摘要
为了更加精准地实现对癫痫脑电分类,提出一种最优通道获取的方法,采用频率切片小波变换时频分析,并结合InceptionNet-V3卷积神经网络进行迁移学习分类,采用10折交叉验证训练模型并评估分类效果,实现对癫痫患者分类准确率最高为95.8%,敏感度为100%。研究结果表明,所提方法可有效地提取出癫痫脑电的特征并进行分类,避免了繁琐的特征提取过程,而迁移学习又避免了机器学习模型需要大量数据进行训练的问题,节省了特征提取和模型训练的时间,实现了高质量的分类效果。
In order to achieve more accurate classification of epileptic EEG,an optimal channel acquisition method is proposed,using frequency slicing wavelet transform time-frequency analysis and combining InceptionNet-V3 convolutional neural network for migration learning classification,using 10-fold crossvalidation training model and evaluating the classification effect,achieving the highest classification accuracy of 95.8% for epileptic patients and sensitivity of 100%. The results show that the proposed method can effectively extract and classify the features of epileptic EEG,avoiding the tedious feature extraction process,while the migration learning avoids the problem of machine learning models requiring large amounts of data for training,saving the time for feature extraction and model training,and achieving high-quality classification results.
作者
杨彬
杨晓利
李振伟
吴晓琴
韩家祺
YANG Bin;YANG Xiaoli;LI Zhenwei;WU Xiaoqin;HAN Jiaqi(School of Medical Technology and Engineering,Henan University of Science and Technology,Luoyang 471000,China)
出处
《电子设计工程》
2023年第3期22-27,共6页
Electronic Design Engineering
基金
河南省重点研发与推广专项(202102310534)
河南省高等学校重点科研项目(20A416002)。
关键词
癫痫
分类
最优通道
迁移学习
epilepsy
classification
optimal channel
transfer learning
