Life support robots are expected to reduce the physical burden on caregivers. By synchronizing the physical rhythm of the caregiver and assisted person, the caregiver gives the assisted person a sense of safety and security. It has also been reported that physical rhythm is also important in embodied interaction between humans and robots. Therefore, we propose a human-like sit-up motion model based on the analysis of human movements. In the experiment, we measure the sit-up motion of the caregiver and assisted person. Optical and magnetic motion capture and microphones were used during the measurements. Then we formulate the neck position and the shoulder posture of the assisted person. During the formulation, two types of waveform patterns were observed. Therefore, the waveform patterns were classified as mountainous-shaped and s-shaped, and applied to the motion model. In addition, we verified that the upper limb movements by the proposed model are similar to those of a caregiver.