Abstract: Grasp planning is one of the key issues of the dexterous hand research. At present master-slave grasp is a feasible solution. The core problem is to map the human grasp into that of the dexterous hands reasonably. Aiming at the anthropomorphic dexterous hands, the method of mapping human hand motion to the robot hand was investigated. The mapping model of master-slave motion was established in the Cartesian space, and the mapping matrix was formulated and simplified considering the similarity of the kinematics of the human and dexterous hands. Through the linear scalar mapping the human hand fingers can control the corresponding fingertip motions of the robot hand independently. The mapping results were analyzed by the experiments, and the fundamental requirements for mapping were discussed. The experimental results illustrate that the mapped fingertip positions of the dexterous hands are basically within their workspaces thus ensure the master-slave motion feasible, and the motion locus are similar thus ensure the master-slave control simple and intuitive.