基于脑电信号的情绪复合特征优选与分类识别  

Emotion Complex Feature Optimization and SVM Classification Recognition based on EEG

作　　者：李忠高 蔡艳平 王涛 陈万 LI Zhong-gao;CAI Yan-ping;WANG Tao;CHEN Wan(Room 305,Rocket Force University of Engineering,Xi'an,Shaanxi 710025,China)

机构地区：[1]火箭军工程大学305室,陕西西安710025

出　　处：《计算机仿真》2025年第3期538-543,共6页Computer Simulation

摘　　要：鉴于情绪影响着人的工作和生活,准确的检测情绪具有重要应用价值,考虑到不同的情绪特征之间的互补性,提出一种基于脑电信号的情绪复合特征提取与识别方法。方法首先对原始信号进行ICA伪迹去除处理,对处理后的信号用小波变化分解为五个节律波,选取包含情绪信息较多的γ频段,提取微分熵、功率谱密度和模糊熵特征,然后对提取的特征的离群值使用最接近的非离群值进行填充,并用串行拼接进行融合,采用Fscore方法对复合特征进行优选,最后使用SVM进行情绪分类识别。在SEED-Ⅳ数据集上进行实验,结果表明所提方法能够更全面的包含情绪信息,有效提高识别精确度。In view of the influence of emotions on people's work and life,accurate detection of emotions is of great application value.Considering the complementarity between different emotional features,a method of emotional complex feature extraction and recognition based on EEG was proposed.Firstly,ICA artifact removal was carried out on the original signal,and the processed signal was decomposed into five rhythm waves by wavelet change.The differential entropy,power spectral density and fuzzy entropy features were extracted from the frequency band containing more emotional information.Then,the outlier values of the extracted features were filled with the closest non-outlier values,and the serial splicings were used for fusion.The Fscore method was used to optimize the composite features,and finally,an SVM was used for emotion classification recognition.Experiments were carried out on the SEED-Ⅳ data set,and the results showed that the proposed method could include emotional information more comprehensively and effectively improve recognition accuracy.

关 键 词：情绪 特征提取 微分熵 功率谱密度 模糊熵 

分 类 号：TN911.7[电子电信—通信与信息系统] R318[电子电信—信息与通信工程]