Classifying EEG-Based Motor Imagery for Drone Control Using Artificial Neural Network

The progress in brain-computer interface (BCI) technology has created new possibilities for controlling external devices through electroencephalogram (EEG) signals. This study investigates the classification of EEG-based motor imagery (MI) signals for drone control using artificial neural networks (ANN). EEG data from six participants was collected using the Emotiv Epoc X device, focusing on MI related to drone movements. Feature extraction was performed using the discrete wavelet transform (DWT), followed by classification with a two-layer ANN with statistical parameters as input. The results indicate the significance of the ANN regression model in classifying motor imagery, although variations in model performance were observed across subjects. While overfitting was noted in one subject, indicating the need for validation sets, the models for other subjects showed good generalization. The ANN model achieved high accuracy in classifying movements “take-off” (98%) and “forward” (99%), but struggled with the “turn-left” (82%), suggesting the need for further model adjustments or additional data to improve generalization.