Problems of dyskinesia and cognitive impairment for the elderly are very common. To solve these problems，an easy-to-use and feature-rich service robot for the elderly was designed. To Reduce the operation difficulty of the service robot for the elderly, and optimize the user’s experience, the human-like motion control for the manipulator was achieved by establishing the spatial joint mapping model between human arm and manipulator. Firstly, according to the structure and degree of freedom distribution of human arm and manipulator, the human-machine joint mapping model was initially established. The working space of human arm and manipulator was simulated and compared using the Monte Carlo method. Because of the obvious difference in the spatial structure and degree of freedom distribution between human arm and manipulator, the working space of manipulator was too little, and the manipulator couldn’t imitate the movement of human arm. So, a corresponding improved scheme was proposed, and the human-machine joint mapping model was reestablished. The spatial pose information of the human joints was collected using the Kinect visual sensor， and the human coordinate system with the human left shoulder joint as the origin was established. The spatial vector method and the inverse kinematics method of human arm based on elbow constraint were used to solve the pose changes of human joints. Finally, an experimental platform for service robot for the elderly was established, and the validity and rationality of the improved human-machine joint mapping model and related control algorithm were verified. The experimental results showed that the manipulator could highly revert to the human arm's motion after joint remapping. The research results can be used for reference to improve the robot's interactive ability, and have a significant effect on reducing the robot operation difficulty, which also lays a foundation for the follow-up research on more complex human-like motion robot.