基于动态功能连接的握力运动参数脑电识别

EEG Recognition of Grip Motion Parameters Based on Dynamic Functional Connectivity

为了研究握力运动过程中脑功能网络(Brain Functional Network,BFN)的动态变化特征及对握力运动参数任务识别的影响,提出一种加权相位滞后指数法构建动态脑功能网络。通过对预处理的EEG时间序列按照一定的非重叠窗口截断成等长的子时间序列,按时间顺序,利用相位同步特性对所有子序列进行筛选;筛选出的有效数据段用加权相位滞后指数法估计功能连接大小,构建相应节点的脑功能网络;提取4种不同网络特征参数,对其进行串行融合获得特征向量;最后以支持向量机作为分类器进行分类。与静息态和不均匀子时段划分方法相比,所提方法的平均识别率提高到了74%。

In order to study the dynamic change characteristics of Brain Functional Network(BFN)during grip exercise and its influence on the task recognition of grip exercise parameters,this paper proposes a weighted phase lag index method to construct a dynamic brain functional network.By truncating the pre-processed EEG time series into equal length sub-series according to a certain nonoverlapping window,all the sub-series are screened in time order by using the phase synchronization property;the screened valid data segments are estimated by the weighted phase lag index method to estimate the functional connection size and construct the brain functional network of corresponding nodes;the four different network feature parameters are extracted and serially fused to obtain the feature vectors.Finally,a support vector machine is used as a classifier for classification.Compared with the resting-state and inhomogeneous sub-time segmentation methods,the average recognition rate of the method in this paper is improved to 74%.