Abstract: This paper quantizes phase synchronism between two pairs of electrodes by phase locking value, and then constructs an incidence matrix. To find the significance nodes, the research extracts the area under the curve of degree and betweenness centrality in different sparsity of networks as features. In addition, this experiment input these features into a support vector machine to train some classified models. The result indicates that the local features of PLV functional network can effectively distinguish among different types of emotional EEG data, which may provide an effective method of emotional recognition based on EEG data.

Key words: emotional recognition, electroencephalograph, phase synchronism, complex networks

摘要： 使用相位锁值（Phase locking value，PLV）来量化任意两个电极通道之间的相位同步性，构建相应的脑功能网络的关联矩阵，提取网络不同稀疏度下的度、中间中心度两个局部属性的曲线下面积作为特征，对不同类型情感的网络特征进行非参数检验，找出显著性的节点。同时采用得到的特征值作为分类依据，训练SVM分类器。实验表明，利用PLV相位同步方法得到功能网络的局部属性，可以有效地区分不同类型的情感脑电数据，为基于脑电数据的情感识别提供了一种有效的方法。

关键词: 情感识别, 脑电图, 相位同步, 复杂网络