基于脑网络的多特征融合情感识别方法

Emotional recognition method for multi-feature fusion based on brain network

为提高情感识别准确率,通过结合脑电信号和脑功能网络,提取功率谱密度、差分熵、样本熵特征,将网络模块度特征引入情感识别,分别从单一特征和融合特征的角度分析情感分类的准确率。结果表明,相比于三种单一特征,融合网络模块度特征后的分类准确率在各频段均有提高,分类准确率最高的Gamma频段可达到90.92%,融合特征后的分类准确率比单一特征提高了7.64%。

This paper aims to improve the accuracy of emotion recognition.The study involves combining EEG signals with brain functional network;extracting power spectral density,differential entropy,and sample entropy feature;introducing the network modularity feature into emotion recognition;and analyzing the accuracy of emotion classification from the perspective of single feature and fused feature.The results show that the classification accuracy after fusing network modularity feature is improved in all frequency bands among the three singal features,and the highest classification accuracy can reach 90.92%in Gamma band;and the classification accuracy after feature fusion is 7.64%higher than that of single feature.