基于多尺度符号转移熵的脑电信号分析

无论从全局还是局部的角度出发,采用多尺度转移熵表示全局和局部两类脑电(electroencephalography,EEG)信号,并分析其动态和不对称信息。采用比例系数从1到199、步长为2的多尺度方法处理正常人和癫痫病患者的脑电信号;然后采用维度为3的全局排列方法表示序列。将正向和反向符号序列作为转移熵的输入。比例因子的间隔和全局路径分别为(37,57)和(65,85),分析发现两组EEG信号的熵值在该处较容易区分。当比例系数为67时,健康对照组和癫痫病患者的转移熵值分别为0.113 7和0.102 8,差异最大。在比例系数是165时,全局变量的相应值为0.064 1和0.060 1。研究结果表明,合适的排列有助于更好的区分脑电数据信息,采用多尺度符号转移熵分析EEG信号更加有效。

EEG Analysis based on Multi-scale symbolic transfer entropy

From both global and local perspectives, we symbolize two kinds of EEG and analyze their dynamic and asymmetrical information using multi-scale transfer entropy. Multi-scale process with scale factor from 1 to 199 and step size of 2 is applied to EEG of healthy people and epileptic patients, and then the permutation with embedding dimension of 3 and global approach are used to symbolize the sequences. The forward and reverse symbol sequences are taken as the inputs of transfer entropy. Scale factor intervals of permutation and global way are (37, 57) and (65, 85) where the two kinds of EEG have satisfied entropy distinctions. When scale factor is 67, transfer entropy of the healthy and epileptic subjects of permutation, 0.1137 and 0.1028, have biggest difference. And the corresponding values of the global symbolization is 0.0641 and 0.0601 which lies in the scale factor of 165. Research results show that permutation which takes contribution of local information has better distinction and is more effectively applied to our multi-scale transfer entropy analysis of EEG.