This paper presents an exoskeleton for the assist of forearm motion (elbow flexion-extension and forearm pronation-supination motion) in daily activity and rehabilitation. The exoskeleton is controlled based on the activation patterns of the electromyogram (EMG) signals of the patient's muscles, which directly reflects the motion intention of the patient, in order to realize natural motion assist. The sophisticated real-time neuro-fuzzy control method, in which the effect of a muscle common to both motions is taken into account, is proposed. The proposed control method enables the cooperative motion of elbow and forearm of the patient by learning the muscle activation patterns of each patient. The effectiveness of the proposed method was evaluated by experiment.