Abstract: At present, offline analysis and research of motor imagery brain computer interface (MIBCI) is relatively mature, but asynchronous online MIBCI is always challenging. Aiming at the recognition rate and control method of online BCI system, this paper proposes the use of common spatial pattern (CSP) algorithm for feature extraction of motor imagery (MI) and asynchronous control with alpha wave. A simple and practical autonomous control ball movement MIBCI experimental system was constructed. Four subjects participated in the online experiment, and two of them had an online motion imaging recognition rate of up to 100%. The experimental results verify the feasibility and practicability of the system built in this paper.