基于传递熵关键因果连接的情感识别方法

Emotion recognition method based on key causal connection of transfer entropy

人脑在情绪活动中呈现的信息流是复杂多变的,因此理解脑区间的动态交互过程至关重要,但是基于原始脑电信号构建的情绪网络包含了许多与情绪无关的冗余信息。针对此问题,提出一种在不丢失关键因果信息的前提下去除情绪无关网络连接的方法,并验证其在情感识别过程中的有效性。首先,基于传递熵因果分析方法对积极、中性和消极情绪构建归一化传递熵矩阵,再从积极、消极情绪矩阵中减去中性情绪矩阵,最后基于简化后的矩阵构建因效性脑网络并利用图论分析不同情绪的网络连通性。通过在DEAP数据集上的验证发现,该方法有效地提高了情感识别准确率。

The information flow presented by human brain in emotional activities is complex and changeable,so it is crucial to understand the dynamic interaction process of brain regions.However,using raw EEG signals to build emotional networks contains a lot of redundant information that has nothing to do with emotions.To solve this problem,this paper proposed a method to remove emotionally irrelevant network connections without losing key causal information,and verified its effectiveness in the process of emotion recognition.Firstly,this method used the transfer entropy causality analysis method to construct the norma-lized transfer entropy(NTE)matrix for positive,neutral,and negative emotions,and then subtracted the neutral emotion matrix from the matrix of positive and negative emotions.Finally,it used the simplified matrix to construct an effective brain network and used graph theory to analyze the network connectivity of different emotions.Through the verification on DEAP dataset,it is found that this method can effectively improve the accuracy of emotion recognition.