Electroencephalography Signal Analysis for Human Activities Classification: A Solution Based on Machine Learning and...

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Electroencephalography (EEG) measures brain electrical activity, crucial for understanding motor functions.
• Brain-Computer Interfaces (BCIs) leverage EEG signals for assistive technologies, particularly for individuals with motor impairments.
• Extracting meaningful patterns from complex EEG signals often requires advanced algorithms like machine learning (ML).

Purpose of the Study:

• To develop and evaluate a Random Forest (RF) based ML algorithm for classifying EEG signals during real and imagined motor activities.
• To assess the performance of the RF algorithm using both consumer-grade and research-grade EEG systems.
• To explore the potential of cognitive process-controlled tools through accurate EEG signal interpretation.

Main Methods:

• Implementation of a Random Forest (RF) machine learning algorithm.
• Classification of EEG signals recorded during real and imagery motor tasks.
• Evaluation of the RF algorithm's efficacy using data from both consumer and research-grade EEG devices.

Main Results:

• The Random Forest algorithm demonstrated efficiency in distinguishing between real and imagined motor activities.
• The algorithm could accurately identify the body part associated with the motor activity, even when using a consumer-grade EEG system.
• Interpersonal variability in EEG signals was identified as a factor negatively impacting classification accuracy.

Conclusions:

• The developed Random Forest ML model shows promise for BCI applications, enabling the classification of motor imagery and real motor execution.
• Consumer-grade EEG systems can be viable for BCI development with appropriate ML algorithms like Random Forest.
• Addressing interpersonal variability in EEG signals is essential for further improving BCI performance and reliability.