Feature Extraction of Emotional EEG Based on Fusion of EMD and CSP
EEG is a kind of non-linear and non-stationary time-varying signal.In order to extract effective features of emo-tional EEG,this paper proposes a fusion method of empirical mode decomposition(EMD)and common spatial pattern algorithm(CSP),and conducts two classification experiments of positive and negative emotions on open data sets deap and seed.Firstly,the optimal CSP feature parameters are selected through experiments.Secondly,IMFs are used as the input of CSP to extract emotion features with three dimensions of time frequency space,which are verified by support vector machine(SVM).Finally,the relation-ship between different brain regions and emotion is further analyzed.The experimental results show that m=5 and m=4 are the opti-mal CSP characteristic parameters of deap and seed respectively.And the low-frequency intrinsic mode function(IMF)component is easier to distinguish negative and positive emotions.The classification accuracy of EMD-CSP method on deap and seed data sets reaches 88.7%and 99.2%respectively.
EEGemotion recognitionempirical mode decompositioncommon space pattern algorithm
万方数据
维普
