چكيده به لاتين
This research investigates a brain-computer interface (BCI) system based on EEG signals implemented in a virtual reality (VR) environment to enhance motor rehabilitation.
The primary objective of this study is to translate motor imagery into real-time interactive commands within a VR environment. In the initial phase, the training model is designed and saved. For this purpose, the BCI Competition IV 2a dataset was used, which consists of EEG signals collected from humans. Necessary preprocessing steps, including band-pass filtering and removal of physiological and non-physiological artifacts, were performed on the data. Extracted features were classified using SVM and Logistic Regression methods.
The modeling accuracy achieved in this study was 72% for the SVM classifier and 77% for the Logistic Regression classifier, showing an improvement in classification accuracy compared to previous studies. Additionally, parameter optimization for the SVM model was performed using GridSearchCV, yielding the best values of C = 0.1, γ = 1, and kernel 'rbf'. The Logistic Regression model was configured with parameters such as max_iter = 100, warm_start = True, and n_jobs = 6, enabling continuous updates with new data.
In the second phase, the process of real-time interaction between the user and the virtual reality (VR) environment was designed and implemented. The VR environment in this study was developed using the Unity game engine, featuring a motivational and cheerful space with natural elements such as trees and greenery along the path. Train wagons were used as obstacles. Upon receiving requests from the VR execution system via the UDP protocol on port 8000, the detected commands (Left, Right, No Motor Imagery) were sent to the VR environment, allowing the user to begin interacting and moving within the VR space.
During the user's activity in VR, metrics such as collected coins, elapsed time, and the number of commands (Non-MI, Right, Left) were recorded. If the user collided with obstacles, a "Game Over" message was displayed, while reaching the finish line triggered a "Congratulations" message. Therefore, the final results for each execution cycle included the final score, elapsed time, and the count of MI/Non-MI commands, enabling the assessment of real-time motor imagery improvements for a user.
The analysis of results demonstrated that with increased experience, users exhibited better control and improved performance in using the real-time BCI-VR system. For instance, most participants showed longer movement durations in the VR space and achieved higher scores during their third attempt with the BCI-VR system, indicating an enhancement in their real-time motor imagery abilities.
Overall, this research aimed to advance the integration of BCI and VR technologies in the field of neurorehabilitation, contributing to further studies and potential industrial applications.
