基于增强深度特征的TSK模糊分类器用于癫痫脑电信号识别
作者:
作者单位:

1.湖州师范学院信息工程学院;2.湖州师范学院

作者简介:

通讯作者:

中图分类号:

TP181

基金项目:

国家自然科学基金面上项目61771193,61802123,61772199


TSK fuzzy classifier based on enhanced deep feature for epilepsy EEG signal recognition
Author:
Affiliation:

1.Huzhou university;2.fsfa

Fund Project:

The National Natural Science Foundation of China under Grants 61771193,61802123,61772199

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    摘 要: 识别癫痫脑电信号的关键在于获取有效的特征和构建可解释的分类器。为此,本文提出了一种基于增强深度特征的TSK模糊分类器(ED-TSK-FC)。首先,ED-TSK-FC使用一维卷积神经网络(1D-CNN)自动获取癫痫脑电信号的深度特征与潜在类别信息,并将深度特征和潜在类别信息合并为增强深度特征;然后,将增强深度特征作为ED-TSK-FC模糊规则前件与后件部分的训练变量,保证了原始输入的深度特征及其潜在意义都出现在模糊规则中,进而对增强深度特征作出了良好的解释;此外,在不显著降低分类准确度的情况下,ED-TSKFC的廉价训练方法可以缩短模型的训练时间;最后,在Bonn癫痫数据集上,分别从分类性能、学习效率和可解释性三方面,验证了ED-TSK-FC的优势。

    Abstract:

    Abstract: The key to recognize epileptic EEG signals is to obtain effective features and construct an interpretable classifier. In this paper, a novel Takagi-Sugeno-Kang (TSK) fuzzy classifier based on enhanced deep feature (ED-TSKFC) for epilepsy EEG signal recognition is proposed. ED-TSK-FC adopts an one-dimension convolution neural network (1D-CNN) to extract deep feature and label information, label information is expanded to deep feature space to generate enhanced deep feature; As the antecedent-consequent variable of ED-TSK-FC, enhanced deep feature can provide deep feature and label information for each fuzzy rule, hence fuzzy rules open the black box of 1D-CNN in an interpretable manner; In addition, ED-TSK-FC also formulates a cheap strategy to decrease training time; Finally, experiments show that this method provides an excellent performance on the Bonn Epilepsy Dataset in terms of classification performance, training time and interpretability.

    参考文献
    相似文献
    引证文献
引用本文
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2021-06-12
  • 最后修改日期:2022-08-12
  • 录用日期:2021-09-28
  • 在线发布日期: 2021-11-01
  • 出版日期: