Scholarlink

<p>Motor Imagery Brain-Computer Interfaces (MI-BCIs) employ machine learning to interpret brain activity patterns associated with imagined movements, translating them into commands for external devices. In recent years, Visibility Graph (VG) methods have been widely used for analyzing biological signals such as electroencephalography (EEG), demonstrating promising results. This paper introduces a novel MI-BCI system based on VG analysis. Our proposed method for classifying left- and right-hand motor imagery consists of two approaches: global and local. In the global approach, each EEG channel is transformed into a graph, and fifteen features are extracted. This graph-based transformation preserves the complex relationships within EEG signals, enhancing the feature extraction process. The features are then reduced using the Minimum Redundancy Maximum Relevance (mRMR) method, which improves computational efficiency and reduces overfitting risks. The classification is performed using Linear Discriminant Analysis (LDA), K-Nearest Neighbors (KNN), and Random Forest (RF). In the local approach, each trial is divided into three segments, and each segment is converted into a graph with fifteen extracted features. The mRMR method is again applied for feature selection, and classification is performed on each segment, with majority voting determining the final class. The results indicate that the proposed method outperforms the global approach, achieving mean accuracy, sensitivity, precision, and F1-score of 74.62%, 76.36%, 74.77%, and 74.74%, respectively, for the classification of motor imagery tasks involving left- and right-hand movements. A comparison with previous studies using the same dataset confirms the superior performance of the proposed method.</p>

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

Leveraging Segmentation and Visibility Graph Analysis To Enhance Motor Imagery Classification in EEG Signals

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月12日

查看全文
  • Fatemeh Mohammady,
  • Sekineh Asadi Amiri,
  • Zeynab Mohammadpoory

摘要

Motor Imagery Brain-Computer Interfaces (MI-BCIs) employ machine learning to interpret brain activity patterns associated with imagined movements, translating them into commands for external devices. In recent years, Visibility Graph (VG) methods have been widely used for analyzing biological signals such as electroencephalography (EEG), demonstrating promising results. This paper introduces a novel MI-BCI system based on VG analysis. Our proposed method for classifying left- and right-hand motor imagery consists of two approaches: global and local. In the global approach, each EEG channel is transformed into a graph, and fifteen features are extracted. This graph-based transformation preserves the complex relationships within EEG signals, enhancing the feature extraction process. The features are then reduced using the Minimum Redundancy Maximum Relevance (mRMR) method, which improves computational efficiency and reduces overfitting risks. The classification is performed using Linear Discriminant Analysis (LDA), K-Nearest Neighbors (KNN), and Random Forest (RF). In the local approach, each trial is divided into three segments, and each segment is converted into a graph with fifteen extracted features. The mRMR method is again applied for feature selection, and classification is performed on each segment, with majority voting determining the final class. The results indicate that the proposed method outperforms the global approach, achieving mean accuracy, sensitivity, precision, and F1-score of 74.62%, 76.36%, 74.77%, and 74.74%, respectively, for the classification of motor imagery tasks involving left- and right-hand movements. A comparison with previous studies using the same dataset confirms the superior performance of the proposed method.