分段复合多尺度模糊熵和IGWO-SVM的脑电情感识别

为提高脑电的情感识别率，提出了分段复合多尺度模糊熵算法，采用分段粗粒化和计算复合多尺度模糊熵的策略，使提取特征较好地解决了数据缺失和计算不准确的问题；同时构造了应用余弦非线性收敛因子和动静态位置更新的灰狼算法优化支持向量机分类模型。为证明所提两种算法的有效性，进行了仿真实验验证，并在公开DEAP数据库下与几种常见的支持向量机优化模型比较脑电的情感识别率，结果表明在提出的模型下，效价、唤醒度、优势度、喜欢度的平均识别率分别为87.27%、87.81%、89.06%、87.58%，均高于其他算法。另外对比了高/低喜欢度下效价和唤醒度的分类，实验表明喜欢度低时情感识别率较高。

EEG emotion recognition based on piecewise complex multi-scale fuzzy entropy and IGWO-SVM algorithm

This paper used a piecewise complex multi-scale fuzzy entropy to increase the recognition rate. It proposed piecewise coarse graining and calculating complex multi-scale fuzzy entropy. It could solve the problem of missing data and calculating inaccurate entropy. This paper constructed an IGWO-SVM classification model which used cosine nonlinear convergence factor and dynamic and static position. To prove the validity of the algorithms, the paper performed the simulation experiment. Comparing the EEG emotion recognition rate with several common SVM optimization models in the public DEAP database, the results show that the average recognition rates of valance, arousal, dominance and liking are 87.27%, 87.81%, 89.06% and 87.58% in proposed model, which is higher than other algorithms. Comparing the classification of valance and arousal under high/low liking, experiments show that the recognition rate is higher when the liking is lower.