Abstract: This paper explored the feasibility of recognizing motor imagery (MI) and motor execution (ME) in the same motion, as well as the affection of prefrontal cortex on the classification accuracy of MI and ME. We measured changes of oxygenated hemoglobin HbO2 and deoxygenated hemoglobinon (Hb) on prefrontal cortex (PFC) and motor cortex (MC) when 15 subjects performed hand extension and finger tapping tasks. Then mean, slope, quadratic coefficient and approximate entropy features were extracted from HbO2 as the input of support vector machine. With the four-class classifiers of brain-computer interface using functional near-infrared (fNIRS-BCI spectroscopy), 87.65% and 87.58% classification accuracy were realized corresponding to hand extension and finger tapping tasks. The classification accuracy increased significantly after adding PFC fNIRS signal, and greater increase emerged in finger napping than hand extension. In conclusion, it is effective for fNIRS-BCI to recognize MI and ME in the same motion, and the PFC region is sensitive in the four-class fNIRS-BCI classifiers.

Key words: functional near-infrared spectroscopy(fNIRS), brain-computer interface(BCI), motor imagery(MI), motor execution (ME), support vector machine (SVM)