Symbolic representation of imitation learning is a convenient and feasible way for fri-co(coexisting- cooperative-cognitive) robots to improve their intelligence, and it also provides a practical and feasible solution to solve the learning problems of complex and multi-steps tasks. The automatic segmentation of the teaching trajectory and the acquisition of its movement primitives are the prerequisition of the successful application of this learning method. In view of this, this paper firstly analyzes the basic requirements of the automatic segmentation mehods based on the introduction of imitation learning in symbolic representation. Then, it classifies them into two categories according to the prior knowledge of teaching tasks and introduces the typical segmentation methods of each category in detail. Finally, it analyzed these two categories comparatively and proposes the prospect of the development direction of the automatic segmentation method.