[1]曲思霖,王从庆,李建亮,等.基于小波变换共空间模式的脑电信号解码[J].郑州大学学报(工学版),2022,43(03):31-36.[doi:10.13705/j.issn.1671-6833.2021.06.003]
　QU Silin,WANG Congqing,LI Jianliang,et al.EEG Decoding Based on Wavelet Transform and Common Space Pattern[J].Journal of Zhengzhou University (Engineering Science),2022,43(03):31-36.[doi:10.13705/j.issn.1671-6833.2021.06.003]

点击复制

基于小波变换共空间模式的脑电信号解码()

分享到：

《郑州大学学报(工学版)》[ISSN:1671-6833/CN:41-1339/T]

卷:

43

期数:

2022年03期

页码:

31-36

栏目:

出版日期:

2022-04-10

文章信息/Info

Title:

EEG Decoding Based on Wavelet Transform and Common Space Pattern

作者:

曲思霖¹; 2; 王从庆¹; 2; 李建亮¹; 2; 展文豪²; 张 民¹

1.南京航空航天大学自动化学院;2.中国航天员科研训练中心人因工程国防科技重点实验室;

Author(s):

QU Silin¹; 2;  WANG Congqing¹; 2;  LI Jianliang¹; 2;  ZHAN Wenhao²;  ZHANG Min¹

1.School of Automation,Nanjing University of Aeronautics and Astronautics, Nanjing 210016，China; 

2.National Key Laboratory of Human Factors Engineering, China Astronauts Research and Training Center, Beijing 100094，China

关键词:

脑机接口; 脑电信号解码; 长短期记忆网络; 空间机械臂; 共空间模式

Keywords:

BCI;  EEG decoding;  long short term memory network;  space manipulator;  CSP

分类号:

TP391. 4

DOI:

10.13705/j.issn.1671-6833.2021.06.003

文献标志码:

A

摘要:

针对运动想象脑电信号实现任务少、识别准确率低等问题,提出了一种基于小波包分解的共空间模式脑电信号特征提取方法。该方法通过长短期记忆网络进行脑电信号解码,采用独立成分分析的方法将运动想象信号进行盲源分离,采用小波包分解方法将每个通道脑电信号按频率分为8组。计算每组信号的功率值,采用递归特性消除方法去除对分类不重要的10个节点特征,将被选择的节点信号采用1对1共空间模式提取空域特征,将特征矩阵输入长短期记忆网络进行脑电信号解码,得到4类运动想象信号分类结果。采用本文方法对公开的脑机接口竞赛数据集(包括左手想象信号、右手想象信号、舌头想象信号、双脚想象信号)前3位受试者数据进行验证,结果表明:本文方法的识别准确率分别为90.28%,94.25%、96.55% ,平均识别准确率达到93.69%。与其他方法对比,本文方法识别准确率较高。用识别的脑电信号作为解码控制信号,控制虚拟太空环境中的空间机械臂顺时针或逆时针运动,达到抓取空间中目标物体的目的。

Abstract:

Aiming at the problem of fewer tasks and low accuracy of recognition for motion imagination electroencephalogram(EEG)signals, in this paper a common space pattern(CSP)method based on wavelet packet decomposition(WPD)was proposed to extract the features of EEG signals.The long short term memory network was used to decode the EEG signals.Motor imagination signals were separated from blind sources by independent component analysis(ICA), and each channel EEG signal was divided into 8 groups by frequency using wavelet packet decomposition.The power value of each group of signals was calculated, 10 features were removed for classification by recurrence feature elimination(RFE).The selected signals were extracted by one-to-one common space pattern filters.The feature matrix were input into a long shortterm memory network for EEG decoding, and the classification results of 4 categories of motion imagination signals were obtained.The proposed method was used to verify the open data set of brain computer interface(BCI)competition(including four kinds of EEG signals: left hand imagination signal, right hand imagination signal, tongue imagination signal, and foot imagination signal).The recognition accuracy of three subjects was 90.28%, 94.25% and 96.55% respectively, and the average recognition accuracy could reach 93.69%.Compared with other feature extraction and classification methods, this method had a high classification accuracy.The decoded EEG signals were used to control the clockwise or counterclockwise movement of the space manipulator to achieve the purpose of grasping the target in the virtual space environment.

常用功能