In this paper, we propose an online motor imagery detect algorithm including Electroencephalography (EEG) preprocessing, common spatial patterns feature extraction?, SVM classification, and threshold mechanism. In designing the light wireless switch hardware module, we develop the light switch system controlled by EEG. Participants can use their brain waves to remote control the light turn on, turn off , which especially provide a new direct communication channel for patients suffering from Amyotrophic Lateral Sclerosis(ALS) with the nature. Five healthy subjects participated in the training experiment and online experiment, the experimental results show that: after a specific training procedure, participants can effectively control the light switch system with an average accuracy of 90% and an average time of 4 seconds for each single command.