The dangerous body movements(DBM) of lactating sows, such as the change from standing, sitting to lying, is a major cause for the death of pre-weaning piglets. Although, the utilities of farrowing crate and anti-crushing rail will effectively reduce the mortality caused by the DBM of lactating sows, there are still significant differences for the mortality of piglets in the same breeding environment, therefore, the difference maternal ability of individual sow will play an important role. The maternal index of sow was determined by the changing frequencies and types of these DBM. Five lactating sows of Xiao Meishan which were bred in the farm of Jiangsu Polytechnic College of Agriculture and Forestry were used in this study. Firstly, a set of monitoring node with MPU6050 and bluetooth module was tighten on the neck of a sow to collect the three-axis acceleration data during 10 d. Then, the histogram method was used to compute the threshold of the kinetic energy in order to locate the area of a DBM quickly. Finally, a classifier to identify the type of the DBM was built according to the characteristics of the body postures occurred in pre- or pro-DBM. The results showed that the accuracy of for marking DBM area for 327 samples was 77.4%,while the accuracy of DBM recognition was 81.7%. The main reason for the different accuracy mentioned above is that the errors caused by piglet crashing the node were corrected by the recognition algorithm of posture. Thus, the bend index of a lactating sow can be calculated further more and the data basis for the scientific evaluation of maternal ability in the breeding sow will be provided.