Abstract

Info

Title:: Feature extraction method based on enhanced power spectral density for emotion analysis using EEG

Author(s):: LUO Gang1; WANG Mingxun2; LI Ming1; HUANG Min2; CHEN Hao1; 1. School of Information Engineering, Nanchang Hangkong University, Nanchang 330063, China 2. School of Music, Nanchang Hangkong University, Nanchang 330063, China

Keywords:: Keywords: enhanced power spectral density α frequency image feature feature fusion emotion analysis

Abstract:: Aiming at the problem that power spectral density (PSD) has single feature and cannot effectively represent the differences between frequencies in electroencephalogram (EEG)-based emotion analysis, a feature extraction method based on enhanced PSD is proposed to realize the analysis on emotions and the assessment of the significance of difference. After obtaining power spectral density image by α frequency power spectral density of EEG signal, the color feature, texture feature and similarity feature are extracted by image feature extraction algorithm. Then the redundant features are eliminated based on correlation criterion, and the final feature subset is obtained by taking the minimum average value of the significance of difference (P value) as the target, thus effectively fusing different image features. Finally, the emotions of subjects are analyzed, and the significance of the difference is assessed. The experimental results show that the proposed method can effectively quantify the emotional differences of the subjects in the SEED dataset. In the self-designed emotional EEG test, the significance of difference obtained by the proposed method is smaller than other methods, which proves the feasibility and effectiveness of the method.