In this paper, a motion recognition method that uses haptic information of the human hand is proposed. The haptic information is measured by master and slave robot hands. The robot has five degrees of freedom, and bilateral control of the robot is implemented. The operator wears the master robot hand and manipulates an object through the slave robot hand. A motion database is prepared, containing haptic information in the form of reference vectors for eight kinds of human motion. The motion database utilizes the cosine similarity to distinguish different human motions on the basis of the haptic information acquired by the master robot. Expansion and contraction of the time axes are corrected by dynamic programming matching, and combination motion is then recognized. The validity of the proposed method is experimentally confirmed. © 2013 Wiley Periodicals, Inc. Electron Comm Jpn, 96(12): 15–23, 2013; Published online in Wiley Online Library (wileyonlinelibrary.com). DOI 10.1002/ecj.11566