The highly accurate BCI using alpha waves was developed for controlling the robot arm, and real-time operation was succeeded by using noninvasive electrodes. The significant components of the alpha wave were identified by spectral analysis and confirmation of the amplitude of the alpha wave. When the alpha wave was observed in the subject, the subjects were instructed to select the multiple decision branches, concerning 7 motions (including "STOP") of a robot arm. As a result, high accuracy (70-95%) was obtained, and the subject succeeded in transferring a small box by controlling the robot arm. Since high accuracy was obtained by use of this method, it can be applied to control equipments such as a robot arm. Since the alpha wave can be easily generated, the BCI using alpha waves does not need more training than that using other signals. Moreover, we tried to reduce the false positive errors by effectively detecting artifacts using spectral analysis and detecting signals of 50 µV or more. As a result, the false positive errors could be reduced from 25% to 0%. Therefore, this technique shows great promise in the area of communication and the control of other external equipments, and will make great contribution in the improvement of Quality of Life (QOL) of mobility disabled.