摘要
目的探究与情绪密切相关的脑电特征,并藉此优化最少导联集合。方法 30名被试观看四类典型情绪图片(轻松、沮丧、愉悦、恐惧)并记录脑电信号。Fisher分数(Fscore)算法筛选每导联脑电特征,使用支持向量机方法(SVM)分类四种情绪。结果 24名被试情绪诱发有效,四个特征组合(β频带、γ频带、信息熵、微分熵)的F-score均值作为情绪有效性评价指标,筛选出分类准确率高达81.15%的5个导联(FT7、T7、FC4、TP10、O1)。结论利用校正后的F-score算法首次筛选出脑电信号的特征组合,获得与情绪密切相关的最优导联集合,极大地降低了运算时间,该结果对实现情绪的快速识别有重要价值。
Objective To find the least channels and to explore electroencepharologram(EEG)features which are closely related to emotion.Methods Thirty subjects were invited to watch four typical emotional pictures(relaxed,depressed,delightful and fearful)with their EEGs simultaneously recorded.EEG features were selected from each channel and the classification accuracy rates of four categories emotion were assessed by support vector machine(SVM)algorithm.Results 24 subjects were induced four categoriesemotion effectively.The average of the F-score ofβ-wave,γ-wave,information entropy and the differential entropy were used as evaluation indexes of emotion validity for each channel.The classification accuracy rate of the screened five channels(FT7,T7,FC4,TP10,O1)was 81.15%.Conclusion Corrected F-score algorithm is used to select the combination of features and the optimal channel set closely related to emotion,which can greatly reduce the computation time and have great value in realizing fast and real-time on-line recognition of emotion.
作者
李彤
王永宗
张艺耀
彭宏
朱玲玲
赵永岐
Li Tong;Wang Yongzong;Zhang Yiyao(Graduate School of Anhui Medical University,Hefei 230032;Laboratory of Military Cognitive Science and Stress Medicine,Military Institute of Cognition and Brian Sciences,Beijing 100850;Gansu Provincial Key Laboratory of Wearable Computing,School of Information Science and Engineering Lanzhou University, Lanzhou 730000;Dept of Clinical Examination,Air Force General Hospital of PLA,Beijing 100042)
出处
《安徽医科大学学报》
CAS
北大核心
2018年第10期1517-1521,共5页
Acta Universitatis Medicinalis Anhui
基金
军事医学科学院军事医学创新基金(编号：2015CXJJ011)
关键词
情绪
情绪识别
脑电信号
特征选取
F-score
emotion
emotion recognition
electroencepharologram
feature extraction
F-score
