Originally, no additional feature extraction steps were used, but later the Common Spatial Patterns (CSP) algorithm was applied to improve the model's accuracy. However, it was found that CSP was not suitable for domain adjacent classes, such as differentiating between imagined left and right hand movements.