In this paper, a functional interface between brain and central pattern generator (CPG) is designed in an engineering perspective, which may serve in a human-machine system. Steady-state visual evoked potential (SSVEP) based brain-computer interface (BCI) is used to recognize five types of intention related to human walking. After feature extraction, classification and command translation on electroencephalography (EEG) signals, the human intention can control CPG to generate desired motor patterns for walking. Four subjects take part in BCI experiment, the successful classification accuracy are above 80%. Also the CPG model can accomplish the desired changes using online EEG data.