基于微分熵与深度残差网络的脑电信号情感识别  

EMOTION RECOGNITION OF EEG SIGNAL BASED ON DIFFERENTIAL ENTROPY AND DEEP RESIDUAL NETWORK

作　　者：杜秀丽[1,2] 马振倩 郭庆汝 邱少明 吕亚娜[1,2] Du Xiuli;Ma Zhenqian;Guo Qingru;Qiu Shaoming;LüYana(Key Laboratory of Communication and Network,Dalian University,Dalian 116622,Liaoning,China;College of Information Engineering,Dalian University,Dalian 116622,Liaoning,China)

机构地区：[1]大连大学通信和网络重点实验室,辽宁大连116622 [2]大连大学信息工程学院,辽宁大连116622

出　　处：《计算机应用与软件》2023年第6期160-165,共6页Computer Applications and Software

基　　金：辽宁省“百千万人才工程”项目(2018921080)。

摘　　要：现有的脑电信号情感识别方法大多是挑选出与情感变化相关度较大的几个单导联脑电信号进行特征提取与选择。针对脑电信号情感识别中没有考虑到导联间存在的整体空间拓扑结构问题,提出一种基于微分熵与深度残差网络的识别方法。该方法将全部导联脑电信号作为一个整体,把各个频带(全频段、γ段、β段和α段)的微分熵特征按照相应的电极空间位置、频段顺序映射为脑电信号微分熵二维特征;利用深度残差网络实现二维特征的自动提取,以充分利用各个导联的脑电信号信息,挖掘导联间隐匿的空间拓扑结构特征。在国际公开数据集SEED上的仿真实验结果表明,该方法的识别平均准确率可以达到95.10%。Most of the existing EEG signal emotion recognition methods select several single-lead EEG signals that have a greater correlation with emotional changes for feature extraction and selection.Aimed at the problem of EEG signal emotion recognition without considering the overall spatial topological structure between leads,a recognition method based on differential entropy and deep residual network is proposed.This method took all lead EEG signals as a whole,and mapped the differential entropy characteristics of each frequency band(full frequency band,γsegment,βsegment,andαsegment)into two-dimensional features of EEG signal differential entropy according to the corresponding electrode spatial position and frequency sequence.The deep residual network was used to realize the automatic extraction of two-dimensional features,so as to make full use of the EEG signal information of each lead and mine the hidden spatial topological structure features between the leads.The simulation experiment results on the international public dataset SEED show that the average recognition accuracy of this method can reach 95.10%.

关 键 词：脑电信号 情感识别 微分熵 深度残差网络 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]