基于变尺度符号化补偿传递熵的EEG压力情感分析

基于脑电信号的情感识别,对于相关情感疾病的诊断与治疗有着重要的临床和科研意义。如何有效地提取特征,提高识别率,减少计算时间是研究的重点。从研究脑通道间定向信息交互的角度出发,结合对瞬时因果效应的补偿算法,提出以变尺度符号化补偿传递熵(VSSCTE)为特征的情感分析方法,并以此构建情感因效性脑网络,使用网络测度与ReliefF特征优化选择算法进行通道选择。结果显示,在使用DAEP数据集中处理后的127个压力状态和125个平静状态的数据时,较传统的二元传递熵方法,VSSCTE的特征提取方法在压力和平静情感二分类上的准确率提升约15%,达到96.74%;进行脑电通道优化后,当通道数由32个降至15个时,分类准确率仅下降约2%(分类准确率为94.36%),计算时间从51.27 s降至23.84 s。所提出的VSSCTE脑电特征提取方法可以有效分析情感变化时脑区之间的信息交互,为情感分析和计算提供新的思路。

EEG Stress Emotion Analysis Based on Variable-Scale Symbolic Compensation Transfer Entropy

Emotion recognition based on EEG signals has important clinical and scientific significance for the diagnosis and treatment of related emotional diseases. How to effectively extract features, improve recognition rate and reduce calculation time is the focus of this paper. From the perspective of studying the directional information interaction between brain channels, this paper combined the compensation algorithm for instantaneous causal effects and proposed an emotional analysis method of Variable-Scale Symbolic Compensation Transfer Entropy. This method was used to construct an emotional causal effect brain network, the network measurement and ReliefF feature optimization selection algorithm were used for channel selection. The results showed that the feature extraction method of VSSCTE improved the accuracy of emotion classification by about 15% to 96.74% over the conventional binary transfer entropy method when using data from the DAEP dataset of 127 stresses and 125 calms. After optimization of EEG channels, when the number of channels was reduced from 32 to 15, the classification accuracy rate only droped by about 2% (the classification accuracy rate was 94.36%), but the calculation time was reduced by about 110%. Overall, the VSSCTE method proposed in this paper was able to effectively analyze the information interaction between brain regions of different emotional states, providing a new method and ideas for emotional analysis.