Robotic Arm Control Using Machine Learning-Based EOG Signal Classifier

Abstract

Electrooculography (EOG) signals are a type of signal analyzed for detecting eye movements. In this project, features were extracted from EOG data collected from various subjects, and these features were then fed into machine learning algorithms k-Nearest Neighbor (KNN), Long-Short Term Memory (LSTM), Support Vector Machine (SVM), Decision Tree (DT)). The outputs of the algorithms were compared with findings from relevant literature studies. As a result, it was determined that the classification outcomes of the KNN algorithm were more successful compared to the other algorithms, with an achieved result of 0.9996. Consequently, appropriate parameters for the KNN algorithm were established using the extracted features from EOG signals. Using the determined parameters, the classification of new input values was performed, and the obtained results were transferred to the designed robotic arm, enabling the control of the arm. Consequently, successful outcomes were achieved and validated in the detection and classification of eye movements.