This paper presents the method of using facial expressions based on brain waves to control robot arms. In this study, two facial expressions including raising eyebrows and closing eyes are set to move the robotic arms to the left and right respectively. Facial expressions are recorded by Emotive Epoc device in the form of brain waves then converted into a typical model through a training process with the Emotive Control Epoc software. This model is also used to classify facial expressions into three types including natural faces, closed eyes and raised eyebrows. The results of this classification process are converted and output the coresponding keyboard values. In the case of facial expressions, for example, defected as closed eyes or raised eyebrows, the system will output characters respectively "a" or "b" sending commands to move the robot arms to the left or right. If the facial expression is natural, a command which signals to stop all the movements of the robot arms is operative. Experimental results prove that the average accuracy rate of these two commands is 82.5%.