关于脑电信号的情感优化识别仿真  

Emotion Optimization Identification Emulation for EEG Signal

作　　者：王薇蓉 张雪英[1] 孙颖[1] 畅江[1] WANG Wei - rong;ZHANG Xue - ying;SUN Ying;CHANG Jiang(College of Information Engineering, Taiyuan University of Technology, Taiyuan Shanxi 030024, Chin)

机构地区：[1]太原理工大学信息工程学院,山西太原030024

出　　处：《计算机仿真》2018年第6期426-431,共6页Computer Simulation

基　　金：国家自然科学基金(61371193)

摘　　要：针对支持向量机在脑电信号的情感识别中存在计算复杂度高、核函数选择具有局限性、计算时间长等问题,提出采用一种与支持向量机类似但不受核函数和惩罚因子限制的相关向量机算法对脑电信号的情感识别进行优化。设计情感脑电识别实验验证相关向量机在脑电信号中的情感识别性能,并与支持向量机的仿真结果进行分析比较。首先,选用实验室自主采集的情感脑电数据中的四种情感(悲伤、生气、惊奇和中性)作为实验数据来源。其次,提取脑电波中与情感相关性较大的β节律,并进一步提取其非线性特征(功率谱熵)作为分类器的输入特征向量,分别采用相关向量机和支持向量机进行了情感识别。仿真结果表明,相关向量机对情感脑电信号的识别率高于支持向量机,可以更有效地区分情感脑电信号,从而提高人类脑电信号的情感识别率。Considering the high computational complexity, the limitation of kernel function selection and the long computation time of super vector machine（SVM） in the emotion recognition of electroencephalogram（EEG） signal, this paper proposes a method, the relevance vector machine （RVM） algorithm, to optimize the emotion recognition of EEG signals, which is similar to SVM but not limited by kernel function and penalty factor. Emotional EEG signal recognition experiments were designed to verify the performance of the RVM, and the simulation results were compared with the SVM. First, four emotions （sadness, anger, surprise, and neutrality） in the EEG database collected by the laboratory were selected as experimental data sources. Secondly, the β band of the EEG signal was extracted and its power spectrum entropy was calculated. And then this entropy was used as the input feature vector of RVM and SVM for emotion recognition. The simulation results show that the emotional EEG recognition rate of relevance vector machine is higher than that of support vector machine, which can distinguish the emotional EEG signal more effectively and improve the emotional recognition rate of human EEG signals.

关 键 词：情感识别 脑电信号 相关向量机 支持向量机 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP399[自动化与计算机技术—控制科学与工程]