In human-robot interaction, human hand gesture is an important way to deliver message or teach by demonstration, and it also indicates the content of the safety. Good performance in human motion recognition can not only provide robots with correct decisions but also avoid the potential safety hazards. In this paper, Empirical Copula is introduced for recognizing human motions for the first time using the proposed motion template and matching algorithm. The huge computational cost of Empirical Copula is alleviated by the proposed re-sampling processing. The experiments with human hand motions including grasps and in-hand manipulations demonstrate Empirical Copula outperforms the Time Clustering (TC), Gaussian Mixture Models (GMMs) and Hidden Markov Models (HMMs) in terms of recognition rate.