【摘要】 为实现对脑卒中疾病的高效分类预测，提出一种基于多维融合脑电特征的脑卒中分类预测方法。提出基于优化经验模态分解的多重分形去趋势波动分析算法，采用Pearson相关系数优化经验模态分解实现对脑电信号趋势项的选取，以解决多重分形去趋势波动分析中趋势项确定难、不连续等问题。基于分层模糊熵提出不对称熵特征和不对称熵指数，分析两类脑卒中脑电信号整体和局部熵值的差异性。对多维融合脑电特征进行脑卒中分类预测，结果表明，提出的多维融合脑电特征分类预测性能优异，准确率达到94.90%，特异性达到99.89%，表现出较强的脑卒中分类预测性能。更多还原

【Abstract】 To realize efficient classification prediction of stroke,a method of stroke classification prediction based on multi-dimensional fusion of EEG feature is proposed. An improved multifractal detrending fluctuation analysis algorithm based on optimized empirical mode decomposition is proposed.Pearson correlation coefficient is used to optimize the empirical mode decomposition to select the trend items of EEG signals,to solve the problems of difficult and discontinuous trend item determination in multifractal detrending fluctuation analysis. Based on hierarchical fuzzy entropy,the asymmetric entropy feature and asymmetric entropy index are proposed to analyze the differences between the global and local entropy values of EEG signals in two types of stroke. The multi-dimensional fusion of EEG feature is used for stroke classification. The results show that the proposed multi-dimensional fusion of EEG feature classification performance is excellent,with an accuracy of 94.90% and a specificity of 99.89%,showing a strong stroke classification performance.更多还原