Abstract: Sport data based quantitative assessment of actions is the basis of the development of intelligent sport science. However， conventional action quantitative assessment methods usually match the pattern or signal of an action with its standard version， and therefore cannot grasp the essential attributes and cannot assess the action in a refined and comprehensive way. In order to assess an action from its structure and essential attributes， we firstly implement the SWAB （Sliding Windows and Bottom-up） curve segmentation algorithm to perform a fine-grained segmentation on the process of the action such that the action is decomposed into a number of posture fragments. We then assess each posture fragment from the viewpoints of standard degree， speed， and integrity， and finally obtain the comprehensive assessment score of the action. To deal with the difficulty in matching the key points of two posture fragments in the above assessment， we transform this point-matching problem into a dynamic programming problem and obtain the optimal matching manner by solving it with the conventional optimization algorithm. It is demonstrated by the experimental results on a sensor based badminton action dataset that our proposed comprehensive action assessment method with posture segmentation provides more fine-grained comprehensive assessment of sport actions， and can effectively guide players to improve their skills.