Scholarlink

In recent years, electroencephalography (EEG) has been extensively utilized to study brain activation across various cognitive tasks. This study aimed to implement an acquisition protocol alongside an analysis framework to accurately estimate participants’ cognitive workload and the task in which they were engaged. Five participants were participated, using a 20-channel dry-electrode EEG device while performing three distinct cognitive tasks: a Baseline task involving verbal counting, a Language task involving Verbal Fluency Tests, and a Reasoning task using Raven’s Advanced Progressive Matrices Test. Multiple features from individual EEG channels accurately discriminated between the Baseline vs. Language and Baseline vs. Reasoning tasks, while a combination of different features from multiple EEG channels was used to distinguish the Language vs. Reasoning task. The dominant features in the analysis were the Band Power Beta and Theta, Mean Teager Energy, Arithmetic Mean, Median Value, and Maximum Value. The proposed analysis framework efficiently discriminated different workloads, achieving up to 96% accuracy in each two-class classification problem. This resulted in an efficient model for estimating participants’ cognitive workload and accurately identifying the task they were engaged in, showing the potential for future real-time cognitive workload monitoring systems.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A Machine Learning Framework for Automatic Cognitive Task Classification Using Dry Electrode EEG Data

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Eleftherios Kontopodis,
  • Christodoulos Serafeim,
  • Dionisis Cavouras,
  • Ioannis Kalatzis,
  • Errikos Ventouras,
  • Ioannis Kakkos,
  • Aikaterini Skouroliakou

摘要

In recent years, electroencephalography (EEG) has been extensively utilized to study brain activation across various cognitive tasks. This study aimed to implement an acquisition protocol alongside an analysis framework to accurately estimate participants’ cognitive workload and the task in which they were engaged. Five participants were participated, using a 20-channel dry-electrode EEG device while performing three distinct cognitive tasks: a Baseline task involving verbal counting, a Language task involving Verbal Fluency Tests, and a Reasoning task using Raven’s Advanced Progressive Matrices Test. Multiple features from individual EEG channels accurately discriminated between the Baseline vs. Language and Baseline vs. Reasoning tasks, while a combination of different features from multiple EEG channels was used to distinguish the Language vs. Reasoning task. The dominant features in the analysis were the Band Power Beta and Theta, Mean Teager Energy, Arithmetic Mean, Median Value, and Maximum Value. The proposed analysis framework efficiently discriminated different workloads, achieving up to 96% accuracy in each two-class classification problem. This resulted in an efficient model for estimating participants’ cognitive workload and accurately identifying the task they were engaged in, showing the potential for future real-time cognitive workload monitoring systems.