Background: The purpose of this research is to design a Brain-Computer Interface to discriminate the brain signals while the brain images four main directions. To be innovative, the subjects have imaged the aimed directions by power of imagination, and for the first time, the ICA algorithm has been used to detect the aimed signal and to eliminate the artifacts.

Materials and Methods: In this descriptive-ana alytic study, signals are recorded by using a Micromed device and a 19-channel helmet in unipolar mode. The statistical population included three persons in the age range of 25 to 30 and the designed task consisted of 24 slides of four main directions.

Results: Simulations have shown that the best classification accuracy was the outcome of the 2.5-second time windowing and the best choice for extracting features was the AR coefficients of 15 order. There was no significant difference between the classification accuracy of different implementation of the Artificial Neural Network classifier with different number of layers and neurons and different classification functions. In comparison with the Neural Network, the Linear Discriminant Analysis (LDA) showed better classification accuracies.

Conclusion: The results of this research are in accordance with the results of the methods such as FMRI and methods based on the brain signals in vowel imagination. In this research, the best classification accuracy was obtained from the Linear Discriminant Analysis classifier by extracting the target signal from the output of the ICA algorithm and extracting the AR coefficients as feature and the 2.5-second time windowing. The Linear Discriminant Analysis classifier result the best classification accuracies.