井冈山大学学报（自然科学版）2026，Vol.47Issue(3)：67-78,12.DOI:10.3969/j.issn.1674-8085.2026.03.009
基于DE-MI-STFT融合特征的情绪识别方法
Emotion recognition method based on DE-MI-STFT fusion features
摘要
Abstract
The application of EEG to identify the level of pleasure and excitement of patients is of great significance for the depression diagnosis.The current research on electroencephalogram(EEG)emotion recognition has some problems such as insufficient extraction of signal features from different brain regions and low accuracy of three-classification due to single-dimensional feature modeling.To solve the above problems,this paper proposes an EEG emotion recognition method based on DE-MI-STFT fusion features:differential entropy(DE),mutual information(MI),and short-time Fourier transform(STFT)features are extracted to construct a multi-dimensional set,which is separately modeled and trained through a deptwise separable neural networks,and then the multi-model decision results are fused via D-S evidence theory to fully utilize the complementary information of features.Three-classification experiments on the DEAP public dataset show that the classification accuracy of the proposed method in the valence and arousal dimensions is 14.3%and 13.6%higher than that of the existing optimal models,respectively.The study indicates that the proposed method can effectively capture the spatiotemporal features of EEG,significantly improve the three-classification accuracy through the fusion of multiple features and multiple models,and provide technical support for the auxiliary diagnosis of emotion-related diseases.关键词
EEG/情绪识别/D-S证据合成理论Key words
electroencephalogram/emotion recognition/D-S evidence theory分类
信息技术与安全科学引用本文复制引用
刘云凤,张雪茹,周全,万思佳,赵转哲,刘永明..基于DE-MI-STFT融合特征的情绪识别方法[J].井冈山大学学报（自然科学版）,2026,47(3):67-78,12.基金项目
安徽省高等学校科研计划项目(2022AH050995) （2022AH050995）
安徽省重点实验室开放基金项目(DQKJ202410,APELDE2023A005) （DQKJ202410,APELDE2023A005）
芜湖市重大科技成果工程化项目(WJ-KG-CG-202403002) （WJ-KG-CG-202403002）
