By using a brain-computer interface system (BCIs) humans can be enable direct communication with a computer or electronic device. In our previous works, we proved that gaze on the different rotation vanes causes a different effects on EEG signals. This paper proffers a novel BCI system based on this issue. Our BCI system proposes to identify four different rotating vane from EEG Signals that represents commands in a limited visual space. The feature extraction method from the 1-sec epoch of the EEG signal is done by using Discrete Wavelet Transform (DWT). And then MATLAB Classification Learner App is implemented to classify these features. Results of Subspace Discriminant and Quadratic-Support Vector Machine (SVM) were better than other classifiers. Therefore, these classifiers were selected to compare with Partial Least Squares Regression (PLSR). The results show that PLSR is better than other classifiers.