Related Experiment Video
Updated: Sep 13, 2025
06:37
Electroencephalography Network Indices as Biomarkers of Upper Limb Impairment in Chronic Stroke
Published on: July 14, 2023
976
Classification of finger movements through optimal EEG channel and feature selection
Murside Degirmenci1, Yilmaz Kemal Yuce2, Matjaž Perc3,4,5,6,7
1Kutahya Vocational School, Kutahya Health Sciences University, Kutahya, Türkiye.
Frontiers in Human Neuroscience
|July 31, 2025
Summary
This study enhances brain-computer interface (BCI) accuracy for predicting finger movements using electroencephalography (EEG) by analyzing optimal EEG channels and features, including the no mental task state.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Signal Processing
Background:
- Electrencephalography (EEG) is a key non-invasive neuroimaging technique for brain-computer interfaces (BCIs).
- Accurate prediction of individual finger movements using EEG-based BCIs is an emerging research area.
- Previous studies often excluded the no mental task (NoMT) state, potentially impacting prediction accuracy.
Purpose of the Study:
- To develop a practical system for predicting five finger movements and the NoMT state from EEG signals.
- To analyze the impact of statistical significance-based feature selection across various feature domains (time, frequency, time-frequency, nonlinear).
- To evaluate the performance of different classifiers using selected EEG channels and features.
Main Methods:
- Investigated 1102 EEG features from four domains and their combinations across 19 EEG channels.
- Employed a statistical significance-based feature selection method.
- Tested feature sets with eight prominent machine learning classifiers, including Support Vector Machine (SVM).
Main Results:
- Subject-dependent analysis achieved a maximum accuracy of 59.17% using SVM with selected features and all EEG channels.
- Subject-independent analysis yielded a maximum accuracy of 39.30% with SVM, utilizing specific feature subsets and EEG channels.
- Statistical feature reduction generally improved prediction performance across most classifiers.
Conclusions:
- The study demonstrates a modest but considerable advancement in finger movement prediction accuracy.
- Comprehensive analysis of EEG channels and features, along with feature selection, is crucial for improving BCI performance.
- Further research with larger, more diverse datasets is needed for generalizable conclusions.
