基于深度学习的脑电信号特征检测方法  

Method for EEG signal feature detection based on deep learning

作　　者：张成 汤璇[2] 杨冬平[2,3] 刘藤子 ZHANG Cheng;TANG Xuan;YANG Dongping;LIU Tengzi(School of Electrical Engineering and Automation,Fuzhou University,Fuzhou 350000,China;Quanzhou Institute of Equipment Manufacturing,Haixi Institutes,Chinese Academy of Sciences,Quanzhou 362200,China;Research Center for Human-Machine Augmented Intelligence,Research Institute of Artificial Intelligence,Zhejiang Lab,Hangzhou 311101,China)

机构地区：[1]福州大学电气工程与自动化学院,福建福州350000 [2]中国科学院海西研究院泉州装备制造研究中心,福建泉州362200 [3]之江实验室人工智能研究院混合增强智能研究中心,浙江杭州311101

出　　处：《电子设计工程》2024年第15期156-160,共5页Electronic Design Engineering

基　　金：国家自然科学基金面上项目(12175242)。

摘　　要：针对传统手工制作脑电(EEG)信号特征所带来的性能较差问题,提出了一种基于深度学习的脑电特征检测方法。采用EEG片段分割和带通滤波作为数据预处理方法,并结合卷积神经网络与Transformer自动地提取脑电信号特征,进而利用10折交叉验证训练模型以评估其有效性。同时以CHB-MIT数据集对所提出的方法进行验证,得到了该方法对癫痫脑电信号的分类平均准确率为98.51%,平均灵敏度为98.13%,平均特异性则为98.93%。实验结果表明,所提方法避免了繁琐的特征提取过程,能够有效完成脑电信号的自动检测及分类任务。To address the problem of poor performance associated with the traditional manual design of Electroencephalographic(EEG)signal features.A deep learning-based EEG feature detection method is proposed in the paper.EEG segmentation and band-pass filtering are used as data pre-processing methods,and convolutional neural networks and Transformer are combined to extract EEG signal features automatically,and a 10-fold cross-validation training model is used to evaluate the effectiveness of the model.The proposed method was also validated with the CHB-MIT dataset,achieving an average classification accuracy of 98.51%,an average sensitivity of 98.13%and an average specificity of 98.93%for epileptic EEG signals.The experimental results show that the proposed method avoids the tedious feature extraction process and can better perform the task of automatic detection and classification of EEG signals.

关 键 词：脑电信号 特征检测 卷积神经网络 TRANSFORMER 

分 类 号：TN911.7[电子电信—通信与信息系统] TP181[电子电信—信息与通信工程]