基于多特征融合的脑电信号注意力识别  

Attention recognition based on multi-feature fusion of electroencephalogram signals

作　　者：徐彬 甘良志 栾声扬 李伟 XU Bin;GAN Liangzhi;LUAN Shengyang;LI Wei(School of Electrical Engineering and Automation,Jiangsu Normal University,Xuzhou 221116,China;Xuzhou Brain Signal Acquisition and Application Engineering Center,Xuzhou 221111)

机构地区：[1]江苏师范大学电气工程及自动化学院,徐州221116 [2]徐州市脑信号采集及应用工程中心,徐州221111

出　　处：《生物医学工程研究》2024年第6期462-467,共6页Journal Of Biomedical Engineering Research

基　　金：国家自然科学基金项目(61801197);江苏省自然科学基金项目(BK20181004);江苏高校青蓝工程;徐州市科技计划资助项目(KC22290);江苏师范大学研究生科研与实践创新计划资助项目(2024XKT0266)。

摘　　要：为充分利用脑电(electroencephalogram,EEG)信号特征,克服单一的特征提取方法无法全面表征EEG信号的局限,本研究提出了一种基于排列熵、模糊熵和平均能量的多特征融合算法。首先,利用经验模态分解去除EEG中的眼电、肌电伪迹;其次,对信号进行多特征提取,将特征融合后输入支持向量机(support vector machine,SVM)分类。通过对采集的40组EEG数据进行实验,结果显示,注意力集中和放松信号的识别准确率为90.53%。结果表明,多特征融合算法可增强特征表达能力,提高EEG信号的识别准确率。In order to make full use of electroencephalogram(EEG)signals features and overcome the limitation that a single feature extraction method can′t fully represent EEG information,we proposed a multi-feature fusion algorithm based on permutation entropy,fuzzy entropy and average energy.Firstly,the eye and myoelectric artifacts in EEG were removed by empirical mode decomposition.Secondly,the signals were extracted with multiple features,and the features were fused into the support vector machine(SVM)for classification 40 sets of collected EEG data were experimented and the results showed that the recognition accuracy of attention and relaxation signals reached 90.53%.The result indicates that the multi-feature fusion algorithm can enhance the feature expression ability and improve the accuracy of EEG recognition.

关 键 词：脑电信号 注意力识别 排列熵 平均能量 支持向量机 特征融合 

分 类 号：R318[医药卫生—生物医学工程] TP391[医药卫生—基础医学]