| Abstrakt: |
An interface between a human brain and a computer (or any external device) can be implemented for interchanging orders using a brain–computer interface (BCI) system. Motor imagery (MI), which represents human intention to execute actions or movements, can be captured and analyzed using brain signals such as electroencephalograms (EEGs). The present study focuses on a synchronous control system with a BCI based on MI for robot navigation. We employ a new feature extraction technique using common spatial pattern (CSP) filtering combined with band power to form feature vectors. Linear discriminant analysis (LDA) is employed to classify two types of MI tasks (right hand and left hand). In addition, we have developed posture-dependent control architecture that translates the obtained MI into four robot motion commands: going forward, turning left, turning right, and stopping. The EEGs of eight healthy volunteer male subjects were recorded and employed to navigate a simulated robot to a goal in a virtual environment. On a predefined task, the developed BCI robot control system achieved its task in170 s with a collision number of 0.65, distance of 23.92 m, and successful command rate of 80%. Although the performance of the complete system varied from one subject to another, the robot always reached its final position successfully. The developed BCI robot control system yields promising results compared to manual controls. |
Full text from SpringerLink