EEG-Based Brain-ComputerInterface for Robotic Arm Movement Using Machine Learning Techniques

Abstract

Brain-Computer Interface (BCI) is a communication pathway that connects a human being or an animal to a part of some external equipment, allowing control of the equipment using signals from the brain. In this research, Electroencephalography (EEG)  signals using motor imagery(MI)  related to hand movement are modeled for artificial limp movement. Hand engineered features are extracted using signal processing techniques and statistical methods. Furthermore, the features are modeled by machine Learning (ML) algorithms, such as  Decision Tree, Support-Vector Machine (SVM),  Kernel Support-Vector Machine, Logistic Regression, and Random Forest. Popular deep learning algorithm Convolutional Neural Network (CNN) is also trained with handcrafted features to identify the hand movements.  Our empirical results prove that the best average accuracy achieved is 87%, with CNN. The achieved k-fold accuracy of ML algorithms are between 70% to 72%.