The aim of the present study was to propose an effective and low-delayed asynchronous SSVEPs-based BCI system for practical wheelchair control. The paradigm was based on the discrimination of Steady-state visually evoked potential (SSVEP) which is widely applied to various audiences. Bayesian Classifier and a low-delayed asynchronous detection mechanism were devised and integrated to enable the user to control the wheelchair flexibly. In particular, comparing with the traditional method using a fix threshold or a simple classification model to distinguish idle state and task state, our detection mechanism exhibited higher accuracy and possessed a better performance for wheelchair. Five subjects took part in our offline task and two of them continued the on-line task on a real wheelchair. In average, we achieved a classification accuracy of 87.17% in task state and 92.70% in idle state and two subjects accomplished on-line task using 187 s and 298 s, respectively.