Considering the limitations of stimulator redundancy and user-unfriendliness in the mainstream brain control paradigm, a facial expression assisted brain control method had been proposed by our research group for the realtime and precise needs of peripheral control. Focused on further research, this paper proposed an updated version with semi-asynchronous strategy and up to eight instruction sets, in which EEGs with 100 ms window length were selected in realizing the real-time decoding. An algorithm based on ‘one vs one’ CSP combined with SVM was applied in processing signals. By taking Normal as the idle state, combined with other eight assisting facial expressions, the semi-asynchronous detection and its real-time output were realized by the classifier for nine targets. Finally, the research results show the accuracy rate of semi-asynchronous real-time processing is 88.8% with eight instructions and 92.8% with abridged six instructions.