Analysis Of Eeg Signal And Sensitive Brain Region For Music Emotioms

Ageing population problem not only place great burden on family working-age adult,but also increase the number of senior people who live alone.The aggravation of this social phenomenon,resulting in the morbidity of sicknesses,such as neurodegenerative diseases(Alzheimer's disease)and mental illnesses(depression and anxiety)increased rapidly.These emotional relevant neurological diseases have brought great load to public health care and rehabilitation,while music therapy has become a nonpharmacological intervention potential treatment.In the music individual therapy,the treatment effect in different emotional state is with great differences.Therefore,it is the key issue to detect music emotional state effectively,and the most important thing is to recognize emotional state and explore sensitive brain area.In the analysis of emotional state,the key is to extract the effective and high correlation with emotion.In this paper,based on the emotional electroencephalogram(EEG)evoked by music,different features of six emotions(happy,exciting,mellow,melancholy,sad and terrible)were extracted: wavelet coefficient energy,wavelet entropy,and features of ?,?,? and ? rhythms,namely intrinsic mode function,approximate entropy,hurst exponent and lateralization features.The support vector machine(SVM)based on the particle swarm optimization(PSO)was used to assess the state of emotion.Contrasting recognition of different emotions' different features,showed that the classification result of single feature was not ideal,the average accuracy rate of which was only 50%,indicating the single feature could not fully reflect the specificity of emotion.If the feature combination was simple,it was possible to increase the redundancy.Based on principal component analysis(PCA),this paper proposed a method to select the feature parameters that were highly correlated with emotion.The research found that the classification result of the selected features was much superior than the combined feature,and its accuracy rate could improve 30% at lest.Most emotional sensitive brain regions researches are aiming at functional magnetic resonance imaging technique,which is high cost and inconvenient.To solve this problem,this paper proposed an emotion-sensitive brain area analysis method based on EEG.A total of 12 channel data were extracted from the prefrontal,frontal,frontal side,central,temporal and parietal regions.Comparison of the recognition result and floating of each channel,showed that the sensitive area of positive emotional were concentrated,mainly in the left frontal area.The sensitive area of negative emotional were more dispersed,and were in the left frontal area,left central,right parietal area,right and left frontal area,right temporal region.In order to show classification results of different brain area more intuitively,this paper based on Visual Studio 2010,SQL Server 2008 and MATLAB 2009 a,designed an emotional state assessment system.The system would realize an effective assessment of the music emotional state.