Brain computer interface (BCI) is a communication system to establish a communication interface between the brain and the device. Motor imagery-based brain computer interface is one of asynchronous BCIs, and have been widely developed in recent years. However, most of motor imagery-based BCI require bulk EEG machines, and lots of EEG data have to be transmitted to the back-end computer for analysis. This also causes the limitation of daily applications. In this study, a wearable motor-imagery-based BCI system with the least number of EEG channels was developed. In this system, the wearable mechanical design and the wireless EEG acquisition module were designed for measuring real time EEG in daily life. The modified BCI algorithm was also developed, and here only three channels are used to distinguish different events of motor imagery in this study. Finally, the proposed BCI system was also validated by the hand-motor imagery experiment.