基于ECA和三分支卷积融合网络的脑电信号解码研究  

Research on EEG signal decoding based on ECA and three⁃branch convolutional fusion network

作　　者：周凯 艾尔肯·亥木都拉[1] ZHOU Kai;ARKIN Hamdulla(School of Mechanical Engineering,Xinjiang University,Urumqi 830017,China)

机构地区：[1]新疆大学机械工程学院,新疆乌鲁木齐830017

出　　处：《现代电子技术》2024年第23期89-97,共9页Modern Electronics Technique

基　　金：国家自然科学基金项目(10971177)。

摘　　要：基于运动想象脑电信号的脑机接口系统可以实现大脑和外部设备的交互,能够帮助残疾人控制辅助设备,提高他们的生活质量。然而,有限的脑电信号解码性能限制了脑机接口产业的大规模发展。文中提出一种基于高效通道注意力(ECA)和三分支卷积融合网络的ECA-TBCFNet模型用于基于脑电图的运动想象(MI-EEG)信号解码。ECA模块可自动捕捉脑电信号中的跨通道交互,三分支卷积融合网络能够多尺度地提取信号中的时空特征。ECA-TBCFNet模型在BCI竞赛IV-2a数据集上的四分类任务中取得了83.3%的准确率和0.78的kappa系数;此外,在Physionet MI-EEG数据集上两分类和四分类任务中,ECA-TBCFNet模型的准确率分别为87.87%和69.01%。结果表明,提出的ECA-TBCFNet模型可以有效提高运动想象脑电信号的识别准确率,并具有较高的鲁棒性。Brain⁃computer interface(BCI)systems based on electroencephalogram⁃based motor imagery(MI⁃EEG)signals can facilitate interaction between the brain and external devices,assisting disabled individuals in controlling assistive devices and thereby improving their quality of life.However,the limited decoding performance of EEG signals has hindered the development of the BCI industry.In view of this,an ECA⁃TBCFNet model based on efficient channel attention(ECA)and a three⁃branch convolutional fusion network for decoding MI⁃EEG signals is proposed.The ECA module can capture the cross⁃channel interactions in EEG signals automatically.The three⁃branch convolutional fusion network can extract spatio⁃temporal features from the signals at multiple scales.The ECA⁃TBCFNet model achieves accuracy rate of 83.3%and kappa coefficient of 0.78 in the four⁃class task on the dataset BCI Competition IV⁃2a.The ECA⁃TBCFNet model achieves accuracy rate of 87.87%and 69.01%in the two⁃class tasks and four⁃class tasks on the dataset Physionet MI⁃EEG,respectively.The results demonstrate that the proposed ECA⁃TBCFNet model can improve the recognition accuracy rate of MI⁃EEG signals effectively,and has high robustness.

关 键 词：脑机接口 脑电图 运动想象 高效通道注意力 三分支卷积神经网络 特征融合 

分 类 号：TN911.7-34[电子电信—通信与信息系统] TP391.9[电子电信—信息与通信工程]