Strokes are a leading cause of adult disability that often induces motor functional deficits. Synchronous activation of neurons along motor and sensory pathway is thought to facilitate synaptic reconnection. We developed an electroencephalogram (EEG)-based brain computer interface (BCI) system to control the magnitude of functional electrical stimulation (FES) for an affected tibialis anterior muscle of a stroke patient. The proposed EEG-FES system used event related desynchronization (ERD) of motor intention, which is extracted from an EEG electrode located on the lower limb region of primary motor area. In this paper, we will report the training effect of the EEG-FES system on the motor function of affected lower limb of a severely impaired patient. After 20 minutes training, the range of motion of an ankle joint and the electromyography (EMG) amplitude of the tibialis anterior muscle of a patient's paralyzed lower limb were significantly enlarged. The results suggest that the EEG-FES system has the potential to improve the motor function of severely paralyzed muscles.