EEG Acquisition and Motor Imagery Classification for Robotic Control.

Area of Science:

• Neuroscience
• Robotics
• Machine Learning

Background:

• Brain-computer interfaces (BCIs) are increasingly used for controlling robotic systems via motor imagery.
• Minimally invasive electroencephalography (EEG) devices offer a pathway for practical BCI implementation.

Purpose of the Study:

• To validate the effectiveness of a portable, dry-electrode EEG device combined with machine learning for controlling robotic vehicle movements.
• To demonstrate the practical application of motor imagery-based BCI for robot control.

Main Methods:

• Acquired EEG signals from five participants using an 8-dry-electrode portable EEG device.
• Utilized sliding window segmentation and Common Spatial Pattern (CSP) for feature extraction.
• Implemented Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) for classification tasks (4-class and 2-class).

Main Results:

• Personalized models were developed for each participant.
• Binary classification tasks achieved a promising average accuracy of approximately 61%.
• Four-class classification tasks showed lower, less notable accuracy.

Conclusions:

• The study demonstrates the potential of dry-electrode EEG-based BCIs for robot control.
• Motor imagery classification with machine learning shows promise for practical robotic applications.
• Further research may improve accuracy in more complex BCI tasks.