《计算机应用》唯一官方网站 ›› 2023, Vol. 43 ›› Issue (11): 3418-3427.DOI: 10.11772/j.issn.1001-9081.2022101590

• 人工智能 • 上一篇    

基于通道选择和多维特征融合的脑电信号分类

杨淑莹(), 国海铭, 李欣   

  1. 天津理工大学 计算机科学与工程学院,天津 300384
  • 收稿日期:2022-10-24 修回日期:2023-01-31 接受日期:2023-01-31 发布日期:2023-04-12 出版日期:2023-11-10
  • 通讯作者: 杨淑莹
  • 作者简介:杨淑莹(1964—),女,四川成都人,教授,博士,主要研究方向:模式识别、智能机器人 yangshuying@email.tjut.edu.cn
    国海铭(1998—),男,河北衡水人,硕士研究生,主要研究方向:模式识别
    李欣(1998—),男,安徽六安人,硕士研究生,主要研究方向:模式识别。
  • 基金资助:
    2019年天津市教育科学规划院教学成果奖重点培育项目(PYGJ?015);2020年天津理工大学校级重点教学基金资助项目(ZD20?04)

EEG classification based on channel selection and multi-dimensional feature fusion

Shuying YANG(), Haiming GUO, Xin LI   

  1. School of Computer Science and Engineering,Tianjin University of Technology,Tianjin 300384,China
  • Received:2022-10-24 Revised:2023-01-31 Accepted:2023-01-31 Online:2023-04-12 Published:2023-11-10
  • Contact: Shuying YANG
  • About author:YANG Shuying, born in 1964, Ph. D., professor. Her research interests include pattern recognition, intelligent robots.
    GUO Haiming, born in 1998, M. S. candidate. His research interest includes pattern recognition.
    LI Xin, born in 1998, M. S. candidate. His research interest includes pattern recognition.
  • Supported by:
    2019 Key Cultivation Project of Teaching Achievement Award of Tianjin Institute of Educational Science Planning(PYGJ-015);School Level Teaching Fund of Tianjin University of Technology(ZD20-04)

摘要:

针对多通道脑电信号(EEG)相互干扰、存在个体差异性导致分类结果不同和单域特征识别率低等问题,提出一种通道选择和特征融合的方法。首先,对获取到的EEG进行预处理,使用梯度提升决策树(GBDT)选出重要通道;其次,采用广义预测控制(GPC)模型构建重要通道的预测信号,辨析多维相关信号之间的细微差别,再使用SE?TCNTA(Squeeze and Excitation block-Temporal Convolutional Network-Temporal Attention)模型提取不同帧之间的时序特征;然后,使用皮尔逊相关系数计算通道间的关系,提取EEG的频域特征和预测信号的控制量作为输入,建立空间图结构,并采用图卷积网络(GCN)提取频域、空域的特征;最后,将上述二者特征输入全连接层进行特征融合,实现EEG的分类。在公共数据集BCICIV_2a上的实验结果表明,在进行通道选择的情况下,与首个用于ERP检测的EEG-Inception模型以及同样采用双分支提取特征的DSCNN (Shallow Double-branch Convolutional Neural Network)模型方法相比,所提方法的分类准确率分别提升了1.47%和1.69%,Kappa值分别提升了1.25%和2.53%。所提方法能够提高EGG的分类精度,同时减少冗余数据对特征提取的影响,因此更适用于脑机接口(BCI)系统。

关键词: 脑电信号, 特征融合, 通道选择, 图卷积网络, 时序卷积网络, 广义预测控制模型

Abstract:

To solve the problems of the mutual interference of multi-channel ElectroEncephaloGraphy (EEG), the different classification results caused by individual differences, and the low recognition rate of single domain features, a method of channel selection and feature fusion was proposed. Firstly, the acquired EEG was preprocessed, and the important channels were selected by using Gradient Boosting Decision Tree (GBDT). Secondly, the Generalized Predictive Control (GPC) model was used to construct the prediction signals of important channels and distinguish the subtle differences among multi-dimensional correlation signals, then the SE-TCNTA (Squeeze and Excitation block-Temporal Convolutional Network-Temporal Attention) model was used to extract temporal features between different frames. Thirdly, the Pearson correlation coefficient was used to calculate the relationship between channels, the frequency domain features of EEG and the control values of prediction signals were extracted as inputs, the spatial graph structure was established, and the Graph Convolutional Network (GCN) was used to extract the features of frequency domain and spatial domain. Finally, the above two features were input to the fully connected layer for feature fusion in order to realize the classification of EEG. Experimental results on public dataset BCICIV_2a show that in the case of channel selection, compared with the first EEG-inception model for ERP detection and DSCNN (Shallow Double-branch Convolutional Neural Network) model that also uses double branch feature extraction, the proposed method has the classification accuracy increased by 1.47% and 1.69% respectively, and has the Kappa value increased by 1.25% and 2.53% respectively. The proposed method can improve the classification accuracy of EEG and reduce the influence of redundant data on feature extraction, so it is more suitable for Brain-Computer Interface (BCI) systems.

Key words: ElectroEncephaloGraphy (EEG), feature fusion, channel selection, Graph Convolution Network (GCN), Temporal Convolutional Network (TCN), Generalized Predictive Control (GPC) model

中图分类号: