现代电子技术2025，Vol.48Issue(19)：92-98,7.DOI:10.16652/j.issn.1004-373x.2025.19.015
基于CAPTCNet的运动想象脑电分类
Motor imagery EEG classification based on CAPTCNet
摘要
Abstract
This paper proposes a hybrid feature extraction and multi-scale parallel temporal convolutional network(CAPTCNet)based on coordinate attention,so as to fully utilize the various feature information in the motor imagery electroencephalogram(MI-EEG)signals and improve the decoding accuracy of MI-EEG signals.The network mainly consists of three modules.The hybrid feature extraction module extracts shallow temporal and spatial features from MI-EEG signals by multi-layer convolution.The coordinate attention module highlights the most valuable information in MI-EEG signals.The parallel multi-scale temporal convolutional module is used to extract deeper temporal features at different scales.The average accuracy rate of the designed network model tested on the dataset BCI Competition Ⅳ-2a was 84.14%,with an average Kappa coefficient of 0.79.The experimental results show that the proposed method has higher classification accuracy rate and more satisfactory Kappa coefficient in comparison with the other methods.The ablation experiment further demonstrated the effectiveness of each module,among which the parallel multi-scale temporal convolutional module contributed the most to the improvement of the model performance.关键词
脑机接口/运动想象/卷积神经网络/时间卷积网络/深度学习/注意力机制Key words
brain-computer interface/motor imagery/convolutional neural network/temporal convolutional network/deep learning/attention mechanism分类
信息技术与安全科学引用本文复制引用
高洪睿,毕峰..基于CAPTCNet的运动想象脑电分类[J].现代电子技术,2025,48(19):92-98,7.基金项目
辽宁省教育厅高校基本科研项目(LJKMZ20221756) （LJKMZ20221756）
