Abstract: In view of the imperfection of the existing nonlinear characteristics of EEG signals which represent emotional information?, this paper introduces phase space reconstruction theory into the feature extraction of emotional EEG signals, three nonlinear geometric features of trajectory-based descriptive contours under the reconstructed phase space are extracted as the new emotional EEG characteristic parameters. Combining the power spectrum entropy and nonlinear attribute feature（approximate entropy, maximum Lyapunov exponent, Hurst exponent） of EEG signals, a nonlinear global feature（nonlinear geometric feature + nonlinear attribute feature） and power spectrum entropy fusion algorithm based on Principal Component Analysis（PCA） is proposed, Support Vector Machine（SVM） is employed to classify for emotion recognition. The results show that the nonlinear global feature can realize emotion recognition more effectively, and the rate of two-classification emotion recognition is about 90%. ?The fusion emotion feature based on PCA can achieve better emotion recognition performance than a single feature, and the average recognition rate can reach 86.42% in four classification experiments. The results show that the recognition rate of nonlinear global feature is higher than that of nonlinear attribute feature, and the combination of nonlinear global feature and power spectrum entropy can construct better emotional EEG feature parameters.

Key words: phase space reconstruction, nonlinear geometric features, nonlinear global feature, feature fusion

摘要： 针对现有表征情感信息的脑电信号的非线性特征提取不完善的问题，将相空间重构技术引入情感脑电的识别中，提取了在相空间重构下基于轨迹的描述轮廓的三种非线性几何特征作为新的情感脑电特征。结合脑电信号的功率谱熵以及非线性属性特征（近似熵、最大Lyapunov指数、Hurst指数），提出了基于主成分分析（PCA）的非线性全局特征（非线性几何特征+非线性属性特征）和功率谱熵的融合算法，以支持向量机（SVM）为分类器进行情感识别。结果显示，非线性全局特征能更有效地实现情感识别，二分类情感识别率约90%左右。基于PCA的融合情感特征相比单一特征能达到更佳的情感识别性能，四分类实验中平均识别率可达86.42%。结果表明，非线性全局特征相比非线性属性特征情感识别率有所提高，非线性全局特征以及功率谱熵的结合可以构造出更佳的情感脑电特征参数。

关键词: 相空间重构, 非线性几何特征, 非线性全局特征, 特征融合