MULTI-FEATURE FUSION EMOTION RECOGNITION BASED ON RESTING EEG

• Published in: Journal of mechanics in medicine and biology Vol. 22; no. 3
• Main Authors: ZHANG, JUN-AN, GU, LIPING, CHEN, YONGQIANG, ZHU, GENG, OU, LANG, WANG, LIYAN, LI, XIAOOU, ZHONG, LICHANG
• Format: Journal Article
• Language: English
• Published: Singapore World Scientific Publishing Company 01.04.2022; World Scientific Publishing Co. Pte., Ltd
• Subjects: Cluster analysis; Clustering; Correlation coefficients; Electroencephalography; Emotion recognition; Emotions; Feature extraction; Feature recognition; Human-computer interface; Machine learning; Signal processing; Vector quantization; Multi-feature fusion; EEG signals; machine learning; cluster analysis; emotion recognition; resting-state
• ISSN: 0219-5194; 1793-6810
• DOI: 10.1142/S0219519422400024

An important task of brain–computer interface (BCI) research is to read or decode mental content from a large number of electroencephalographic (EEG) signals, which is an exciting, attractive, and challenging goal. In this paper, resting EEG data are collected from spontaneous EEG, and 16 random features such as correlation coefficient, covariance, and brainpower spectrum are extracted as reference vectors. In the subsequent emotion recognition experiment, the same feature information is extracted and separated from the EEG signal, and the translation and normalization processing are carried out based on the resting-state features. Finally, with the machine learning methods such as k -means clustering and multi-feature fusion, the positive, negative, and neutral emotional characteristic parameters were correctly separated. In a group of 12 subjects, the correct recognition rate of visual evoked positive, negative, and neutral emotions reached 83.9%, which was better than the literature mentioned in this paper. Another highlight of this method is that it can quickly, accurately, and efficiently select the number of features with the best matching and the least resource consumption from multiple features and multiple potential acquisition points. Further analysis and comparison of EEG characteristics can find the relationship between specific stimuli and corresponding EEG characteristic signals.