The paper proposes a new method to control external device in real-time using electroencephalography-based brain signals. The brain signals of a healthy female student at F7 and F8 channels are recorded from Emotiv Epoc+. They are filtered using a combination of wavelet approach and recursive least square estimation to remove unwanted noises. Open and closed eyes states are extracted from filtered brain signals. The support vector machine approach is applied to classify two states of eyes (open and closed). The classified eyes states are utilized to generate the on and off commands, respectively. Those commands are sent to an Arduino control board to control on and off states of the light. Experimental results showed that the average accuracy of two control commands is 81.6%. The obtained results promise for extraction of more commands that can be utilized for applications in daily life.