多特征融合的脑电情绪分类

为进一步探究不同类型特征互补性对脑电情绪分类的影响，提出一种基于多特征融合的脑电情绪分类新方法。对预处理后的脑电信号进行DE、MST和SampEn特征提取，采用双样本T检验去除冗余筛选出最优特征并融合，采用SVM分类模型来识别不同的情绪状态。在SEED-Ⅳ数据集上的实验结果表明，单一特征中DE的平均分类准确率最高（77.86%），而融合非线性SampEn特征与功能连接MST属性后平均分类准确率得到进一步提升（84.58%），不同时间段采集的数据上重测实验则证明了该方法的有效性与稳定性。

EEG Emotion Classification Based on Multi-Feature Fusion

In order to further explore the influence of the complementarity of different types of features on EEG emotion classification, this paper proposes a new method of EEG emotion classification based on multi-feature fusion. First, it extracts the features of DE, MST and SampEn from the pre-processed EEG signals, then uses the sample t-test to remove the redundancy to screen out the optimal features and fusion them, and finally uses the SVM classification model to identify different emotional states. The experimental results on the SEED-IV dataset show that the average classification accuracy of DE in a single feature is the highest（77.86%）, and the average classification accuracy is further improved（84.58%） after fusing the nonlinear SampEn feature and the function to connect the MST attribute. The retest experiments on the data collected in different time periods have proved the effectiveness and stability of this method.