方法计算复杂度。该方法有效地克服了经典LZC复杂度不能描述脑电精细结构的不足。提出了当脑电信号在粗粒化后出现局部周期，且达到一定阈值后(如beta波或具有相似幅值特性的脑电信号达到一定阈值)，4重复杂度值小于10重复杂度值的假设，初步建立了脑电节律与多重LZC的联系。仿真计算和实际数据验证表明，多重LZC方法结果更准确、合理，而且能反映脑电节律成分特性。

EEG Complexity Method Based on Multiple Coarse-Grained Sequences

In order to describe the fine structure and complexity of EEG components accurately, a EEG complexity method is proposed based on multiple coarse-grained EEG components. The coarse-grained thresholds of the method are selected based on the normalized statistics characters of the EEG. The complexity value is computed according to the formula of the classical LZC method. The simulation results based on the logistic map validate that the multiple LZC is much better than conventional methods and the trend of the LZC is much closer to the Lyapunov exponents. The method overcomes the defect of classical LZC complexity method, which cannot describe the fine structure of EEG characteristics effectively. Finally an assumption is proposed based on the results of the BCI IV EEG data, that is, the quadruple LZC is less than the decuple LZC when the EEG was partial periodicity and the period reached a certain threshold. The assumption is validated by the simulation data based on the EEG characters, a relationship between EEG rhythms and multiple LZC is established by which.