Many kinds of wearable robot such as power-assist robots have been proposed.Those robots are useful for the elderly or physically weak persons.To reflect a user motion intention to the motion of the robot,biological signals are used as input signals for the robot.An EMG signals one of the widely used biological signals.However,a person who cannot generate EMG signals is not able to use those robo ts.0n the other hand,in recent years,EEG signals are used as input signals to control the robot.Those robots can be used by persons even if their EMG signals cannot be measured.In this study,EEG signals are used to control a upper-limb power-assist robot.To use EEG signals as input signals for those robots,it is necessary to estimate a user's motion intention from measured EEG signals.In this paper,certain EEG channels are selected and the hand velocity is estimated based on selected EEG signals.