In the human-robot interaction, teaching robots to learn human skills is a challenging work. Throwing is a process of projecting an object to a distant target location, and it is usually considered as one of the skills possessed by humans. To enable robots to perform throwing skills, two steps are required: firstly, modeling and outputting the motion sequences of human throwing motion; secondly, transferring the motion sequences to the robotic arm and realizing the throwing skill on the robotic arm. This paper aims to model the motion of human throwing. First of all, raw datas of human throwing motion are obtained through human demonstration. Afterwards, feature vectors are extracted through the analysis of the raw datas. By applying the HMM to train the feature vectors, the model of human throwing motion \({\lambda }_{final}\) has ultimately been finalized. The result shows that our model can effectively describe the behavior of human throwing and generate the motion sequences of human throwing motion.