Stroke rehabilitation plays a vital role in restoring the normal lives of stroke survivors who have impairments due to the stroke. Rehabilitation robots are emerging to replace their traditional counterparts and show promise in providing efficient rehabilitation training. Bio-signal modalities such as electromyography, electroencephalography, or near-infrared spectroscopy, when combined with rehabilitation robots, use the motion intention of their wearers to enhance the effect of rehabilitation therapy. However, many technological advancements are needed in this integration, particularly in understanding the information contained in these signals relevant to robotic control. This paper presents a brain-computer interface that can estimate human motion intention with higher accuracy and improved temporal performance for a hand neuro-rehabilitation robot, SMOVE.